Innodb 1.1 for mysql 5.5 user's guide
0 likes2,153 views
InnoDB 1.1 is the latest version of the InnoDB storage engine plugin for MySQL 5.5 that provides performance and scalability improvements over previous versions. It includes features from the InnoDB Plugin 1.0 for MySQL 5.1 as well as new features for MySQL 5.5. InnoDB 1.1 supports data compression, fast index creation, and provides additional INFORMATION_SCHEMA tables for monitoring transactions and compression statistics. However, it is not recommended for production use and the generally available version is InnoDB Plugin 1.0.7 and higher.
![InnoDB File Format Management
The file format tag is a “high water mark”, and as such it is increased after the server is started, if a table in a “higher” format is
created or an existing table is accessed for read or write (assuming its format is supported). If you access an existing table in a
format higher than the format the running software supports, the system tablespace tag is not updated, but table-level compatibility
checking applies (and an error is issued), as described in Section 4.2.2, “Compatibility Check When a Table Is Opened”. Any time
the high water mark is updated, the value of innodb_file_format_check is updated as well, so the command SELECT
@@innodb_file_format_check; displays the name of the newest file format known to be used by tables in the currently
open ib-file set and supported by the currently executing software.
To best illustrate this behavior, consider the scenario described in Table 4.1, “InnoDB Data File Compatibility and Related InnoDB
Parameters”. Imagine that some future version of InnoDB supports the “Cheetah” format and that an ib-file set has been used with
that version.
Table 4.1. InnoDB Data File Compatibility and Related InnoDB Parameters
innodb innodb file Highest file Highest file Result
file format format used format sup-
format in ib-file set ported by In-
check noDB
OFF Antelope or Barracuda Barracuda Database can be opened; tables can be created which require
Barracuda “Antelope” or “Barracuda” file format
OFF Antelope or Cheetah Barracuda Database can be opened with a warning, since the database con-
Barracuda tains files in a “too new” format; tables can be created in
“Antelope” or “Barracuda” file format; tables in “Cheetah” format
cannot be accessed
OFF Cheetah Barracuda Barracuda Database cannot be opened; innodb_file_format cannot be
set to “Cheetah”
ON Antelope or Barracuda Barracuda Database can be opened; tables can be created in “Antelope” or
Barracuda “Barracuda” file format
ON Antelope or Cheetah Barracuda Database cannot be opened, since the database contains files in a
Barracuda “too new” format (“Cheetah”)
ON Cheetah Barracuda Barracuda Database cannot be opened; innodb_file_format cannot be
set to “Cheetah”
4.2.2. Compatibility Check When a Table Is Opened
When a table is first accessed, InnoDB (including some releases prior to InnoDB 1.0) checks that the file format of the tablespace
in which the table is stored is fully supported. This check prevents crashes or corruptions that would otherwise occur when tables
using a “too new” data structure are encountered.
All tables using any file format supported by a release can be read or written (assuming the user has sufficient privileges). The set-
ting of the system configuration parameter innodb_file_format can prevent creating a new table that uses specific file
formats, even if they are supported by a given release. Such a setting might be used to preserve backward compatibility, but it does
not prevent accessing any table that uses any supported format.
As noted in Chapter 4, InnoDB File Format Management [13], versions of MySQL older than 5.0.21 cannot reliably use database
files created by newer versions if a new file format was used when a table was created. To prevent various error conditions or cor-
ruptions, InnoDB checks file format compatibility when it opens a file (for example, upon first access to a table). If the currently
running version of InnoDB does not support the file format identified by the table type in the InnoDB data dictionary, MySQL re-
ports the following error:
ERROR 1146 (42S02): Table 'test.t1' doesn't exist
InnoDB also writes a message to the error log:
InnoDB: table test/t1: unknown table type 33
The table type should be equal to the tablespace flags, which contains the file format version as discussed in Section 4.3,
“Identifying the File Format in Use”.
Versions of InnoDB prior to MySQL 4.1 did not include table format identifiers in the database files, and versions prior to MySQL
5.0.21 did not include a table format compatibility check. Therefore, there is no way to ensure proper operations if a table in a “too
new” format is used with versions of InnoDB prior to 5.0.21.
The file format management capability in InnoDB 1.0 and higher (tablespace tagging and run-time checks) allows InnoDB to verify
15](https://image.slidesharecdn.com/innodb1-1formysql5-5usersguide-120515213218-phpapp02/85/Innodb-1-1-for-mysql-5-5-user-s-guide-23-320.jpg)
![Performance and Scalability Enhancements
The default value of innodb_spin_wait_delay is 6. The spin wait delay is a dynamic, global parameter that can be specified
in the MySQL option file (my.cnf or my.ini) or changed at runtime with the command SET GLOBAL in-
nodb_spin_wait_delay=delay, where delay is the desired maximum delay. Changing the setting requires the SUPER
privilege.
For performance considerations for InnoDB locking operations, see Optimizing Locking Operations.
7.15. Making Buffer Cache Scan Resistant
Rather than using a strictly LRU algorithm, InnoDB uses a technique to minimize the amount of data that is brought into the buffer
cache and never accessed again. The goal is to make sure that frequently accessed (“hot”) pages remain in the buffer cache, even as
read-ahead and full table scans bring new blocks in that might or might not be accessed afterward.
Newly read blocks are inserted into the middle of the list representing the buffer cache. of the LRU list. All newly read pages are
inserted at a location that by default is 3/8 from the tail of the LRU list. The pages are moved to the front of the list (the most-
recently used end) when they are accessed in the buffer cache for the first time. Thus pages that are never accessed never make it to
the front portion of the LRU list, and “age out” sooner than with a strict LRU approach. This arrangement divides the LRU list into
two segments, where the pages downstream of the insertion point are considered “old” and are desirable victims for LRU eviction.
For an explanation of the inner workings of the InnoDB buffer pool and the specifics of its LRU replacement algorithm, see The
InnoDB Buffer Pool.
Starting with InnoDB 1.0.5, you can control the insertion point in the LRU list, and choose whether InnoDB applies the same op-
timization to blocks brought into the buffer pool by table or index scans.
The configuration parameter innodb_old_blocks_pct controls the percentage of “old” blocks in the LRU list. The default
value of innodb_old_blocks_pct is 37, corresponding to the original fixed ratio of 3/8. The value range is 5 (new pages in
the buffer pool age out very quickly) to 95 (only 5% of the buffer pool reserved for hot pages, making the algorithm close to the fa-
miliar LRU strategy).
The optimization that keeps the buffer cache from being churned by read-ahead can avoid similar problems due to table or index
scans. In these scans, a data page is typically accessed a few times in quick succession and is never touched again. The configura-
tion parameter innodb_old_blocks_time specifies the time window (in milliseconds) after the first access to a page during
which it can be accessed without being moved to the front (most-recently used end) of the LRU list. The default value of in-
nodb_old_blocks_time is 0, corresponding to the original behavior of moving a page to the most-recently used end of the
buffer pool list when it is first accessed in the buffer pool. Increasing this value makes more and more blocks likely to age out
faster from the buffer pool.
Both the new parameters innodb_old_blocks_pct and innodb_old_blocks_time are dynamic, global and can be spe-
cified in the MySQL option file (my.cnf or my.ini) or changed at runtime with the SET GLOBAL command. Changing the set-
ting requires the SUPER privilege.
To help you gauge the effect of setting these parameters, some additional statistics are reported by SHOW ENGINE INNODB
STATUS command. The BUFFER POOL AND MEMORY section now looks like:
Total memory allocated 1107296256; in additional pool allocated 0
Dictionary memory allocated 80360
Buffer pool size 65535
Free buffers 0
Database pages 63920
Old database pages 23600
Modified db pages 34969
Pending reads 32
Pending writes: LRU 0, flush list 0, single page 0
Pages made young 414946, not young 2930673
1274.75 youngs/s, 16521.90 non-youngs/s
Pages read 486005, created 3178, written 160585
2132.37 reads/s, 3.40 creates/s, 323.74 writes/s
Buffer pool hit rate 950 / 1000, young-making rate 30 / 1000 not 392 / 1000
Pages read ahead 1510.10/s, evicted without access 0.00/s
LRU len: 63920, unzip_LRU len: 0
I/O sum[43690]:cur[221], unzip sum[0]:cur[0]
• Old database pages is the number of pages in the “old” segment of the LRU list.
• Pages made young and not young is the total number of “old” pages that have been made young or not respectively.
• youngs/s and non-young/s is the rate at which page accesses to the “old” pages have resulted in making such pages
young or otherwise respectively since the last invocation of the command.
• young-making rate and not provides the same rate but in terms of overall buffer cache accesses instead of accesses just
to the “old” pages.
31](https://image.slidesharecdn.com/innodb1-1formysql5-5usersguide-120515213218-phpapp02/85/Innodb-1-1-for-mysql-5-5-user-s-guide-39-320.jpg)
![Changes for Flexibility, Ease of Use and Reliability
idates input syntax and data values. Since strict mode is relatively new, some statements that execute without errors with earlier
versions of InnoDB might generate errors unless you disable strict mode.
The setting of InnoDB strict mode affects the handling of syntax errors on the CREATE TABLE, ALTER TABLE and CREATE
INDEX statements. Starting with InnoDB 1.0.2, the strict mode also enables a record size check, so that an INSERT or UPDATE
never fails due to the record being too large for the selected page size.
Using the clauses and settings for ROW_FORMAT and KEY_BLOCK_SIZE on CREATE TABLE, ALTER TABLE, and CREATE
INDEX statements can be confusing when not running in strict mode. Unless you run in strict mode, InnoDB ignores certain syntax
errors and creates the table or index, with only a warning in the message log. However, if InnoDB strict mode is on, such problems
generate an immediate error and the table or index is not created, thus saving time by catching the error at the time the statement is
issued.
InnoDB strict mode is set with the configuration parameter innodb_strict_mode, which can be specified as on or off. You
can set the value on the command line when you start mysqld, or in the configuration file my.cnf (Unix operating systems) or
my.ini (Windows). You can also enable or disable InnoDB strict mode at runtime with the statement SET
[GLOBAL|SESSION] innodb_strict_mode=mode, where mode is either ON or OFF. Changing the GLOBAL setting re-
quires the SUPER privilege and affects the operation of all clients that subsequently connect. Any client can change the SESSION
setting for innodb_strict_mode, which affects only that client.
8.5. Controlling Optimizer Statistics Estimation
The MySQL query optimizer uses estimated statistics about key distributions to select or avoid using an index in an execution plan,
based on the relative selectivity of the index. Previously, InnoDB sampled 8 random pages from an index to get an estimate of the
cardinality of (i.e., the number of distinct values in) the index. (This page sampling technique is frequently described as “index
dives”.) This small number of page samples frequently was insufficient, and could give inaccurate estimates of an index's selectiv-
ity and thus lead to poor choices by the query optimizer.
To give users control over the quality of the statistics estimate (and thus better information for the query optimizer), the number of
sampled pages now can be changed using the parameter innodb_stats_sample_pages.
This feature addresses user requests such as that as expressed in MySQL Bug#25640: InnoDB Analyze Table Should Allow User
Selection of Index Dives.
You can change the number of sampled pages using the global parameter innodb_stats_sample_pages, which can be set at
runtime. The default value for this parameter is 8, preserving the same behavior as in past releases.
Note that the value of innodb_stats_sample_pages affects the index sampling for all tables and indexes. Note that there
are the following potentially significant impacts when you change the index sample size:
• Small values like 1 or 2 can result in very inaccurate estimates of cardinality.
• Values much larger than 8 (say, 100), can cause a big slowdown in the time it takes to open a table or execute SHOW TABLE
STATUS.
• The optimizer might choose very different query plans based on different estimates of index selectivity.
The cardinality estimation can be disabled for metadata statements such as SHOW TABLE STATUS by executing the statement
SET GLOBAL innodb_stats_on_metadata=OFF (or 0). Before InnoDB 1.0.2, this variable could only be set in the
MySQL option file (my.cnf or my.ini), and changing it required shutting down and restarting the server.
The cardinality (the number of different key values) in every index of a table is calculated when a table is opened, at SHOW TABLE
STATUS and ANALYZE TABLE and in other circumstances such as when the table has changed significantly. Note that all tables
are opened, and the statistics are re-estimated, when the mysql client starts if the auto-rehash setting is set on (the default). The
auto-rehash feature enables automatic name completion of database, table, and column names for interactive users. You can turn
auto-rehash off to improve the start up time of the mysql client.
Because the statistics are automatically recalculated at various times other than on execution of ANALYZE TABLE, it does not
make sense to increase the index sample size, then run ANALYZE TABLE and decrease sample size to attempt to obtain better stat-
istics. The “better” statistics calculated by ANALYZE running with a high value of innodb_stats_sample_pages will be
wiped away.
The estimated cardinality for an index is more accurate with a larger number of samples, but each sample might require a disk read,
so you do not want to make the sample size too large. Choose a value for innodb_stats_sample_pages that results in reas-
onably accurate estimates for all tables in your database without requiring excessive I/O.
Although it is not possible to specify the sample size on a per-table basis, smaller tables generally require fewer index samples than
larger tables do. If your database has many large tables, consider using a higher value for innodb_stats_sample_pages
than if you have mostly smaller tables.
37](https://image.slidesharecdn.com/innodb1-1formysql5-5usersguide-120515213218-phpapp02/85/Innodb-1-1-for-mysql-5-5-user-s-guide-45-320.jpg)
![InnoDB Glossary
See also Barracuda.
See also buffer pool.
See also compressed row format.
See also hot backup.
See also index.
See also plugin.
See also table.
See also undo buffer.
concatenated index See composite index.
concurrency The ability of multiple operations (in database terminology, transactions) to run simultaneously,
without interfering with each other. Concurrency is also involved with performance, because
ideally the protection for multiple simultaneous transactions works with a minimum of perform-
ance overhead, using efficient mechanisms for locking.
See also ACID.
See also locking.
See also transaction.
configuration file The file that holds the option values used by MySQL at startup. Traditionally, on Linux and
UNIX this file is named my.cnf, and on Windows it is named my.ini. You can set a number
of options related to InnoDB under the [mysqld] section of the file.
When you use the MySQL Enterprise Backup product, you typically use two configuration
files: one that specifies where the data comes from and how it is structured (which could be the
original configuration file for your real server), and a stripped-down one containing only a small
set of options that specify where the backup data goes and how it is structured. The configuration
files used with the MySQL Enterprise Backup product must contain certain options that are
typically left out of regular configuration files, so you might need to add some options to your ex-
isting configuration file for use with MySQL Enterprise Backup.
See also my.cnf.
See also option file.
consistent read A read operation that uses snapshot information to present query results based on a point in time,
regardless of changes performed by other transactions running at the same time. If queried data
has been changed by another transaction, the original data is reconstructed based on the contents
of the undo log. This technique avoids some of the locking issues that can reduce concurrency
by forcing transactions to wait for other transactions to finish.
With the repeatable read isolation level, the snapshot is based on the time when the first read
operation is performed. With the read committed isolation level, the snapshot is reset to the time
of each consistent read operation.
Consistent read is the default mode in which InnoDB processes SELECT statements in READ
COMMITTED and REPEATABLE READ isolation levels. Because a consistent read does not
set any locks on the tables it accesses, other sessions are free to modify those tables while a con-
sistent read is being performed on the table.
See also ACID.
See also concurrency.
See also isolation level.
See also locking.
See also MVCC.
59](https://image.slidesharecdn.com/innodb1-1formysql5-5usersguide-120515213218-phpapp02/85/Innodb-1-1-for-mysql-5-5-user-s-guide-67-320.jpg)
Innodb 1.1 for mysql 5.5 user's guide
- 1. InnoDB 1.1 for MySQL 5.5 User’s Guide
- 2. InnoDB 1.1 for MySQL 5.5 User’s Guide This is the User’s Guide for the InnoDB storage engine 1.1 for MySQL 5.5, generated on 2011-01-24 (revision: 24873) . Abstract Beginning with MySQL version 5.1, it is possible to swap out one version of the InnoDB storage engine and use an- other (the “plugin”). This manual documents the latest InnoDB plugin, version 1.1, which works with MySQL 5.5 and features cutting-edge improvements in performance and scalability. This User's Guide documents the procedures and features that are specific to the InnoDB storage engine 1.1 for MySQL 5.5. It supplements the general InnoDB information in the MySQL Reference Manual. InnoDB 1.1 is recommended for performance, functionality, and compatibility testing, but not yet for deployment in mission-critical situations. The generally available (GA) version of the InnoDB Plugin is 1.0.7 and higher, for MySQL 5.1. WARNING: Because the InnoDB storage engine 1.0 and above introduces a new file format, restrictions apply to the use of a database created with the InnoDB storage engine 1.0 and above, with earlier versions of InnoDB, when using mysqldump or MySQL replication and if you use the older InnoDB Hot Backup product rather than the newer MySQL Enterprise Backup product. See Section 1.4, “Compatibility Considerations for Down- grade and Backup”. Copyright © 1997, 2011, Oracle and/or its affiliates. All rights reserved. This software and related documentation are provided under a license agreement containing restrictions on use and disclosure and are protected by intellectual property laws. Except as expressly permitted in your license agreement or allowed by law, you may not use, copy, reproduce, translate, broadcast, modify, license, transmit, distribute, exhibit, perform, publish, or display any part, in any form, or by any means. Reverse engineering, disassembly, or decompilation of this software, unless required by law for interoperability, is prohibited. The information contained herein is subject to change without notice and is not warranted to be error-free. If you find any errors, please report them to us in writing. If this software or related documentation is delivered to the U.S. Government or anyone licensing it on behalf of the U.S. Government, the follow- ing notice is applicable: U.S. GOVERNMENT RIGHTS Programs, software, databases, and related documentation and technical data delivered to U.S. Government cus- tomers are "commercial computer software" or "commercial technical data" pursuant to the applicable Federal Acquisition Regulation and agency- specific supplemental regulations. As such, the use, duplication, disclosure, modification, and adaptation shall be subject to the restrictions and li- cense terms set forth in the applicable Government contract, and, to the extent applicable by the terms of the Government contract, the additional rights set forth in FAR 52.227-19, Commercial Computer Software License (December 2007). Oracle USA, Inc., 500 Oracle Parkway, Redwood City, CA 94065. This software is developed for general use in a variety of information management applications. It is not developed or intended for use in any inher- ently dangerous applications, including applications which may create a risk of personal injury. If you use this software in dangerous applications, then you shall be responsible to take all appropriate fail-safe, backup, redundancy, and other measures to ensure the safe use of this software. Oracle Corporation and its affiliates disclaim any liability for any damages caused by use of this software in dangerous applications. Oracle is a registered trademark of Oracle Corporation and/or its affiliates. MySQL is a trademark of Oracle Corporation and/or its affiliates, and shall not be used without Oracle's express written authorization. Other names may be trademarks of their respective owners. This software and documentation may provide access to or information on content, products, and services from third parties. Oracle Corporation and its affiliates are not responsible for and expressly disclaim all warranties of any kind with respect to third-party content, products, and services. Oracle Corporation and its affiliates will not be responsible for any loss, costs, or damages incurred due to your access to or use of third-party con- tent, products, or services. This document in any form, software or printed matter, contains proprietary information that is the exclusive property of Oracle. Your access to and use of this material is subject to the terms and conditions of your Oracle Software License and Service Agreement, which has been executed and with which you agree to comply. This document and information contained herein may not be disclosed, copied, reproduced, or distributed to any- one outside Oracle without prior written consent of Oracle or as specifically provided below. This document is not part of your license agreement nor can it be incorporated into any contractual agreement with Oracle or its subsidiaries or affiliates. This documentation is NOT distributed under a GPL license. Use of this documentation is subject to the following terms: You may create a printed copy of this documentation solely for your own personal use. Conversion to other formats is allowed as long as the actual content is not altered or edited in any way. You shall not publish or distribute this documentation in any form or on any media, except if you distrib- ute the documentation in a manner similar to how Oracle disseminates it (that is, electronically for download on a Web site with the software) or on a CD-ROM or similar medium, provided however that the documentation is disseminated together with the software on the same medium. Any oth- er use, such as any dissemination of printed copies or use of this documentation, in whole or in part, in another publication, requires the prior writ- ten consent from an authorized representative of Oracle. Oracle and/or its affiliates reserve any and all rights to this documentation not expressly granted above. For more information on the terms of this license, for details on how the MySQL documentation is built and produced, or if you are interested in doing a translation, please visit MySQL Contact & Questions.
- 3. For additional licensing information, including licenses for third-party libraries used by MySQL products, see Preface and Notes. If you want help with using MySQL, please visit either the MySQL Forums or MySQL Mailing Lists where you can discuss your issues with other MySQL users. For additional documentation on MySQL products, including translations of the documentation into other languages, and downloadable versions in variety of formats, including HTML and PDF formats, see the MySQL Documentation Library.
- 5. Table of Contents 1. Introduction to InnoDB 1.1 ...............................................................................................................1 1.1. Features of the InnoDB Storage Engine ......................................................................................1 1.2. Obtaining and Installing the InnoDB Storage Engine ......................................................................1 1.3. Viewing the InnoDB Storage Engine Version Number ....................................................................1 1.4. Compatibility Considerations for Downgrade and Backup ................................................................2 2. Fast Index Creation in the InnoDB Storage Engine ...................................................................................3 2.1. Overview of Fast Index Creation ..............................................................................................3 2.2. Examples of Fast Index Creation ..............................................................................................3 2.3. Implementation Details of Fast Index Creation ..............................................................................3 2.4. Concurrency Considerations for Fast Index Creation .......................................................................4 2.5. How Crash Recovery Works with Fast Index Creation ....................................................................4 2.6. Limitations of Fast Index Creation ............................................................................................4 3. InnoDB Data Compression ...............................................................................................................6 3.1. Overview of Table Compression ..............................................................................................6 3.2. Enabling Compression for a Table ............................................................................................6 3.2.1. Configuration Parameters for Compression ........................................................................6 3.2.2. SQL Compression Syntax Warnings and Errors ..................................................................7 3.3. Tuning InnoDB Compression ..................................................................................................8 3.3.1. When to Use Compression ...........................................................................................8 3.3.2. Monitoring Compression at Runtime ............................................................................. 10 3.4. How Compression Works in InnoDB ....................................................................................... 10 3.4.1. Compression Algorithms ........................................................................................... 10 3.4.2. InnoDB Data Storage and Compression .......................................................................... 11 3.4.3. Compression and the InnoDB Buffer Pool ....................................................................... 12 3.4.4. Compression and the InnoDB Log Files ......................................................................... 12 4. InnoDB File Format Management ..................................................................................................... 13 4.1. Enabling File Formats ......................................................................................................... 13 4.2. File Format Compatibility .................................................................................................... 13 4.2.1. Compatibility Check When InnoDB Is Started .................................................................. 14 4.2.2. Compatibility Check When a Table Is Opened .................................................................. 15 4.3. Identifying the File Format in Use .......................................................................................... 16 4.4. Downgrading the File Format ................................................................................................ 16 4.5. Future InnoDB File Formats ................................................................................................. 16 5. Storage of Variable-Length Columns ................................................................................................. 17 5.1. Overview ........................................................................................................................ 17 5.2. COMPACT and REDUNDANT Row Format ................................................................................. 17 5.3. DYNAMIC Row Format ....................................................................................................... 17 5.4. Specifying a Table’s Row Format ........................................................................................... 17 6. InnoDB INFORMATION_SCHEMA tables ........................................................................................... 18 6.1. Information Schema Tables about Compression .......................................................................... 18 6.1.1. INNODB_CMP and INNODB_CMP_RESET ..................................................................... 18 6.1.2. INNODB_CMPMEM and INNODB_CMPMEM_RESET .......................................................... 18 6.1.3. Using the Compression Information Schema Tables ........................................................... 18 6.2. Information Schema Tables about Transactions ........................................................................... 19 6.2.1. INNODB_TRX ....................................................................................................... 19 6.2.2. INNODB_LOCKS .................................................................................................... 19 6.2.3. INNODB_LOCK_WAITS ........................................................................................... 19 6.2.4. Using the Transaction Information Schema Tables ............................................................. 20 6.3. Notes on Locking in InnoDB ................................................................................................. 23 6.3.1. Understanding InnoDB Locking .................................................................................. 23 6.3.2. Granularity of INFORMATION_SCHEMA Data ................................................................. 23 6.3.3. Possible Inconsistency with PROCESSLIST ................................................................... 24 7. Performance and Scalability Enhancements .......................................................................................... 25 7.1. Overview of InnoDB Performance .......................................................................................... 25 7.2. Faster Locking for Improved Scalability ................................................................................... 25 7.3. Using Operating System Memory Allocators .............................................................................. 25 7.4. Controlling InnoDB Change Buffering ..................................................................................... 26 7.5. Controlling Adaptive Hash Indexing ........................................................................................ 27 7.6. Changes Regarding Thread Concurrency .................................................................................. 27 7.7. Changes in the Read-Ahead Algorithm ..................................................................................... 28 7.8. Multiple Background I/O Threads ........................................................................................... 28 7.9. Asynchronous I/O on Linux .................................................................................................. 29 7.10. Group Commit ................................................................................................................ 29 7.11. Controlling the Master Thread I/O Rate .................................................................................. 29 v
- 6. InnoDB 1.1 for MySQL 5.5 User’s Guide 7.12. Controlling the Flushing Rate of Dirty Pages ............................................................................ 30 7.13. Using the PAUSE instruction in InnoDB spin loops .................................................................... 30 7.14. Control of Spin Lock Polling ............................................................................................... 30 7.15. Making Buffer Cache Scan Resistant ...................................................................................... 31 7.16. Improvements to Crash Recovery Performance .......................................................................... 32 7.17. Integration with MySQL PERFORMANCE_SCHEMA ............................................................... 32 7.18. Improvements to Performance from Multiple Buffer Pools ............................................................ 33 7.19. Better Scalability with Multiple Rollback Segments .................................................................... 33 7.20. Better Scalability with Improved Purge Scheduling ..................................................................... 33 7.21. Improved Log Sys Mutex ................................................................................................... 33 7.22. Separate Flush List Mutex .................................................................................................. 34 8. Changes for Flexibility, Ease of Use and Reliability ................................................................................ 35 8.1. Enabling New File Formats .................................................................................................. 35 8.2. Dynamic Control of System Configuration Parameters .................................................................. 35 8.2.1. Dynamically Changing innodb_file_per_table ....................................................... 35 8.2.2. Dynamically Changing innodb_stats_on_metadata ................................................. 36 8.2.3. Dynamically Changing innodb_lock_wait_timeout ................................................. 36 8.2.4. Dynamically Changing innodb_adaptive_hash_index .............................................. 36 8.3. TRUNCATE TABLE Reclaims Space ...................................................................................... 36 8.4. InnoDB Strict Mode ........................................................................................................... 36 8.5. Controlling Optimizer Statistics Estimation ............................................................................... 37 8.6. Better Error Handling when Dropping Indexes ............................................................................ 38 8.7. More Compact Output of SHOW ENGINE INNODB MUTEX ......................................................... 38 8.8. More Read-Ahead Statistics .................................................................................................. 38 9. Installing the InnoDB Storage Engine ................................................................................................ 40 10. Upgrading the InnoDB Storage Engine ............................................................................................. 41 10.1. Upgrading the Dynamic InnoDB Storage Engine ....................................................................... 41 10.2. Upgrading a Statically Built InnoDB Storage Engine ................................................................... 41 10.3. Converting Compressed Tables Created Before Version 1.0.2 ........................................................ 41 11. Downgrading the InnoDB Storage Engine .......................................................................................... 42 11.1. Overview ...................................................................................................................... 42 11.2. The Built-in InnoDB, the Plugin and File Formats ...................................................................... 42 11.3. How to Downgrade .......................................................................................................... 42 11.3.1. Converting Tables .................................................................................................. 42 11.3.2. Adjusting the Configuration ...................................................................................... 42 11.4. Possible Problems ............................................................................................................ 43 11.4.1. Accessing COMPRESSED or DYNAMIC Tables ............................................................... 43 11.4.2. Issues with UNDO and REDO ................................................................................... 43 11.4.3. Issues with the Doublewrite Buffer .............................................................................. 44 11.4.4. Issues with the Insert Buffer ...................................................................................... 44 12. InnoDB Storage Engine Change History ............................................................................................ 45 12.1. Changes in InnoDB Storage Engine 1.x ................................................................................... 45 12.2. Changes in InnoDB Storage Engine 1.1 (April 13, 2010) .............................................................. 45 12.3. Changes in InnoDB Plugin 1.0.x ........................................................................................... 45 A. Third-Party Software .................................................................................................................... 46 A.1. Performance Patches from Google ......................................................................................... 46 A.2. Multiple Background I/O Threads Patch from Percona .................................................................. 46 A.3. Performance Patches from Sun Microsystems ............................................................................ 47 B. List of Parameters Changed in InnoDB 1.1 and InnoDB Plugin 1.0 ............................................................. 48 B.1. New Parameters ............................................................................................................... 48 B.2. Deprecated Parameters ....................................................................................................... 49 B.3. Parameters with New Defaults .............................................................................................. 50 C. InnoDB Glossary ........................................................................................................................ 51 Index .......................................................................................................................................... 98 vi
- 7. List of Tables 3.1. Meaning of CREATE TABLE and ALTER TABLE options ......................................................................7 3.2. CREATE/ALTER TABLE Warnings and Errors when InnoDB Strict Mode is OFF ..........................................8 4.1. InnoDB Data File Compatibility and Related InnoDB Parameters ............................................................. 15 7.1. Changes to innodb_thread_concurrency ................................................................................ 27 B.1. InnoDB 1.1 New Parameter Summary ............................................................................................. 48 B.2. InnoDB Parameters with New Defaults ............................................................................................ 50 vii
- 8. List of Examples 6.1. Using the Compression Information Schema Tables ............................................................................. 18 6.2. Identifying Blocking Transactions .................................................................................................. 20 6.3. More Complex Example of Transaction Data in Information Schema Tables ................................................ 21 viii
- 9. Chapter 1. Introduction to InnoDB 1.1 InnoDB 1.1 combines the familiar reliability and performance of the InnoDB storage engine, with new performance and usability enhancements. InnoDB 1.1 includes all the features that were part of the InnoDB Plugin for MySQL 5.1, plus new features specific to MySQL 5.5 and higher. Beginning with MySQL version 5.5, InnoDB is the default storage engine, rather than MyISAM, to promote greater data reliability and reducing the chance of corruption. 1.1. Features of the InnoDB Storage Engine The InnoDB Storage Engine for MySQL contains several important new features: • Fast index creation: add or drop indexes without copying the data • Data compression: shrink tables, to significantly reduce storage and i/o • New row format: fully off-page storage of long BLOB, TEXT, and VARCHAR columns • File format management: protects upward and downward compatibility • INFORMATION_SCHEMA tables: information about compression and locking • Performance and scalability enhancements • Other changes for flexibility, ease of use and reliability Upward and Downward Compatibility Note that the ability to use data compression and the new row format require the use of a new InnoDB file format called “Barracuda”. The previous file format, used by the built-in InnoDB in MySQL versions 5.0 and 5.1 is now called “Antelope” and does not support these features, but does support the other features introduced with the InnoDB storage engine. The InnoDB storage engine is upward compatible from standard InnoDB as built in to, and distributed with, MySQL. Existing databases can be used with the InnoDB Storage Engine for MySQL. The new parameter innodb_file_format can help pro- tect upward and downward compatibility between InnoDB versions and database files, allowing users to enable or disable use of new features that can only be used with certain versions of InnoDB. The built-in InnoDB in MySQL since version 5.0.21 has a safety feature that prevents it from opening tables that are in an un- known format. However, as noted in Section 11.2, “The Built-in InnoDB, the Plugin and File Formats”, the system tablespace may contain references to new-format tables that confuse the built-in InnoDB in MySQL. These references are cleared in a “slow” shut- down of the InnoDB storage engine. With previous versions of InnoDB, no error would be returned until you try to access a table that is in a format “too new” for the software. Beginning with version 1.0.1 of the InnoDB storage engine, however, to provide early feedback, InnoDB checks the sys- tem tablespace before startup to ensure that the file format used in the database is supported by the storage engine. See Sec- tion 4.2.1, “Compatibility Check When InnoDB Is Started” for the details. 1.2. Obtaining and Installing the InnoDB Storage Engine Starting with MySQL 5.4.2, you do not need to do anything special to get or install the most up-to-date InnoDB storage engine. From that version forward, the InnoDB storage engine in the server is what was formerly known as the InnoDB Plugin. Earlier ver- sions of MySQL required some extra build and configuration steps to get the Plugin-specific features such as fast index creation and table compression. Report any bugs in the InnoDB storage engine using the MySQL bug database. For general discussions about InnoDB Storage En- gine for MySQL, see http://forums.mysql.com/list.php?22. 1.3. Viewing the InnoDB Storage Engine Version Number InnoDB storage engine releases are numbered with version numbers independent of MySQL release numbers. The initial release of the InnoDB storage engine is version 1.0, and it is designed to work with MySQL 5.1. Version 1.1 of the InnoDB storage engine is for MySQL 5.5 and up. 1
- 10. Introduction to InnoDB 1.1 • The first component of the InnoDB storage engine version number designates a major release level. • The second component corresponds to the MySQL release. The digit 0 corresponds to MySQL 5.1. The digit 1 corresponds to MySQL 5.5. • The third component indicates the specific release of the InnoDB storage engine (at a given major release level and for a specif- ic MySQL release); only bug fixes and minor functional changes are introduced at this level. Once you have installed the InnoDB storage engine, you can check its version number in three ways: • In the error log, it is printed during startup • SELECT * FROM information_schema.plugins; • SELECT @@innodb_version; The InnoDB storage engine writes its version number to the error log, which can be helpful in diagnosis of errors: 091105 12:28:06 InnoDB Plugin 1.0.5 started; log sequence number 46509 Note that the PLUGIN_VERSION column in the table INFORMATION_SCHEMA.PLUGINS does not display the third component of the version number, only the first and second components, as in 1.0. 1.4. Compatibility Considerations for Downgrade and Backup Because InnoDB 1.1 supports the “Barracuda” file format, with new on-disk data structures within both the database and log files, pay special attention to file format compatibility with respect to the following scenarios: • Downgrading from MySQL 5.5 to the MySQL 5.1 or earlier (without the InnoDB Plugin enabled), or otherwise using earlier versions of MySQL with database files created by MySQL 5.5 and higher. • Using mysqldump. • Using MySQL replication. • Using MySQL Enterprise Backup or InnoDB Hot Backup. WARNING: Once you create any tables with the Barracuda file format, take care to avoid crashes and corruptions when using those files with an earlier version of InnoDB. It is strongly recommended that you use a “slow shutdown” (SET GLOBAL in- nodb_fast_shutdown=0) when stopping the MySQL server before downgrading to MySQL 5.1 or earlier. This ensures that the log files and other system information do not cause problems when using a prior version of InnoDB. See Section 11.3, “How to Downgrade”. WARNING: If you dump a database containing compressed tables with mysqldump, the dump file may contain CREATE TA- BLE statements that attempt to create compressed tables, or those using ROW_FORMAT=DYNAMIC in the new database. Therefore, be sure the new database is running the InnoDB storage engine, with the proper settings for innodb_file_format and in- nodb_file_per_table, if you want to have the tables re-created as they exist in the original database. Typically, when the mysqldump file is loaded, MySQL and InnoDB ignore CREATE TABLE options they do not recognize, and the table(s) are cre- ated in a format used by the running server. WARNING: If you use MySQL replication, ensure all slaves are configured with the InnoDB storage engine, with the same set- tings for innodb_file_format and innodb_file_per_table. If you do not do so, and you create tables that require the new “Barracuda” file format, replication errors may occur. If a slave MySQL server is running an older version of MySQL, it ig- nores the CREATE TABLE options to create a compressed table or one with ROW_FORMAT=DYNAMIC, and creates the table un- compressed, with ROW_FORMAT=COMPACT. WARNING: Version 3.0 of InnoDB Hot Backup does not support the new “Barracuda” file format. Using InnoDB Hot Backup Version 3 to backup databases in this format causes unpredictable behavior. MySQL Enterprise Backup, the successor product to InnoDB Hot Backup, does support tables with the Barracuda file format. You can also back up such databases with mysqldump. 2
- 11. Chapter 2. Fast Index Creation in the InnoDB Storage Engine In MySQL 5.5 and higher, or in MySQL 5.1 with the InnoDB Plugin, creating and dropping secondary indexes does not copy the contents of the entire table, making this operation much more efficient than with prior releases. 2.1. Overview of Fast Index Creation Historically, adding or dropping an index on a table with existing data could be very slow. The CREATE INDEX and DROP IN- DEX statements worked by creating a new, empty table defined with the requested set of indexes, then copying the existing rows to the new table one-by-one, updating the indexes as the rows are inserted. After all rows from the original table were copied, the old table was dropped and the copy was renamed with the name of the original table. In InnoDB, the rows of a table are stored in a clustered (or primary key) index, forming what some database systems call an “index-organized table”. Changing the clustered index still requires copying the data. The performance speedup applies only to sec- ondary indexes. This new mechanism also means that you can generally speed the overall process of creating and loading an indexed table by creat- ing the table with only the clustered index, and adding the secondary indexes after the data is loaded. Although no syntax changes are required in the CREATE INDEX or DROP INDEX commands, some factors affect the perform- ance, space usage, and semantics of this operation (see Section 2.6, “Limitations of Fast Index Creation”). 2.2. Examples of Fast Index Creation It is possible to create multiple indexes on a table with one ALTER TABLE statement. This is relatively efficient, because the clustered index of the table needs to be scanned only once (although the data is sorted separately for each new index). For example: CREATE TABLE T1(A INT PRIMARY KEY, B INT, C CHAR(1)) ENGINE=InnoDB; INSERT INTO T1 VALUES (1,2,'a'), (2,3,'b'), (3,2,'c'), (4,3,'d'), (5,2,'e'); COMMIT; ALTER TABLE T1 ADD INDEX (B), ADD UNIQUE INDEX (C); The above statements create table T1 with the clustered index (primary key) on column A, insert several rows, and then build two new indexes on columns B and C. If there were many rows inserted into T1 before the ALTER TABLE statement, this approach would be much more efficient than creating the table with all its indexes before loading the data. You can also create the indexes one at a time, but then the clustered index of the table is scanned (as well as sorted) once for each CREATE INDEX statement. Thus, the following statements are not as efficient as the ALTER TABLE statement above, even though neither requires recreating the clustered index for table T1. CREATE INDEX B ON T1 (B); CREATE UNIQUE INDEX C ON T1 (C); Dropping InnoDB secondary indexes also does not require any copying of table data. You can equally quickly drop multiple in- dexes with a single ALTER TABLE statement or multiple DROP INDEX statements: ALTER TABLE T1 DROP INDEX B, DROP INDEX C; or DROP INDEX B ON T1; DROP INDEX C ON T1; Restructuring the clustered index in InnoDB always requires copying the data in the table. For example, if you create a table without a primary key, InnoDB chooses one for you, which may be the first UNIQUE key defined on NOT NULL columns, or a system-generated key. Defining a PRIMARY KEY later causes the data to be copied, as in the following example: CREATE TABLE T2 (A INT, B INT) ENGINE=InnoDB; INSERT INTO T2 VALUES (NULL, 1); ALTER TABLE T2 ADD PRIMARY KEY (B); When you create a UNIQUE or PRIMARY KEY index, InnoDB must do some extra work. For UNIQUE indexes, InnoDB checks that the table contains no duplicate values for the key. For a PRIMARY KEY index, InnoDB also checks that none of the PRIMARY KEY columns contains a NULL. It is best to define the primary key when you create a table, so you need not rebuild the table later. 2.3. Implementation Details of Fast Index Creation 3
- 12. Fast Index Creation in the InnoDB Storage Engine InnoDB has two types of indexes: the clustered index and secondary indexes. Since the clustered index contains the data values in its B-tree nodes, adding or dropping a clustered index does involve copying the data, and creating a new copy of the table. A sec- ondary index, however, contains only the index key and the value of the primary key. This type of index can be created or dropped without copying the data in the clustered index. Furthermore, because the secondary index contains the values of the primary key (used to access the clustered index when needed), when you change the definition of the primary key, thus recreating the clustered index, all secondary indexes are recreated as well. Dropping a secondary index is simple. Only the internal InnoDB system tables and the MySQL data dictionary tables need to be updated to reflect the fact that the index no longer exists. InnoDB returns the storage used for the index to the tablespace that con- tained it, so that new indexes or additional table rows can use the space. To add a secondary index to an existing table, InnoDB scans the table, and sorts the rows using memory buffers and temporary files in order by the value(s) of the secondary index key column(s). The B-tree is then built in key-value order, which is more efficient than inserting rows into an index in random order with respect to the key values. Because the B-tree nodes are split when they fill, building the index in this way results in a higher fill-factor for the index, making it more efficient for subsequent access. 2.4. Concurrency Considerations for Fast Index Creation While a secondary index is being created or dropped, the table is locked in shared mode. Any writes to the table are blocked, but the data in the table can be read. When you alter the clustered index of a table, the table is locked in exclusive mode, because the data must be copied. Thus, during the creation of a new clustered index, all operations on the table are blocked. Before it can start executing, a CREATE INDEX or ALTER TABLE statement must always wait for currently executing transac- tions that are accessing the table to commit or rollback before it can proceed. In addition, ALTER TABLE statements that create a new clustered index must wait for all SELECT statements that access the table to complete (or their containing transactions to com- mit). Even though the original index exists throughout the creation of the new clustered index, no transactions whose execution spans the creation of the index can be accessing the table, because the original table must be dropped when clustered index is re- structured. Once a CREATE INDEX or ALTER TABLE statement that creates a secondary index begins executing, queries can access the ta- ble for read access, but cannot update the table. If an ALTER TABLE statement is changing the clustered index, all queries must wait until the operation completes. A newly-created secondary index contains only the committed data in the table at the time the CREATE INDEX or ALTER TA- BLE statement begins to execute. It does not contain any uncommitted values, old versions of values, or values marked for deletion but not yet removed from the old index. Because a newly-created index contains only information about data current at the time the index was created, queries that need to see data that was deleted or changed before the index was created cannot use the index. The only queries that could be affected by this limitation are those executing in transactions that began before the creation of the index was begun. For such queries, unpre- dictable results could occur. Newer queries can use the index. 2.5. How Crash Recovery Works with Fast Index Creation No data is lost if the server crashes while an ALTER TABLE statement is executing. Recovery, however, is different for clustered indexes and secondary indexes. If the server crashes while creating a secondary index, upon recovery, InnoDB drops any partially created indexes. You must re-run the ALTER TABLE or CREATE INDEX statement. When a crash occurs during the creation of a clustered index, recovery is somewhat more complicated, because the data in the table must be copied to an entirely new clustered index. Remember that all InnoDB tables are stored as clustered indexes. In the follow- ing discussion, we use the word table and clustered index interchangeably. The InnoDB storage engine creates the new clustered index by copying the existing data from the original table to a temporary ta- ble that has the desired index structure. Once the data is completely copied to this temporary table, the original table is renamed with a different temporary table name. The temporary table comprising the new clustered index is then renamed with the name of the original table, and the original table is then dropped from the database. If a system crash occurs while creating a new clustered index, no data is lost, but you must complete the recovery process using the temporary tables that exist during the process. Since it is rare to re-create a clustered index or re-define primary keys on large tables, and even rarer to encounter a system crashes during this operation, this manual does not provide information on recovering from this scenario. Instead, please see the InnoDB web site: http://www.innodb.com/support/tips. 2.6. Limitations of Fast Index Creation Take following considerations into account when creating or dropping InnoDB indexes: 4
- 13. Fast Index Creation in the InnoDB Storage Engine • During index creation, files are written to the temporary directory ($TMPDIR on Unix, %TEMP% on Windows, or the value of - -tmpdir configuration variable). Each temporary file is large enough to hold one column that makes up the new index, and each one is removed as soon as it is merged into the final index. • If any of the indexed columns use the UTF-8 character encoding, MySQL copies the table instead of using Fast Index Creation. This has been reported as MySQL Bug#33650. • The table is copied, rather than using Fast Index Creation when you create an index on a TEMPORARY TABLE. This has been reported as MySQL Bug#39833. • To avoid consistency issues between the InnoDB data dictionary and the MySQL data dictionary, the table is copied, rather than using Fast Index Creation when you use the ALTER TABLE ... RENAME COLUMN syntax. • The statement ALTER IGNORE TABLE t ADD UNIQUE INDEX does not delete duplicate rows. This has been reported as MySQL Bug#40344. The IGNORE keyword is ignored. If any duplicate rows exist, the operation fails with the following error message: ERROR 23000: Duplicate entry '347' for key 'pl' • As noted above, a newly-created index contains only information about data current at the time the index was created. There- fore, you should not run queries in a transaction that might use a secondary index that did not exist at the beginning of the trans- action. There is no way for InnoDB to access “old” data that is consistent with the rest of the data read by the transaction. See the discussion of locking in Section 2.4, “Concurrency Considerations for Fast Index Creation”. Prior to InnoDB storage engine 1.0.4, unexpected results could occur if a query attempts to use an index created after the start of the transaction containing the query. If an old transaction attempts to access a “too new” index, InnoDB storage engine 1.0.4 and later reports an error: ERROR HY000: Table definition has changed, please retry transaction As the error message suggests, committing (or rolling back) the transaction, and restarting it, cures the problem. • InnoDB storage engine 1.0.2 introduces some improvements in error handling when users attempt to drop indexes. See section Section 8.6, “Better Error Handling when Dropping Indexes” for details. • MySQL 5.5 does not support efficient creation or dropping of FOREIGN KEY constraints. Therefore, if you use ALTER TA- BLE to add or remove a REFERENCES constraint, the child table is copied, rather than using Fast Index Creation. 5
- 14. Chapter 3. InnoDB Data Compression By setting InnoDB configuration options, you can create tables where the data is stored in compressed form. The compression means less data is transferred between disk and memory, and takes up less space in memory. The benefits are amplified for tables with secondary indexes, because index data is compressed also. 3.1. Overview of Table Compression Because processors and cache memories have increased in speed more than disk storage devices, many workloads are I/O-bound. Data compression enables smaller database size, reduced I/O, and improved throughput, at the small cost of increased CPU utiliza- tion. Compression is especially valuable for read-intensive applications, on systems with enough RAM to keep frequently-used data in memory. 3.2. Enabling Compression for a Table The default uncompressed size of InnoDB data pages is 16KB. You can use the attributes ROW_FORMAT=COMPRESSED, KEY_BLOCK_SIZE, or both in the CREATE TABLE and ALTER TABLE statements to enable table compression. Depending on the combination of option values, InnoDB attempts to compress each page to 1KB, 2KB, 4KB, 8KB, or 16KB. Note The term KEY_BLOCK_SIZE refers to the size of compressed pages to use for the table. Compression is applicable to tables, not to individual rows, despite the option name ROW_FORMAT. To avoid adding new SQL keywords, the In- noDB storage engine re-uses the clauses originally defined for MyISAM. To create a compressed table, you might use a statement like this: CREATE TABLE name (column1 INT PRIMARY KEY) ENGINE=InnoDB ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=4; If you specify ROW_FORMAT=COMPRESSED but not KEY_BLOCK_SIZE, the default compressed page size of 8KB is used. If KEY_BLOCK_SIZE is specified, you can omit the attribute ROW_FORMAT=COMPRESSED. Setting KEY_BLOCK_SIZE=16 typically does not result in much compression, since the normal InnoDB page size is 16KB. However, this setting may be useful for tables with many long BLOB, VARCHAR or TEXT columns, because such values often do compress well, and might therefore require fewer “overflow” pages as described in Section 3.4.2.2, “Compressing BLOB, VARCHAR and TEXT Columns”. Note that compression is specified on a table-by-table basis. All indexes of a table (including the clustered index) are compressed using the same page size, as specified on the CREATE TABLE or ALTER TABLE statement. Table attributes such as ROW_FORMAT and KEY_BLOCK_SIZE are not part of the CREATE INDEX syntax, and are ignored if they are specified (although you see them in the output of the SHOW CREATE TABLE statement). 3.2.1. Configuration Parameters for Compression Compressed tables are stored in a format that previous versions of InnoDB cannot process. To preserve downward compatibility of database files, compression can be specified only when the “Barracuda” data file format is enabled using the configuration para- meter innodb_file_format. Table compression is also not available for the InnoDB system tablespace. The system tablespace (space 0, the ibdata* files) may contain user data, but it also contains internal InnoDB system information, and therefore is never compressed. Thus, compres- sion applies only to tables (and indexes) stored in their own tablespaces. To use compression, enable the “file per table” mode using the configuration parameter innodb_file_per_table and enable the “Barracuda” disk file format using the parameter innodb_file_format. If necessary, you can set these parameters in the MySQL option file my.cnf or my.ini, or with the SET statement without shutting down the MySQL server. Specifying ROW_FORMAT=COMPRESSED or a KEY_BLOCK_SIZE in the CREATE TABLE or ALTER TABLE statements if the “Barracuda” file format is not enabled produces these warnings that you can view with the SHOW WARNINGS statement: Level Code Message Warning 1478 InnoDB: KEY_BLOCK_SIZE requires innodb_file_per_table. Warning 1478 InnoDB: KEY_BLOCK_SIZE requires innodb_file_format=1. Warning 1478 InnoDB: ignoring KEY_BLOCK_SIZE=4. 6
- 15. InnoDB Data Compression Level Code Message Warning 1478 InnoDB: ROW_FORMAT=COMPRESSED requires innodb_file_per_table. Warning 1478 InnoDB: assuming ROW_FORMAT=COMPACT. Note These messages are only warnings, not errors, and the table is created as if the options were not specified. When In- noDB “strict mode” (see Section 8.4, “InnoDB Strict Mode”) is enabled, InnoDB generates an error, not a warning, for these cases. In strict mode, the table is not created if the current configuration does not permit using compressed tables. The “non-strict” behavior is intended to permit you to import a mysqldump file into a database that does not support compressed tables, even if the source database contained compressed tables. In that case, the InnoDB storage engine creates the table in ROW_FORMAT=COMPACT instead of preventing the operation. When you import the dump file into a new database, if you want to have the tables re-created as they exist in the original database, ensure the server is running the InnoDB storage engine with the proper settings for the configuration parameters in- nodb_file_format and innodb_file_per_table, 3.2.2. SQL Compression Syntax Warnings and Errors The attribute KEY_BLOCK_SIZE is permitted only when ROW_FORMAT is specified as COMPRESSED or is omitted. Specifying a KEY_BLOCK_SIZE with any other ROW_FORMAT generates a warning that you can view with SHOW WARNINGS. However, the table is non-compressed; the specified KEY_BLOCK_SIZE is ignored). Level Code Message Warning 1478 InnoDB: ignoring KEY_BLOCK_SIZE=n unless ROW_FORMAT=COMPRESSED. If you are running in InnoDB strict mode, the combination of a KEY_BLOCK_SIZE with any ROW_FORMAT other than COM- PRESSED generates an error, not a warning, and the table is not created. Table 3.1, “Meaning of CREATE TABLE and ALTER TABLE options” summarizes how the various options on CREATE TABLE and ALTER TABLE are handled. Table 3.1. Meaning of CREATE TABLE and ALTER TABLE options Option Usage Description ROW_FORMAT= Storage format used prior to MySQL Less efficient than ROW_FORMAT=COMPACT; for backward compat- REDUNDANT 5.0.3 ibility ROW_FORMAT= Default storage format since MySQL Stores a prefix of 768 bytes of long column values in the clustered in- COMPACT 5.0.3 dex page, with the remaining bytes stored in an overflow page ROW_FORMAT= Available only with innodb_file Store values within the clustered index page if they fit; if not, stores DYNAMIC _format=Barracuda only a 20-byte pointer to an overflow page (no prefix) ROW_FORMAT= Available only with innodb_file Compresses the table and indexes using zlib to default compressed COMPRESSED _format=Barracuda page size of 8K bytes; implies ROW_FORMAT=DYNAMIC KEY_BLOCK_ Available only with innodb_file Specifies compressed page size of 1, 2, 4, 8 or 16K bytes; implies SIZE=n _format=Barracuda ROW_FORMAT=DYNAMIC and ROW_FORMAT=COMPRESSED Table 3.2, “CREATE/ALTER TABLE Warnings and Errors when InnoDB Strict Mode is OFF” summarizes error conditions that occur with certain combinations of configuration parameters and options on the CREATE TABLE or ALTER TABLE statements, and how the options appear in the output of SHOW TABLE STATUS. When InnoDB strict mode is OFF, InnoDB creates or alters the table, but may ignore certain settings, as shown below. You can see the warning messages in the MySQL error log. When InnoDB strict mode is ON, these specified combinations of options generate errors, and the table is not created or altered. You can see the full description of the error condition with SHOW ERRORS. For ex- ample: mysql> CREATE TABLE x (id INT PRIMARY KEY, c INT) -> ENGINE=INNODB KEY_BLOCK_SIZE=33333; ERROR 1005 (HY000): Can't create table 'test.x' (errno: 1478) 7
- 16. InnoDB Data Compression mysql> SHOW ERRORS; +-------+------+-------------------------------------------+ | Level | Code | Message | +-------+------+-------------------------------------------+ | Error | 1478 | InnoDB: invalid KEY_BLOCK_SIZE=33333. | | Error | 1005 | Can't create table 'test.x' (errno: 1478) | +-------+------+-------------------------------------------+ 2 rows in set (0.00 sec) Table 3.2. CREATE/ALTER TABLE Warnings and Errors when InnoDB Strict Mode is OFF Syntax Warning or Error Condition Resulting ROW_FORMAT, as shown in SHOW TABLE STATUS ROW_FORMAT=REDUNDANT None REDUNDANT ROW_FORMAT=COMPACT None COMPACT ROW_FORMAT=COMPRESSED or Ignored unless both COMPACT ROW_FORMAT=DYNAMIC or innodb_file_format=Barracuda and in- KEY_BLOCK_SIZE is specified nodb_file_per_table are enabled Invalid KEY_BLOCK_SIZE is KEY_BLOCK_SIZE is ignored the requested one, or COMPACT by specified (not 1, 2, 4, 8 or 16) default ROW_FORMAT=COMPRESSED None; KEY_BLOCK_SIZE specified is used, not the COMPRESSED and valid KEY_BLOCK_SIZE 8K default are specified KEY_BLOCK_SIZE is specified KEY_BLOCK_SIZE is ignored REDUNDANT, COMPACT or DYNAM- with REDUNDANT, COMPACT or IC DYNAMIC row format ROW_FORMAT is not one of RE- Ignored if recognized by the MySQL parser. Other- COMPACT or N/A DUNDANT, COMPACT, DYNAMIC wise, an error is issued. or COMPRESSED When InnoDB strict mode is ON (innodb_strict_mode=1), the InnoDB storage engine rejects invalid ROW_FORMAT or KEY_BLOCK_SIZE parameters. For compatibility with earlier versions of InnoDB,, InnoDB strict mode is not enabled by default; instead, InnoDB issues warnings (not errors) for ignored invalid parameters. Note that it is not possible to see the chosen KEY_BLOCK_SIZE using SHOW TABLE STATUS. The statement SHOW CREATE TABLE displays the KEY_BLOCK_SIZE (even if it was ignored by InnoDB). The real compressed page size inside InnoDB cannot be displayed by MySQL. 3.3. Tuning InnoDB Compression Most often, the internal optimizations in InnoDB described in section Section 3.4.2, “InnoDB Data Storage and Compression” en- sure that the system runs well with compressed data. However, because the efficiency of compression depends on the nature of your data, there are some factors you should consider to get best performance. You need to choose which tables to compress, and what compressed page size to use. You might also adjust the size of the buffer pool based on run-time performance characteristics, such as the amount of time the system spends compressing and uncompressing data. 3.3.1. When to Use Compression In general, compression works best on tables that include a reasonable number of character string columns and where the data is read far more often than it is written. Because there are no guaranteed ways to predict whether or not compression benefits a partic- ular situation, always test with a specific workload and data set running on a representative configuration. Consider the following factors when deciding which tables to compress. 3.3.1.1. Data Characteristics and Compression A key determinant of the efficiency of compression in reducing the size of data files is the nature of the data itself. Recall that com- pression works by identifying repeated strings of bytes in a block of data. Completely randomized data is the worst case. Typical data often has repeated values, and so compresses effectively. Character strings often compress well, whether defined in CHAR, VARCHAR, TEXT or BLOB columns. On the other hand, tables containing mostly binary data (integers or floating point numbers) or data that is previously compressed (for example JPEG or PNG images) may not generally compress well, significantly or at all. Compression is chosen on a table by table basis with the InnoDB storage engine, and a table and all of its indexes use the same (compressed) page size. It might be that the primary key (clustered) index, which contains the data for all columns of a table, com- presses more effectively than the secondary indexes. For those cases where there are long rows, the use of compression might result in long column values being stored “off-page”, as discussed in Section 5.3, “DYNAMIC Row Format”. Those overflow pages may 8
- 17. InnoDB Data Compression compress well. Given these considerations, for many applications, some tables compress more effectively than others, and you might find that your workload performs best only with a subset of tables compressed. Experimenting is the only way to determine whether or not to compress a particular table. InnoDB compresses data in 16K chunks corresponding to the uncompressed page size, and in addition to user data, the page format includes some internal system data that is not compressed. Compression utilities compress an entire stream of data, and so may find more repeated strings across the entire input stream than InnoDB would find in a table compressed in 16K chunks. But you can get a sense of how compression efficiency by using a utility that implements LZ77 compression (such as gzip or WinZip) on your data file. Another way to test compression on a specific table is to copy some data from your uncompressed table to a similar, compressed ta- ble (having all the same indexes) and look at the size of the resulting file. When you do so (if nothing else using compression is running), you can examine the ratio of successful compression operations to overall compression operations. (In the INNODB_CMP table, compare COMPRESS_OPS to COMPRESS_OPS_OK. See INNODB_CMP for more information.) If a high percentage of com- pression operations complete successfully, the table might be a good candidate for compression. 3.3.1.2. Compression and Application and Schema Design Decide whether to compress data in your application or in the InnoDB table. It is usually not sensible to store data that is com- pressed by an application in an InnoDB compressed table. Further compression is extremely unlikely, and the attempt to compress just wastes CPU cycles. 3.3.1.2.1. Compressing in the Database The InnoDB table compression is automatic and applies to all columns and index values. The columns can still be tested with oper- ators such as LIKE, and sort operations can still use indexes even when the index values are compressed. Because indexes are often a significant fraction of the total size of a database, compression could result in significant savings in storage, I/O or processor time. The compression and decompression operations happen on the database server, which likely is a powerful system that is sized to handle the expected load. 3.3.1.2.2. Compressing in the Application If you compress data such as text in your application, before it is inserted into the database, You might save overhead for data that does not compress well by compressing some columns and not others. This approach uses CPU cycles for compression and uncom- pression on the client machine rather than the database server, which might be appropriate for a distributed application with many clients, or where the client machine has spare CPU cycles. 3.3.1.2.3. Hybrid Approach Of course, it is possible to combine these approaches. For some applications, it may be appropriate to use some compressed tables and some uncompressed tables. It may be best to externally compress some data (and store it in uncompressed InnoDB tables) and allow InnoDB to compress (some of) the other tables in the application. As always, up-front design and real-life testing are valu- able in reaching the right decision. 3.3.1.3. Workload Characteristics and Compression In addition to choosing which tables to compress (and the page size), the workload is another key determinant of performance. If the application is dominated by reads, rather than updates, fewer pages need to be reorganized and recompressed after the index page runs out of room for the per-page “modification log” that InnoDB maintains for compressed data. If the updates predomin- antly change non-indexed columns or those containing BLOBs or large strings that happen to be stored “off-page”, the overhead of compression may be acceptable. If the only changes to a table are INSERTs that use a monotonically increasing primary key, and there are few secondary indexes, there is little need to reorganize and recompress index pages. Since InnoDB can “delete-mark” and delete rows on compressed pages “in place” by modifying uncompressed data, DELETE operations on a table are relatively ef- ficient. For some environments, the time it takes to load data can be as important as run-time retrieval. Especially in data warehouse envir- onments, many tables may be read-only or read-mostly. In those cases, it might or might not be acceptable to pay the price of com- pression in terms of increased load time, unless the resulting savings in fewer disk reads or in storage cost is significant. Fundamentally, compression works best when the CPU time is available for compressing and uncompressing data. Thus, if your workload is I/O bound, rather than CPU-bound, you might find that compression can improve overall performance. Therefore when you test your application performance with different compression configurations, it is important to test on a platform similar to the planned configuration of the production system. 3.3.1.4. Configuration Characteristics and Compression Reading and writing database pages from and to disk is the slowest aspect of system performance. Therefore, compression attempts to reduce I/O by using CPU time to compress and uncompress data, and thus is most effective when I/O is a relatively scarce re- source compared to processor cycles. This is often especially the case when running in a multi-user environment with fast, multi-core CPUs. When a page of a com- 9
- 18. InnoDB Data Compression pressed table is in memory, InnoDB often uses an additional 16K in the buffer pool for an uncompressed copy of the page. The ad- aptive LRU algorithm in the InnoDB storage engine attempts to balance the use of memory between compressed and uncompressed pages to take into account whether the workload is running in an I/O-bound or CPU-bound manner. Nevertheless, a configuration with more memory dedicated to the InnoDB buffer pool tends to run better when using compressed tables than a configuration where memory is highly constrained. 3.3.1.5. Choosing the Compressed Page Size The optimal setting of the compressed page size depends on the type and distribution of data that the table and its indexes contain. The compressed page size should always be bigger than the maximum record size, or operations may fail as noted in Sec- tion 3.4.2.1, “Compression of B-Tree Pages”. Setting the compressed page size too large wastes some space, but the pages do not have to be compressed as often. If the com- pressed page size is set too small, inserts or updates may require time-consuming recompression, and the B-tree nodes may have to be split more frequently, leading to bigger data files and less efficient indexing. Typically, one would set the compressed page size to 8K or 4K bytes. Given that the maximum InnoDB record size is around 8K, KEY_BLOCK_SIZE=8 is usually a safe choice. 3.3.2. Monitoring Compression at Runtime The current version of the InnoDB storage engine provides only a limited means to monitor the performance of compression at runtime. Overall application performance, CPU and I/O utilization and the size of disk files are the best indicators of how effective compression is for your application. The InnoDB storage engine does include some Information Schema tables (see Example 6.1, “Using the Compression Information Schema Tables”) that reflect the internal use of memory and the rates of compression used overall. The INNODB_CMP tables report information about compression activity for each compressed page size (KEY_BLOCK_SIZE) in use. The information in these tables is system-wide, and includes summary data across all compressed tables in your database. You can use this data to help de- cide whether or not to compress a table by examining these tables when no other compressed tables are being accessed. The key statistics to consider are the number of, and amount of time spent performing, compression and uncompression operations. Since InnoDB must split B-tree nodes when they are too full to contain the compressed data following a modification, you should also compare the number of “successful” compression operations with the number of such operations overall. Based on the inform- ation in the INNODB_CMP tables and overall application performance and hardware resource utilization, you might make changes in your hardware configuration, adjust the size of the InnoDB buffer pool, choose a different page size, or select a different set of tables to compress. If the amount of CPU time required for compressing and uncompressing is high, changing to faster CPUs, or those with more cores, can help improve performance with the same data, application workload and set of compressed tables. Increasing the size of the InnoDB buffer pool might also help performance, so that more uncompressed pages can stay in memory, reducing the need to uncompress pages that exist in memory only in compressed form. A large number of compression operations overall (compared to the number of INSERT, UPDATE and DELETE operations in your application and the size of the database) could indicate that some of your compressed tables are being updated too heavily for ef- fective compression. If so, choose a larger page size, or be more selective about which tables you compress. If the number of “successful” compression operations (COMPRESS_OPS_OK) is a high percentage of the total number of compres- sion operations (COMPRESS_OPS), then the system is likely performing well. However, if the ratio is low, then InnoDB is being caused to reorganize, recompress and split B-tree nodes more often than is desirable. In this case, avoid compressing some tables, or choose a larger KEY_BLOCK_SIZE for some of the compressed tables. You might want to turn off compression for tables that cause the number of “compression failures” in your application to be more than 1% or 2% of the total (although such a failure ratio might be acceptable during a temporary operation such as a data load). 3.4. How Compression Works in InnoDB This section describes some internal implementation details about compression in InnoDB. The information presented here may be helpful in tuning for performance, but is not necessary to know for basic use of compression. 3.4.1. Compression Algorithms Some operating systems implement compression at the file system level. Files are typically divided into fixed-size blocks that are compressed into variable-size blocks, which easily leads into fragmentation. Every time something inside a block is modified, the whole block is recompressed before it is written to disk. These properties make this compression technique unsuitable for use in an update-intensive database system. The InnoDB storage engine implements a novel type of compression with the help of the well-known zlib library, which imple- ments the LZ77 compression algorithm. This compression algorithm is mature, robust, and efficient in both CPU utilization and in 10
- 19. InnoDB Data Compression reduction of data size. The algorithm is “lossless”, so that the original uncompressed data can always be reconstructed from the compressed form. LZ77 compression works by finding sequences of data that are repeated within the data to be compressed. The patterns of values in your data determine how well it compresses, but typical user data often compresses by 50% or more. Unlike compression performed by an application, or compression features of some other database management systems, InnoDB compression applies both to user data and to indexes. In many cases, indexes can constitute 40-50% or more of the total database size, so this difference is significant. When compression is working well for a data set, the size of the InnoDB data files (the .idb files) is 25% to 50% of the uncompressed size or possibly smaller. Depending on the workload, this smaller database can in turn lead to a reduction in I/O, and an increase in throughput, at a modest cost in terms of increased CPU utilization. 3.4.2. InnoDB Data Storage and Compression All user data in InnoDB is stored in pages comprising a B-tree index (the so-called clustered index). In some other database sys- tems, this type of index is called an “index-organized table”. Each row in the index node contains the values of the (user-specified or system-generated) primary key and all the other columns of the table. Secondary indexes in InnoDB are also B-trees, containing pairs of values: the index key and a pointer to a row in the clustered in- dex. The pointer is in fact the value of the primary key of the table, which is used to access the clustered index if columns other than the index key and primary key are required. Secondary index records must always fit on a single B-tree page. The compression of B-tree nodes (of both clustered and secondary indexes) is handled differently from compression of overflow pages used to store long VARCHAR, BLOB, or TEXT columns, as explained in the following sections. 3.4.2.1. Compression of B-Tree Pages Because they are frequently updated, B-tree pages require special treatment. It is important to minimize the number of times B-tree nodes are split, as well as to minimize the need to uncompress and recompress their content. One technique InnoDB uses is to maintain some system information in the B-tree node in uncompressed form, thus facilitating cer- tain in-place updates. For example, this allows rows to be delete-marked and deleted without any compression operation. In addition, InnoDB attempts to avoid unnecessary uncompression and recompression of index pages when they are changed. With- in each B-tree page, the system keeps an uncompressed “modification log” to record changes made to the page. Updates and inserts of small records may be written to this modification log without requiring the entire page to be completely reconstructed. When the space for the modification log runs out, InnoDB uncompresses the page, applies the changes and recompresses the page. If recompression fails, the B-tree nodes are split and the process is repeated until the update or insert succeeds. Generally, InnoDB requires that each B-tree page can accommodate at least two records. For compressed tables, this requirement has been relaxed. Leaf pages of B-tree nodes (whether of the primary key or secondary indexes) only need to accommodate one re- cord, but that record must fit in uncompressed form, in the per-page modification log. Starting with InnoDB storage engine version 1.0.2, and if InnoDB strict mode is ON, the InnoDB storage engine checks the maximum row size during CREATE TABLE or CREATE INDEX. If the row does not fit, the following error message is issued: ERROR HY000: Too big row. If you create a table when InnoDB strict mode is OFF, and a subsequent INSERT or UPDATE statement attempts to create an index entry that does not fit in the size of the compressed page, the operation fails with ERROR 42000: Row size too large. Unfortunately, this error message does not name the index for which the record is too large, or mention the length of the index re- cord or the maximum record size on that particular index page. The only remedy is to rebuild the table with ALTER TABLE, to se- lect a larger compressed page size (KEY_BLOCK_SIZE), to shorten any column prefix indexes, or to disable compression entirely with ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPACT. 3.4.2.2. Compressing BLOB, VARCHAR and TEXT Columns In a clustered index, BLOB, VARCHAR and TEXT columns that are not part of the primary key may be stored on separately alloc- ated (“overflow”) pages. We call these “off-page columns” whose values are stored on singly-linked lists of overflow pages. For tables created in ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED, the values of BLOB, TEXT or VARCHAR columns may be stored fully off-page, depending on their length and the length of the entire row. For columns that are stored off- page, the clustered index record only contains 20-byte pointers to the overflow pages, one per column. Whether any columns are stored off-page depends on the page size and the total size of the row. When the row is too long to fit entirely within the page of the clustered index, InnoDB chooses the longest columns for off-page storage until the row fits on the clustered index page. As noted above, if a row does not fit by itself on a compressed page, an error occurs. Tables created in previous versions of InnoDB use the “Antelope” file format, which supports only ROW_FORMAT=REDUNDANT and ROW_FORMAT=COMPACT. In these formats, InnoDB stores the first 768 bytes of BLOB, VARCHAR and TEXT columns in the clustered index record along with the primary key. The 768-byte prefix is followed by a 20-byte pointer to the overflow pages that contain the rest of the column value. When a table is in COMPRESSED format, all data written to overflow pages is compressed “as is”; that is, InnoDB applies the zlib compression algorithm to the entire data item. Other than the data, compressed overflow pages contain an uncompressed header 11
- 20. InnoDB Data Compression and trailer comprising a page checksum and a link to the next overflow page, among other things. Therefore, very significant stor- age savings can be obtained for longer BLOB, TEXT or VARCHAR columns if the data is highly compressible, as is often the case with text data (but not previously compressed images). The overflow pages are of the same size as other pages. A row containing ten columns stored off-page occupies ten overflow pages, even if the total length of the columns is only 8K bytes. In an uncompressed table, ten uncompressed overflow pages occupy 160K bytes. In a compressed table with an 8K page size, they occupy only 80K bytes. Thus, it is often more efficient to use compressed table format for tables with long column values. Using a 16K compressed page size can reduce storage and I/O costs for BLOB, VARCHAR or TEXT columns, because such data of- ten compress well, and might therefore require fewer “overflow” pages, even though the B-tree nodes themselves take as many pages as in the uncompressed form. 3.4.3. Compression and the InnoDB Buffer Pool In a compressed InnoDB table, every compressed page (whether 1K, 2K, 4K or 8K) corresponds to an uncompressed page of 16K bytes. To access the data in a page, InnoDB reads the compressed page from disk if it is not already in the buffer pool, then uncom- presses the page to its original 16K byte form. This section describes how InnoDB manages the buffer pool with respect to pages of compressed tables. To minimize I/O and to reduce the need to uncompress a page, at times the buffer pool contains both the compressed and uncom- pressed form of a database page. To make room for other required database pages, InnoDB may “evict” from the buffer pool an un- compressed page, while leaving the compressed page in memory. Or, if a page has not been accessed in a while, the compressed form of the page may be written to disk, to free space for other data. Thus, at any given time, the buffer pool may contain both the compressed and uncompressed forms of the page, or only the compressed form of the page, or neither. InnoDB keeps track of which pages to keep in memory and which to evict using a least-recently-used (LRU) list, so that “hot” or frequently accessed data tends to stay in memory. When compressed tables are accessed, InnoDB uses an adaptive LRU algorithm to achieve an appropriate balance of compressed and uncompressed pages in memory. This adaptive algorithm is sensitive to whether the system is running in an I/O-bound or CPU-bound manner. The goal is to avoid spending too much processing time un- compressing pages when the CPU is busy, and to avoid doing excess I/O when the CPU has spare cycles that can be used for un- compressing compressed pages (that may already be in memory). When the system is I/O-bound, the algorithm prefers to evict the uncompressed copy of a page rather than both copies, to make more room for other disk pages to become memory resident. When the system is CPU-bound, InnoDB prefers to evict both the compressed and uncompressed page, so that more memory can be used for “hot” pages and reducing the need to uncompress data in memory only in compressed form. 3.4.4. Compression and the InnoDB Log Files Before a compressed page is written to a database file, InnoDB writes a copy of the page to the redo log (if it has been recom- pressed since the last time it was written to the database). This is done to ensure that redo logs will always be usable, even if a fu- ture version of InnoDB uses a slightly different compression algorithm. Therefore, some increase in the size of log files, or a need for more frequent checkpoints, can be expected when using compression. The amount of increase in the log file size or checkpoint frequency depends on the number of times compressed pages are modified in a way that requires reorganization and recompression. Note that the redo log file format (and the database file format) are different from previous releases when using compression. The MySQL Enterprise Backup product does support this latest “Barracuda” file format for compressed InnoDB tables. The older In- noDB Hot Backup product can only back up tables using the file format “Antelope”, and thus does not support InnoDB tables that use compression. 12
- 21. Chapter 4. InnoDB File Format Management As InnoDB evolves, new on-disk data structures are sometimes required to support new features. Compressed tables (see Chapter 3, InnoDB Data Compression), and long variable-length columns stored off-page (see Chapter 5, Storage of Variable- Length Columns) are new features requiring data file formats that are not compatible with prior versions of InnoDB. The other new InnoDB features do not require the use of the new file format. Named File Formats InnoDB 1.1 has the idea of a named file format and a configuration parameter to enable the use of features that require use of that format. The new file format is the “Barracuda” format, and the original InnoDB file format is called “Antelope”. Compressed tables and the new row format that stores long columns “off-page” require the use of the “Barracuda” file format or newer. Future versions of InnoDB may introduce a series of file formats, identified with the names of animals, in ascending alphabetic order. 4.1. Enabling File Formats The configuration parameter innodb_file_format controls whether such statements as CREATE TABLE and ALTER TA- BLE can be used to create tables that depend on support for the “Barracuda” file format. In MySQL 5.5.5 and higher, the default value is “Barracuda”. The file format is a dynamic, global parameter that can be specified in the MySQL option file (my.cnf or my.ini) or changed with the SET GLOBAL command. 4.2. File Format Compatibility InnoDB 1.1 incorporates several checks to guard against the possible crashes and data corruptions that might occur if you run an older release of the MySQL server on InnoDB data files using a newer file format. These checks take place when the server is star- ted, and when you first access a table. This section describes these checks, how you can control them, and error and warning condi- tions that might arise. Backward Compatibility Considerations of backward compatibility only apply when using a recent version of InnoDB (the InnoDB Plugin, or MySQL 5.5 and higher with InnoDB 1.1) alongside an older one (MySQL 5.1 or earlier, with the built-in InnoDB rather than the InnoDB Plu- gin). To minimize the chance of compatibility issues, you can standardize on the InnoDB Plugin for all your MySQL 5.1 and earli- er database servers. In general, a newer version of InnoDB may create a table or index that cannot safely be read or written with a prior version of In- noDB without risk of crashes, hangs, wrong results or corruptions. InnoDB 1.1 includes a mechanism to guard against these condi- tions, and to help preserve compatibility among database files and versions of InnoDB. This mechanism lets you take advantage of some new features of an InnoDB release (such as performance improvements and bug fixes), and still preserve the option of using your database with a prior version of InnoDB, by preventing accidental use of new features that create downward-incompatible disk files. If a version of InnoDB supports a particular file format (whether or not that format is the default), you can query and update any ta- ble that requires that format or an earlier format. Only the creation of new tables using new features is limited based on the particu- lar file format enabled. Conversely, if a tablespace contains a table or index that uses a file format that is not supported by the cur- rently running software, it cannot be accessed at all, even for read access. The only way to “downgrade” an InnoDB tablespace to an earlier file format is to copy the data to a new table, in a tablespace that uses the earlier format. This can be done with the ALTER TABLE statement, as described in Section 4.4, “Downgrading the File Format”. The easiest way to determine the file format of an existing InnoDB tablespace is to examine the properties of the table it contains, using the SHOW TABLE STATUS command or querying the table INFORMATION_SCHEMA.TABLES. If the Row_format of the table is reported as 'Compressed' or 'Dynamic', the tablespace containing the table uses the “Barracuda” format. Other- wise, it uses the prior InnoDB file format, “Antelope”. Internal Details Every InnoDB per-table tablespace (represented by a *.ibd file) file is labeled with a file format identifier. The system tablespace (represented by the ibdata files) is tagged with the “highest” file format in use in a group of InnoDB database files, and this tag is checked when the files are opened. Creating a compressed table, or a table with ROW_FORMAT=DYNAMIC, updates the file header for the corresponding .ibd file 13
- 22. InnoDB File Format Management and the table type in the InnoDB data dictionary with the identifier for the “Barracuda” file format. From that point forward, the ta- ble cannot be used with a version of InnoDB that does not support this new file format. To protect against anomalous behavior, In- noDB version 5.0.21 and later performs a compatibility check when the table is opened. (In many cases, the ALTER TABLE state- ment recreates a table and thus changes its properties. The special case of adding or dropping indexes without rebuilding the table is described in Chapter 2, Fast Index Creation in the InnoDB Storage Engine.) Definition of ib-file set To avoid confusion, for the purposes of this discussion we define the term “ib-file set” to mean the set of operating system files that InnoDB manages as a unit. The ib-file set includes the following files: • The system tablespace (one or more ibdata files) that contain internal system information (including internal catalogs and undo information) and may include user data and indexes. • Zero or more single-table tablespaces (also called “file per table” files, named *.ibd files). • InnoDB log files; usually two, ib_logfile0 and ib_logfile1. Used for crash recovery and in backups. An “ib-file set” does not include the corresponding .frm files that contain metadata about InnoDB tables. The .frm files are cre- ated and managed by MySQL, and can sometimes get out of sync with the internal metadata in InnoDB. Multiple tables, even from more than one database, can be stored in a single “ib-file set”. (In MySQL, a “database” is a logical col- lection of tables, what other systems refer to as a “schema” or “catalog”.) 4.2.1. Compatibility Check When InnoDB Is Started To prevent possible crashes or data corruptions when InnoDB opens an ib-file set, it checks that it can fully support the file formats in use within the ib-file set. If the system is restarted following a crash, or a “fast shutdown” (i.e., innodb_fast_shutdown is greater than zero), there may be on-disk data structures (such as redo or undo entries, or doublewrite pages) that are in a “too-new” format for the current software. During the recovery process, serious damage can be done to your data files if these data structures are accessed. The startup check of the file format occurs before any recovery process begins, thereby preventing the problems de- scribed in Section 11.4, “Possible Problems”. Beginning with version InnoDB 1.0.1, the system tablespace records an identifier or tag for the “highest” file format used by any table in any of the tablespaces that is part of the ib-file set. Checks against this file format tag are controlled by the configuration parameter innodb_file_format_check, which is ON by default. If the file format tag in the system tablespace is newer or higher than the highest version supported by the particular currently ex- ecuting software and if innodb_file_format_check is ON, the following error is issued when the server is started: InnoDB: Error: the system tablespace is in a file format that this version doesn't support You can also set innodb_file_format to a file format name. Doing so prevents InnoDB from starting if the current software does not support the file format specified. It also sets the “high water mark” to the value you specify. The ability to set in- nodb_file_format_check will be useful (with future releases of InnoDB) if you manually “downgrade” all of the tables in an ib-file set (as described in Chapter 11, Downgrading the InnoDB Storage Engine). You can then rely on the file format check at startup if you subsequently use an older version of InnoDB to access the ib-file set. In some limited circumstances, you might want to start the server and use an ib-file set that is in a “too new” format (one that is not supported by the software you are using). If you set the configuration parameter innodb_file_format_check to OFF, In- noDB opens the database, but issues this warning message in the error log: InnoDB: Warning: the system tablespace is in a file format that this version doesn't support Note This is a very dangerous setting, as it permits the recovery process to run, possibly corrupting your database if the pre- vious shutdown was a crash or “fast shutdown”. You should only set innodb_file_format_check to OFF if you are sure that the previous shutdown was done with innodb_fast_shutdown=0, so that essentially no recov- ery process occurs. In a future release, this parameter setting may be renamed from OFF to UNSAFE. (However, until there are newer releases of InnoDB that support additional file formats, even disabling the startup checking is in fact “safe”.) The parameter innodb_file_format_check affects only what happens when a database is opened, not subsequently. Con- versely, the parameter innodb_file_format (which enables a specific format) only determines whether or not a new table can be created in the enabled format and has no effect on whether or not a database can be opened. 14
- 23. InnoDB File Format Management The file format tag is a “high water mark”, and as such it is increased after the server is started, if a table in a “higher” format is created or an existing table is accessed for read or write (assuming its format is supported). If you access an existing table in a format higher than the format the running software supports, the system tablespace tag is not updated, but table-level compatibility checking applies (and an error is issued), as described in Section 4.2.2, “Compatibility Check When a Table Is Opened”. Any time the high water mark is updated, the value of innodb_file_format_check is updated as well, so the command SELECT @@innodb_file_format_check; displays the name of the newest file format known to be used by tables in the currently open ib-file set and supported by the currently executing software. To best illustrate this behavior, consider the scenario described in Table 4.1, “InnoDB Data File Compatibility and Related InnoDB Parameters”. Imagine that some future version of InnoDB supports the “Cheetah” format and that an ib-file set has been used with that version. Table 4.1. InnoDB Data File Compatibility and Related InnoDB Parameters innodb innodb file Highest file Highest file Result file format format used format sup- format in ib-file set ported by In- check noDB OFF Antelope or Barracuda Barracuda Database can be opened; tables can be created which require Barracuda “Antelope” or “Barracuda” file format OFF Antelope or Cheetah Barracuda Database can be opened with a warning, since the database con- Barracuda tains files in a “too new” format; tables can be created in “Antelope” or “Barracuda” file format; tables in “Cheetah” format cannot be accessed OFF Cheetah Barracuda Barracuda Database cannot be opened; innodb_file_format cannot be set to “Cheetah” ON Antelope or Barracuda Barracuda Database can be opened; tables can be created in “Antelope” or Barracuda “Barracuda” file format ON Antelope or Cheetah Barracuda Database cannot be opened, since the database contains files in a Barracuda “too new” format (“Cheetah”) ON Cheetah Barracuda Barracuda Database cannot be opened; innodb_file_format cannot be set to “Cheetah” 4.2.2. Compatibility Check When a Table Is Opened When a table is first accessed, InnoDB (including some releases prior to InnoDB 1.0) checks that the file format of the tablespace in which the table is stored is fully supported. This check prevents crashes or corruptions that would otherwise occur when tables using a “too new” data structure are encountered. All tables using any file format supported by a release can be read or written (assuming the user has sufficient privileges). The set- ting of the system configuration parameter innodb_file_format can prevent creating a new table that uses specific file formats, even if they are supported by a given release. Such a setting might be used to preserve backward compatibility, but it does not prevent accessing any table that uses any supported format. As noted in Chapter 4, InnoDB File Format Management [13], versions of MySQL older than 5.0.21 cannot reliably use database files created by newer versions if a new file format was used when a table was created. To prevent various error conditions or cor- ruptions, InnoDB checks file format compatibility when it opens a file (for example, upon first access to a table). If the currently running version of InnoDB does not support the file format identified by the table type in the InnoDB data dictionary, MySQL re- ports the following error: ERROR 1146 (42S02): Table 'test.t1' doesn't exist InnoDB also writes a message to the error log: InnoDB: table test/t1: unknown table type 33 The table type should be equal to the tablespace flags, which contains the file format version as discussed in Section 4.3, “Identifying the File Format in Use”. Versions of InnoDB prior to MySQL 4.1 did not include table format identifiers in the database files, and versions prior to MySQL 5.0.21 did not include a table format compatibility check. Therefore, there is no way to ensure proper operations if a table in a “too new” format is used with versions of InnoDB prior to 5.0.21. The file format management capability in InnoDB 1.0 and higher (tablespace tagging and run-time checks) allows InnoDB to verify 15
- 24. InnoDB File Format Management as soon as possible that the running version of software can properly process the tables existing in the database. If you permit InnoDB to open a database containing files in a format it does not support (by setting the parameter in- nodb_file_format_check to OFF), the table-level checking described in this section still applies. Users are strongly urged not to use database files that contain “Barracuda” file format tables with releases of InnoDB older than the MySQL 5.1 with the InnoDB Plugin. It is possible to “downgrade” such tables to the “Antelope” format with the procedure de- scribed in Section 4.4, “Downgrading the File Format”. 4.3. Identifying the File Format in Use After you enable a given innodb_file_format, this change applies only to newly created tables rather than existing ones. If you do create a new table, the tablespace containing the table is tagged with the “earliest” or “simplest” file format that is required for the table's features. For example, if you enable file format “Barracuda”, and create a new table that is not compressed and does not use ROW_FORMAT=DYNAMIC, the new tablespace that contains the table is tagged as using file format “Antelope”. It is easy to identify the file format used by a given tablespace or table. The table uses the “Barracuda” format if the Row_format reported by SHOW CREATE TABLE or INFORMATION_SCHEMA.TABLES is one of 'Compressed' or 'Dynamic'. (The Row_format is a separate column; ignore the contents of the Create_options column, which may contain the string ROW_FORMAT.) If the table in a tablespace uses neither of those features, the file uses the format supported by prior releases of In- noDB, now called file format “Antelope”. Then, the Row_format is one of 'Redundant' or 'Compact'. Internal Details The file format identifier is written as part of the tablespace flags (a 32-bit number) in the *.ibd file in the 4 bytes starting at posi- tion 54 of the file, most significant byte first. (The first byte of the file is byte zero.) On some systems, you can display these bytes in hexadecimal with the command od -t x1 -j 54 -N 4 tablename.ibd. If all bytes are zero, the tablespace uses the “Antelope” file format (which is the format used by the standard InnoDB storage engine up to version 5.1). Otherwise, the least significant bit should be set in the tablespace flags, and the file format identifier is written in the bits 5 through 11. (Divide the ta- blespace flags by 32 and take the remainder after dividing the integer part of the result by 128.) 4.4. Downgrading the File Format Each InnoDB tablespace file (with a name matching *.ibd) is tagged with the file format used to create its table and indexes. The way to downgrade the tablespace is to re-create the table and its indexes. The easiest way to recreate a table and its indexes is to use the command: ALTER TABLE t ROW_FORMAT=COMPACT; on each table that you want to downgrade. The COMPACT row format uses the file format “Antelope”. It was introduced in MySQL 5.0.3. 4.5. Future InnoDB File Formats The file format used by the standard built-in InnoDB in MySQL 5.1 is the “Antelope” format. The file format introduced with In- noDB Plugin 1.0 is the “Barracuda” format. Although no new features have been announced that would require additional new file formats, the InnoDB file format mechanism allows for future enhancements. For the sake of completeness, these are the file format names that might be used for future file formats: Antelope, Barracuda, Chee- tah, Dragon, Elk, Fox, Gazelle, Hornet, Impala, Jaguar, Kangaroo, Leopard, Moose, Nautilus, Ocelot, Porpoise, Quail, Rabbit, Shark, Tiger, Urchin, Viper, Whale, Xenops, Yak and Zebra. These file formats correspond to the internal identifiers 0..25. 16
- 25. Chapter 5. Storage of Variable-Length Columns This section discusses how certain InnoDB features, such as table compression and off-page storage of long columns, are con- trolled by the ROW_FORMAT clause of the CREATE TABLE statement. 5.1. Overview All data in InnoDB is stored in database pages comprising a B-tree index (the clustered index or primary key index). The nodes of the index data structure contain the values of the key columns for each primary key value, plus the values of the remaining columns of that row. In some other database systems, a clustered index is called an “index-organized table”. Secondary indexes in InnoDB are also B-trees, containing pairs of values of the index key and the value of the primary key, which acts as a pointer to the row in the clustered index. Variable-length columns are an exception to this rule. Such columns, such as BLOB and VARCHAR, that are too long to fit on a B- tree page are stored on separately allocated disk (“overflow”) pages. We call these “off-page columns”. The values of such columns are stored on singly-linked lists of overflow pages, and each such column has its own list of one or more overflow pages. In some cases, all or a prefix of the long column values is stored in the B-tree, to avoid wasting storage and eliminating the need to read a separate page. The “Barracuda” file format provides a new option (KEY_BLOCK_SIZE) to control how much column data is stored in the clustered index, and how much is placed on overflow pages. 5.2. COMPACT and REDUNDANT Row Format Early versions of InnoDB used an unnamed file format (now called “Antelope”) for database files. With that format, tables were defined with ROW_FORMAT=COMPACT (or ROW_FORMAT=REDUNDANT) and InnoDB stored up to the first 768 bytes of variable- length columns (such as BLOB and VARCHAR) in the index record within the B-tree node, with the remainder stored on the over- flow page(s). To preserve compatibility with those prior versions, tables created with the InnoDB storage engine use the prefix format, unless one of ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED is specified (or implied) on the CREATE TABLE statement. With the “Antelope” file format, if the value of a column is 768 bytes or less, no overflow page is needed, and some savings in I/O may result, since the value is in the B-tree node. This works well for relatively short BLOBs, but may cause B-tree nodes to fill with data rather than key values, thereby reducing their efficiency. Tables with many BLOB columns could cause B-tree nodes to be- come too full of data, and contain too few rows, making the entire index less efficient than if the rows were shorter or if the column values were stored off-page. 5.3. DYNAMIC Row Format When innodb_file_format is set to “Barracuda” and a table is created with ROW_FORMAT=DYNAMIC or ROW_FORMAT=COMPRESSED, long column values are stored fully off-page, and the clustered index record contains only a 20-byte pointer to the overflow page. Whether any columns are stored off-page depends on the page size and the total size of the row. When the row is too long, InnoDB chooses the longest columns for off-page storage until the clustered index record fits on the B-tree page. The DYNAMIC row format maintains the efficiency of storing the entire row in the index node if it fits (as do the COMPACT and REDUNDANT formats), but this new format avoids the problem of filling B-tree nodes with a large number of data bytes of long columns. The DYNAMIC format is based on the idea that if a portion of a long data value is stored off-page, it is usually most effi- cient to store all of the value off-page. With DYNAMIC format, shorter columns are likely to remain in the B-tree node, minimizing the number of overflow pages needed for any given row. 5.4. Specifying a Table’s Row Format The row format used for a table is specified with the ROW_FORMAT clause of the CREATE TABLE and ALTER TABLE state- ments. Note that COMPRESSED format implies DYNAMIC format. See Section 3.2, “Enabling Compression for a Table” for more details on the relationship between this clause and other clauses of these commands. 17
- 26. Chapter 6. InnoDB INFORMATION_SCHEMA tables The INFORMATION_SCHEMA is a MySQL feature that helps you monitor server activity to diagnose capacity and performance is- sues. Several InnoDB-related INFORMATION_SCHEMA tables (INNODB_CMP, INNODB_CMP_RESET, INNODB_CMPMEM, IN- NODB_CMPMEM_RESET, INNODB_TRX, INNODB_LOCKS and INNODB_LOCK_WAITS) contain live information about com- pressed InnoDB tables, the compressed InnoDB buffer pool, all transactions currently executing inside InnoDB, the locks that transactions hold and those that are blocking transactions waiting for access to a resource (a table or row). The Information Schema tables are themselves plugins to the MySQL server, and must be activated by INSTALL statements. If they are installed, but the InnoDB storage engine plugin is not installed, these tables appear to be empty. This section describes the InnoDB-related Information Schema tables and shows some examples of their use. 6.1. Information Schema Tables about Compression Two new pairs of Information Schema tables provided by the InnoDB storage engine can give you some insight into how well compression is working overall. One pair of tables contains information about the number of compression operations and the amount of time spent performing compression. Another pair of tables contains information on the way memory is allocated for compression. 6.1.1. INNODB_CMP and INNODB_CMP_RESET The tables INNODB_CMP and INNODB_CMP_RESET contain status information on the operations related to compressed tables, which are covered in Chapter 3, InnoDB Data Compression. The compressed page size is in the column PAGE_SIZE. These two tables have identical contents, but reading from INNODB_CMP_RESET resets the statistics on compression and uncom- pression operations. For example, if you archive the output of INNODB_CMP_RESET every 60 minutes, you see the statistics for each hourly period. If you monitor the output of INNODB_CMP (making sure never to read INNODB_CMP_RESET), you see the cumulated statistics since InnoDB was started. For the table definition, see Columns of INNODB_CMP and INNODB_CMP_RESET. 6.1.2. INNODB_CMPMEM and INNODB_CMPMEM_RESET The tables INNODB_CMPMEM and INNODB_CMPMEM_RESET contain status information on the compressed pages that reside in the buffer pool. Please consult Chapter 3, InnoDB Data Compression for further information on compressed tables and the use of the buffer pool. The tables INNODB_CMP and INNODB_CMP_RESET should provide more useful statistics on compression. Internal Details The InnoDB storage engine uses a so-called “buddy allocator” system to manage memory allocated to pages of various sizes, from 1KB to 16KB. Each row of the two tables described here corresponds to a single page size, except for rows with PAGE_SIZE<1024, which are implementation artifacts. The smallest blocks (PAGE_SIZE=64 or PAGE_SIZE=128, depend- ing on the server platform) are used for keeping track of compressed pages for which no uncompressed page has been allocated in the buffer pool. Other blocks of PAGE_SIZE<1024 should never be allocated (PAGES_USED=0). They exist because the memory allocator allocates smaller blocks by splitting bigger ones into halves. These two tables have identical contents, but reading from INNODB_CMPMEM_RESET resets the statistics on relocation opera- tions. For example, if every 60 minutes you archived the output of INNODB_CMPMEM_RESET, it would show the hourly statistics. If you never read INNODB_CMPMEM_RESET and monitored the output of INNODB_CMPMEM instead, it would show the cumu- lated statistics since InnoDB was started. For the table definition, see Columns of INNODB_CMPMEM and INNODB_CMPMEM_RESET. 6.1.3. Using the Compression Information Schema Tables Example 6.1. Using the Compression Information Schema Tables The following is sample output from a database that contains compressed tables (see Chapter 3, InnoDB Data Compression, IN- NODB_CMP, and INNODB_CMPMEM). The following table shows the contents of INFORMATION_SCHEMA.INNODB_CMP under light load. The only compressed page size that the buffer pool contains is 8K. Compressing or uncompressing pages has consumed less than a second since the time the statistics were reset, because the columns COMPRESS_TIME and UNCOMPRESS_TIME are zero. 18
- 27. InnoDB INFORMATION_SCHEMA tables page size compress ops compress ops ok compress time uncompress ops uncompress time 1024 0 0 0 0 0 2048 0 0 0 0 0 4096 0 0 0 0 0 8192 1048 921 0 61 0 16384 0 0 0 0 0 According to INNODB_CMPMEM, there are 6169 compressed 8KB pages in the buffer pool. The only other allocated block size is 64 bytes. The smallest PAGE_SIZE in INNODB_CMPMEM is used for block descriptors of those compressed pages for which no uncompressed page exists in the buffer pool. We see that there are 5910 such pages. Indirectly, we see that 259 (6169-5910) com- pressed pages also exist in the buffer pool in uncompressed form. The following table shows the contents of INFORMATION_SCHEMA.INNODB_CMPMEM under light load. We can see that some memory is unusable due to fragmentation of the InnoDB memory allocator for compressed pages: SUM(PAGE_SIZE*PAGES_FREE)=6784. This is because small memory allocation requests are fulfilled by splitting bigger blocks, starting from the 16K blocks that are allocated from the main buffer pool, using the buddy allocation system. The fragment- ation is this low, because some allocated blocks have been relocated (copied) to form bigger adjacent free blocks. This copying of SUM(PAGE_SIZE*RELOCATION_OPS) bytes has consumed less than a second (SUM(RELOCATION_TIME)=0). page size pages used pages free relocation ops relocation time 64 5910 0 2436 0 128 0 1 0 0 256 0 0 0 0 512 0 1 0 0 1024 0 0 0 0 2048 0 1 0 0 4096 0 1 0 0 8192 6169 0 5 0 16384 0 0 0 0 6.2. Information Schema Tables about Transactions Three InnoDB-related Information Schema tables make it easy to monitor transactions and diagnose possible locking problems. The three tables are INNODB_TRX, INNODB_LOCKS and INNODB_LOCK_WAITS. 6.2.1. INNODB_TRX Contains information about every transaction currently executing inside InnoDB, including whether the transaction is waiting for a lock, when the transaction started, and the particular SQL statement the transaction is executing. For the table definition, see INNODB_TRX Columns. 6.2.2. INNODB_LOCKS Each transaction in InnoDB that is waiting for another transaction to release a lock (INNODB_TRX.TRX_STATE='LOCK WAIT') is blocked by exactly one “blocking lock request”. That blocking lock request is for a row or table lock held by another transaction in an incompatible mode. The waiting or blocked transaction cannot proceed until the other transaction commits or rolls back, thereby releasing the requested lock. For every blocked transaction, INNODB_LOCKS contains one row that describes each lock the transaction has requested, and for which it is waiting. INNODB_LOCKS also contains one row for each lock that is block- ing another transaction, whatever the state of the transaction that holds the lock ('RUNNING', 'LOCK WAIT', 'ROLLING BACK' or 'COMMITTING'). The lock that is blocking a transaction is always held in a mode (read vs. write, shared vs. exclusive) incompatible with the mode of requested lock. For the table definition, see INNODB_LOCKS Columns. 6.2.3. INNODB_LOCK_WAITS Using this table, you can tell which transactions are waiting for a given lock, or for which lock a given transaction is waiting. This 19
- 28. InnoDB INFORMATION_SCHEMA tables table contains one or more rows for each blocked transaction, indicating the lock it has requested and the lock(s) that is (are) block- ing that request. The REQUESTED_LOCK_ID refers to the lock that a transaction is requesting, and the BLOCKING_LOCK_ID refers to the lock (held by another transaction) that is preventing the first transaction from proceeding. For any given blocked trans- action, all rows in INNODB_LOCK_WAITS have the same value for REQUESTED_LOCK_ID and different values for BLOCK- ING_LOCK_ID. For the table definition, see INNODB_LOCK_WAITS Columns. 6.2.4. Using the Transaction Information Schema Tables Example 6.2. Identifying Blocking Transactions It is sometimes helpful to be able to identify which transaction is blocking another. You can use the Information Schema tables to find out which transaction is waiting for another, and which resource is being requested. Suppose you have the following scenario, with three users running concurrently. Each user (or session) corresponds to a MySQL thread, and executes one transaction after another. Consider the state of the system when these users have issued the following commands, but none has yet committed its transaction: • User A: BEGIN; SELECT a FROM t FOR UPDATE; SELECT SLEEP(100); • User B: SELECT b FROM t FOR UPDATE; • User C: SELECT c FROM t FOR UPDATE; In this scenario, you may use this query to see who is waiting for whom: SELECT r.trx_id waiting_trx_id, r.trx_mysql_thread_id waiting_thread, r.trx_query waiting_query, b.trx_id blocking_trx_id, b.trx_mysql_thread_id blocking_thread, b.trx_query blocking_query FROM information_schema.innodb_lock_waits w INNER JOIN information_schema.innodb_trx b ON b.trx_id = w.blocking_trx_id INNER JOIN information_schema.innodb_trx r ON r.trx_id = w.requesting_trx_id; waiting waiting waiting query blocking blocking blocking query trx id thread trx id thread A4 6 SELECT b FROM t FOR UPDATE A3 5 SELECT SLEEP(100) A5 7 SELECT c FROM t FOR UPDATE A3 5 SELECT SLEEP(100) A5 7 SELECT c FROM t FOR UPDATE A4 6 SELECT b FROM t FOR UPDATE In the above result, you can identify users by the “waiting query” or “blocking query”. As you can see: • User B (trx id 'A4', thread 6) and User C (trx id 'A5', thread 7) are both waiting for User A (trx id 'A3', thread 5). • User C is waiting for User B as well as User A. You can see the underlying data in the tables INNODB_TRX, INNODB_LOCKS, and INNODB_LOCK_WAITS. The following table shows some sample Contents of INFORMATION_SCHEMA.INNODB_TRX. 20
- 29. InnoDB INFORMATION_SCHEMA tables trx trx trx started trx requested trx wait started trx trx mysql trx query id state lock id weight thread id A3 RUN- 2008-01-15 NULL NULL 2 5 SELECT SLEEP(100) NING 16:44:54 A4 LOCK 2008-01-15 A4:1:3:2 2008-01-15 2 6 SELECT b FROM t WAIT 16:45:09 16:45:09 FOR UPDATE A5 LOCK 2008-01-15 A5:1:3:2 2008-01-15 2 7 SELECT c FROM t WAIT 16:45:14 16:45:14 FOR UPDATE The following table shows some sample contents of INFORMATION_SCHEMA.INNODB_LOCKS. lock id lock lock lock type lock table lock index lock lock lock lock data trx id mode space page rec A3:1:3:2 A3 X RECORD `test`.`t` `PRIMARY` 1 3 2 0x0200 A4:1:3:2 A4 X RECORD `test`.`t` `PRIMARY` 1 3 2 0x0200 A5:1:3:2 A5 X RECORD `test`.`t` `PRIMARY` 1 3 2 0x0200 The following table shows some sample contents of INFORMATION_SCHEMA.INNODB_LOCK_WAITS. requesting trx id requested lock id blocking trx id blocking lock id A4 A4:1:3:2 A3 A3:1:3:2 A5 A5:1:3:2 A3 A3:1:3:2 A5 A5:1:3:2 A4 A4:1:3:2 Example 6.3. More Complex Example of Transaction Data in Information Schema Tables Sometimes you would like to correlate the internal InnoDB locking information with session-level information maintained by MySQL. For example, you might like to know, for a given InnoDB transaction ID, the corresponding MySQL session ID and name of the user that may be holding a lock, and thus blocking another transaction. The following output from the INFORMATION_SCHEMA tables is taken from a somewhat loaded system. As can be seen in the following tables, there are several transactions running. The following INNODB_LOCKS and INNODB_LOCK_WAITS tables shows that: • Transaction 77F (executing an INSERT) is waiting for transactions 77E, 77D and 77B to commit. • Transaction 77E (executing an INSERT) is waiting for transactions 77D and 77B to commit. • Transaction 77D (executing an INSERT) is waiting for transaction 77B to commit. • Transaction 77B (executing an INSERT) is waiting for transaction 77A to commit. • Transaction 77A is running, currently executing SELECT. • Transaction E56 (executing an INSERT) is waiting for transaction E55 to commit. • Transaction E55 (executing an INSERT) is waiting for transaction 19C to commit. • Transaction 19C is running, currently executing an INSERT. Note that there may be an inconsistency between queries shown in the two tables INNODB_TRX.TRX_QUERY and PROCESS- LIST.INFO. The current transaction ID for a thread, and the query being executed in that transaction, may be different in these two tables for any given thread. See Section 6.3.3, “Possible Inconsistency with PROCESSLIST” for an explanation. The following table shows the contents of INFORMATION_SCHEMA.PROCESSLIST in a loaded system. 21
- 30. InnoDB INFORMATION_SCHEMA tables ID USER HOST DB COMMAND TIME STATE INFO 384 root localhost test Query 10 update insert into t2 values … 257 root localhost test Query 3 update insert into t2 values … 130 root localhost test Query 0 update insert into t2 values … 61 root localhost test Query 1 update insert into t2 values … 8 root localhost test Query 1 update insert into t2 values … 4 root localhost test Query 0 preparing SELECT * FROM processlist 2 root localhost test Sleep 566 NULL The following table shows the contents of INFORMATION_SCHEMA.INNODB_TRX in a loaded system. trx id trx state trx started trx requested trx wait started trx trx mysql trx query lock id weight thread id 77F LOCK 2008-01-15 77F:806 2008-01-15 1 876 insert into t09 (D, WAIT 13:10:16 13:10:16 B, C) values … 77E LOCK 2008-01-15 77E:806 2008-01-15 1 875 insert into t09 (D, WAIT 13:10:16 13:10:16 B, C) values … 77D LOCK 2008-01-15 77D:806 2008-01-15 1 874 insert into t09 (D, WAIT 13:10:16 13:10:16 B, C) values … 77B LOCK 2008-01-15 77B:733:12:1 2008-01-15 4 873 insert into t09 (D, WAIT 13:10:16 13:10:16 B, C) values … 77A RUN-NI 2008-01-15 NULL NULL 4 872 select b, c from NG 13:10:16 t09 where … E56 LOCK 2008-01-15 E56:743:6:2 2008-01-15 5 384 insert into t2 val- WAIT 13:10:06 13:10:06 ues … E55 LOCK 2008-01-15 E55:743:38:2 2008-01-15 965 257 insert into t2 val- WAIT 13:10:06 13:10:13 ues … 19C RUN-NI 2008-01-15 NULL NULL 2900 130 insert into t2 val- NG 13:09:10 ues … E15 RUN-NI 2008-01-15 NULL NULL 5395 61 insert into t2 val- NG 13:08:59 ues … 51D RUN-NI 2008-01-15 NULL NULL 9807 8 insert into t2 val- NG 13:08:47 ues … The following table shows the contents of INFORMATION_SCHEMA.INNODB_LOCK_WAITS in a loaded system requesting trx id requested lock id blocking trx id blocking lock id 77F 77F:806 77E 77E:806 77F 77F:806 77D 77D:806 77F 77F:806 77B 77B:806 77E 77E:806 77D 77D:806 77E 77E:806 77B 77B:806 77D 77D:806 77B 77B:806 77B 77B:733:12:1 77A 77A:733:12:1 E56 E56:743:6:2 E55 E55:743:6:2 E55 E55:743:38:2 19C 19C:743:38:2 The following table shows the contents of INFORMATION_SCHEMA.INNODB_LOCKS in a loaded system. 22
- 31. InnoDB INFORMATION_SCHEMA tables lock id lock lock lock type lock table lock index lock lock lock lock data trx id mode space page rec 77F:806 77F AUTO TABLE `test` NULL NULL NULL NULL NULL _INC .`t09` 77E:806 77E AUTO TABLE `test` NULL NULL NULL NULL NULL _INC .`t09` 77D:806 77D AUTO TABLE `test` NULL NULL NULL NULL NULL _INC .`t09` 77B:806 77B AUTO TABLE `test` NULL NULL NULL NULL NULL _INC .`t09` 77B:733:12:1 77B X RECORD `test` `PRIMARY` 733 12 1 supremum .`t09` pseudo-re- cord 77A:733:12:1 77A X RECORD `test` `PRIMARY` 733 12 1 supremum .`t09` pseudo-re- cord E56:743:6:2 E56 S RECORD `test` `PRIMARY` 743 6 2 0, 0 .`t2` E55:743:6:2 E55 X RECORD `test` `PRIMARY` 743 6 2 0, 0 .`t2` E55:743:38:2 E55 S RECORD `test` `PRIMARY` 743 38 2 1922, 1922 .`t2` 19C:743:38:2 19C X RECORD `test` `PRIMARY` 743 38 2 1922, 1922 .`t2` 6.3. Notes on Locking in InnoDB 6.3.1. Understanding InnoDB Locking When a transaction updates a row in a table, or locks it with SELECT FOR UPDATE, InnoDB establishes a list or queue of locks on that row. Similarly, InnoDB maintains a list of locks on a table for table-level locks transactions hold. If a second transaction wants to update a row or lock a table already locked by a prior transaction in an incompatible mode, InnoDB adds a lock request for the row to the corresponding queue. For a lock to be acquired by a transaction, all incompatible lock requests previously entered in- to the lock queue for that row or table must be removed (the transactions holding or requesting those locks either commit or roll- back). A transaction may have any number of lock requests for different rows or tables. At any given time, a transaction may be request- ing a lock that is held by another transaction, in which case it is blocked by that other transaction. The requesting transaction must wait for the transaction that holds the blocking lock to commit or rollback. If a transaction is not waiting for a a lock, it is in the 'RUNNING' state. If a transaction is waiting for a lock, it is in the 'LOCK WAIT' state. The table INNODB_LOCKS holds one or more row for each 'LOCK WAIT' transaction, indicating the lock request(s) that is (are) preventing its progress. This table also contains one row describing each lock in a queue of locks pending for a given row or table. The table INNODB_LOCK_WAITS shows which locks already held by a transaction are blocking locks requested by other transac- tions. 6.3.2. Granularity of INFORMATION_SCHEMA Data The data exposed by the transaction and locking tables represent a glimpse into fast-changing data. This is not like other (user) tables, where the data only changes when application-initiated updates occur. The underlying data is internal system-managed data, and can change very quickly. For performance reasons, and to minimize the chance of misleading JOINs between the INFORMATION_SCHEMA tables, InnoDB collects the required transaction and locking information into an intermediate buffer whenever a SELECT on any of the tables is is- sued. This buffer is refreshed only if more than 0.1 seconds has elapsed since the last time the buffer was used. The data needed to fill the three tables is fetched atomically and consistently and is saved in this global internal buffer, forming a point-in-time “snapshot”. If multiple table accesses occur within 0.1 seconds (as they almost certainly do when MySQL processes a join among these tables), then the same snapshot is used to satisfy the query. A correct result is returned when you JOIN any of these tables together in a single query, because the data for the three tables comes from the same snapshot. Because the buffer is not refreshed with every query of any of these tables, if you issue separate 23
- 32. InnoDB INFORMATION_SCHEMA tables queries against these tables within a tenth of a second, the results are the same from query to query. On the other hand, two separate queries of the same or different tables issued more than a tenth of a second apart may see different results, since the data come from different snapshots. Because InnoDB must temporarily stall while the transaction and locking data is collected, too frequent queries of these tables can negatively impact performance as seen by other users. As these tables contain sensitive information (at least INNODB_LOCKS.LOCK_DATA and INNODB_TRX.TRX_QUERY), for se- curity reasons, only the users with the PROCESS privilege are allowed to SELECT from them. 6.3.3. Possible Inconsistency with PROCESSLIST As just described, while the transaction and locking data is correct and consistent when these INFORMATION_SCHEMA tables are populated, the underlying data changes so fast that similar glimpses at other, similarly fast-changing data, may not be in sync. Thus, you should be careful in comparing the data in the InnoDB transaction and locking tables with that in the MySQL table PROCESS- LIST. The data from the PROCESSLIST table does not come from the same snapshot as the data about locking and transactions. Even if you issue a single SELECT (JOINing INNODB_TRX and PROCESSLIST, for example), the content of those tables is gen- erally not consistent. INNODB_TRX may reference rows that are not present in PROCESSLIST or the currently executing SQL query of a transaction, shown in INNODB_TRX.TRX_QUERY may be different from the one in PROCESSLIST.INFO. The query in INNODB_TRX is always consistent with the rest of INNODB_TRX, INNODB_LOCKS and INNODB_LOCK_WAITS when the data comes from the same snapshot. 24
- 33. Chapter 7. Performance and Scalability Enhancements This section discusses recent InnoDB enhancements to performance and scalability, covering the performance features in InnoDB 1.1 with MySQL 5.5, and the features in the InnoDB Plugin for MySQL 5.1. This information is useful to any DBA or developer who is concerned with performance and scalability. Although some of the enhancements do not require any action on your part, knowing this information can still help you diagnose performance issues more quickly and modernize systems and applications that rely on older, inefficient behavior. 7.1. Overview of InnoDB Performance InnoDB has always been highly efficient, and includes several unique architectural elements to assure high performance and scalability. The latest InnoDB storage engine includes new features that take advantage of advances in operating systems and hard- ware platforms, such as multi-core processors and improved memory allocation systems. In addition, new configuration options let you better control some InnoDB internal subsystems to achieve the best performance with your workload. Starting with MySQL 5.5 and InnoDB 1.1, the Plugin replaces the built-in InnoDB storage engine within MySQL. This change makes these performance and scalability enhancements available to a much wider audience than before. After learning about the In- noDB performance features in this section, continue with Optimization to learn the best practices for overall MySQL performance, and Optimizing for InnoDB Tables in particular for InnoDB tips and guidelines. 7.2. Faster Locking for Improved Scalability In MySQL and InnoDB, multiple threads of execution access shared data structures. InnoDB synchronizes these accesses with its own implementation of mutexes and read/write locks. InnoDB has historically protected the internal state of a read/write lock with an InnoDB mutex. On Unix and Linux platforms, the internal state of an InnoDB mutex is protected by a Pthreads mutex, as in IEEE Std 1003.1c (POSIX.1c). On many platforms, there is a more efficient way to implement mutexes and read/write locks. Atomic operations can often be used to synchronize the actions of multiple threads more efficiently than Pthreads. Each operation to acquire or release a lock can be done in fewer CPU instructions, and thus result in less wasted time when threads are contending for access to shared data struc- tures. This in turn means greater scalability on multi-core platforms. Beginning with InnoDB 1.0.3, InnoDB implements mutexes and read/write locks with the built-in functions provided by the GNU Compiler Collection (GCC) for atomic memory access instead of using the Pthreads approach previously used. More specifically, an InnoDB that is compiled with GCC version 4.1.2 or later uses the atomic builtins instead of a pthread_mutex_t to imple- ment InnoDB mutexes and read/write locks. On 32-bit Microsoft Windows, InnoDB has implemented mutexes (but not read/write locks) with hand-written assembler instruc- tions. Beginning with Microsoft Windows 2000, it is possible to use functions for Interlocked Variable Access that are similar to the built-in functions provided by GCC. Beginning with InnoDB 1.0.4, InnoDB makes use of the Interlocked functions on Win- dows. Unlike the old hand-written assembler code, the new implementation supports read/write locks and 64-bit platforms. Solaris 10 introduced library functions for atomic operations. Beginning with InnoDB 1.0.4, when InnoDB is compiled on Solaris 10 with a compiler that does not support the built-in functions provided by the GNU Compiler Collection (GCC) for atomic memory access, the library functions are used. This change improves the scalability of InnoDB on multi-core systems. Note that the user does not have to set any particular para- meter or option to take advantage of this new feature. This feature is enabled out-of-the-box on the platforms where it is supported. On platforms where the GCC, Windows, or Solaris functions for atomic memory access are not available, InnoDB uses the tradi- tional Pthreads method of implementing mutexes and read/write locks. When MySQL starts, InnoDB writes a message to the log file indicating whether atomic memory access is used for mutexes, for mutexes and read/write locks, or neither. If suitable tools are used to build InnoDB and the target CPU supports the atomic opera- tions required, InnoDB uses the built-in functions for mutexing. If, in addition, the compare-and-swap operation can be used on thread identifiers (pthread_t), then InnoDB uses the instructions for read-write locks as well. Note: If you are building from source, ensure that the build process properly takes advantage of your platform capabilities. If the build is not able to automatically use instructions for atomic memory access where available, consult this Support Tip on the In- noDB website for additional steps. For more information about the performance implications of locking, see Optimizing Locking Operations. 7.3. Using Operating System Memory Allocators When InnoDB was developed, the memory allocators supplied with operating systems and run-time libraries were often lacking in performance and scalability. At that time, there were no memory allocator libraries tuned for multi-core CPUs. Therefore, InnoDB implemented its own memory allocator in the mem subsystem. This allocator is guarded by a single mutex, which may become a 25
- 34. Performance and Scalability Enhancements bottleneck. InnoDB also implements a wrapper interface around the system allocator (malloc and free) that is likewise guarded by a single mutex. Today, as multi-core systems have become more widely available, and as operating systems have matured, significant improve- ments have been made in the memory allocators provided with operating systems. New memory allocators perform better and are more scalable than they were in the past. The leading high-performance memory allocators include Hoard, libumem, mtmal- loc, ptmalloc, tbbmalloc, and TCMalloc. Most workloads, especially those where memory is frequently allocated and re- leased (such as multi-table joins), benefit from using a more highly tuned memory allocator as opposed to the internal, InnoDB-spe- cific memory allocator. Beginning with InnoDB 1.0.3, you control whether InnoDB uses its own memory allocator or an allocator of the operating system, by setting the value of the new system configuration parameter innodb_use_sys_malloc in the MySQL option file (my.cnf or my.ini). If set to ON or 1 (the default), InnoDB uses the malloc and free functions of the underlying system rather than manage memory pools itself. This parameter is not dynamic, and takes effect only when the system is started. To continue to use the InnoDB memory allocator, set innodb_use_sys_malloc to 0. Note that when the InnoDB memory allocator is disabled, InnoDB ignores the value of the parameter in- nodb_additional_mem_pool_size. The InnoDB memory allocator uses an additional memory pool for satisfying alloca- tion requests without having to fall back to the system memory allocator. When the InnoDB memory allocator is disabled, all such allocation requests are fulfilled by the system memory allocator. Furthermore, since InnoDB cannot track all memory use when the system memory allocator is used (in- nodb_use_sys_malloc is ON), the section “BUFFER POOL AND MEMORY” in the output of the SHOW ENGINE INNODB STATUS command only includes the buffer pool statistics in the “Total memory allocated”. Any memory allocated using the mem subsystem or using ut_malloc is excluded. On Unix-like systems that use dynamic linking, replacing the memory allocator may be as easy as making the environment variable LD_PRELOAD or LD_LIBRARY_PATH point to the dynamic library that implements the allocator. On other systems, some relink- ing may be necessary. Please refer to the documentation of the memory allocator library of your choice. For more information about the performance implications of InnoDB memory usage, see Buffering and Caching. 7.4. Controlling InnoDB Change Buffering When INSERT, UPDATE, and DELETE operations are done to a table, often the values of indexed columns (particularly the values of secondary keys) are not in sorted order, requiring substantial I/O to bring secondary indexes up to date. InnoDB has an insert buffer that caches changes to secondary index entries when the relevant page is not in the buffer pool, thus avoiding I/O operations by not reading in the page from the disk. The buffered changes are merged when the page is loaded to the buffer pool, and the up- dated page is later flushed to disk using the normal mechanism. The InnoDB main thread merges buffered changes when the server is nearly idle, and during a slow shutdown. Because it can result in fewer disk reads and writes, this feature is most valuable for workloads that are I/O-bound, for example ap- plications with a high volume of DML operations such as bulk inserts. However, the insert buffer occupies a part of the buffer pool, reducing the memory available to cache data pages. If the working set almost fits in the buffer pool, or if your tables have relatively few secondary indexes, it may be useful to disable insert buffering. If the working set entirely fits in the buffer pool, insert buffering does not impose any extra overhead, because it only applies to pages that are not in the buffer pool. Beginning with InnoDB 1.0.3, you can control whether InnoDB performs insert buffering with the system configuration parameter innodb_change_buffering. Beginning with InnoDB 1.1 with MySQL 5.5, the buffering support is expanded to include de- lete operations (when index records are initially marked for deletion) and purge operations (when index records are physically de- leted). InnoDB 1.1 also changes the default value from inserts to all. The allowed values of innodb_change_buffering are: • all The default value: buffer inserts, delete-marking operations, and purges. • none Do not buffer any operations. • inserts Buffer insert operations. • deletes 26
- 35. Performance and Scalability Enhancements Buffer delete-marking operations. • changes Buffer both inserts and delete-marking. • purges Buffer the physical deletion operations that happen in the background. You can set the value of this parameter in the MySQL option file (my.cnf or my.ini) or change it dynamically with the SET GLOBAL command, which requires the SUPER privilege. Changing the setting affects the buffering of new operations; the merging of already buffered entries is not affected. For more information about speeding up INSERT, UPDATE, and DELETE statements, see Optimizing DML Statements. 7.5. Controlling Adaptive Hash Indexing If a table fits almost entirely in main memory, the fastest way to perform queries on it is to use hash indexes rather than B-tree lookups. InnoDB monitors searches on each index defined for a table. If it notices that certain index values are being accessed fre- quently, it automatically builds an in-memory hash table for that index. Based on the pattern of searches that InnoDB observes, it builds a hash index using a prefix of the index key. The prefix of the key can be any length, and it may be that only some of the val- ues in the B-tree appear in the hash index. InnoDB builds hash indexes on demand for those pages of the index that are often ac- cessed. The adaptive hash index mechanism allows InnoDB to take advantage of large amounts of memory, something typically done only by database systems specifically designed for databases that reside entirely in memory. Normally, the automatic building and use of adaptive hash indexes improves performance. However, sometimes, the read/write lock that guards access to the adaptive hash in- dex may become a source of contention under heavy workloads, such as multiple concurrent joins. You can monitor the use of the adaptive hash index and the contention for its use in the “SEMAPHORES” section of the output of the SHOW ENGINE INNODB STATUS command. If you see many threads waiting on an RW-latch created in btr0sea.c, then it might be useful to disable adaptive hash indexing. The configuration parameter innodb_adaptive_hash_index can be set to disable or enable the adaptive hash index. See Section 8.2.4, “Dynamically Changing innodb_adaptive_hash_index” for details. For more information about the performance characteristics of hash indexes, see Comparison of B-Tree and Hash Indexes. 7.6. Changes Regarding Thread Concurrency InnoDB uses operating system threads to process requests from user transactions. (Transactions may issue many requests to In- noDB before they commit or roll back.) On modern operating systems and servers with multi-core processors, where context switching is efficient, most workloads run well without any limit on the number of concurrent threads. Scalability improvements in InnoDB 1.0.3 and up reduce the need to limit the number of concurrently executing threads inside InnoDB. However, for some situations, it may be helpful to minimize context switching between threads. InnoDB can use a number of tech- niques to limit the number of concurrently executing operating system threads (and thus the number of requests that are processed at any one time). When InnoDB receives a new request from a user session, if the number of threads concurrently executing is at a pre-defined limit, the new request sleeps for a short time before it tries again. A request that cannot be rescheduled after the sleep is put in a first-in/first-out queue and eventually is processed. Threads waiting for locks are not counted in the number of concurrently executing threads. The limit on the number of concurrent threads is given by the settable global variable innodb_thread_concurrency. Once the number of executing threads reaches this limit, additional threads sleep for a number of microseconds, set by the system config- uration parameter innodb_thread_sleep_delay, before being placed into the queue. The default value for innodb_thread_concurrency and the implied default limit on the number of concurrent threads has been changed in various releases of MySQL and InnoDB. Starting with InnoDB 1.0.3, the default value of in- nodb_thread_concurrency is 0, so that by default there is no limit on the number of concurrently executing threads, as shown in Table 7.1, “Changes to innodb_thread_concurrency”. Table 7.1. Changes to innodb_thread_concurrency 27
- 36. Performance and Scalability Enhancements InnoDB Version MySQL Version Default Default limit of con- Value to allow unlim- value current threads ited threads Built-in Earlier than 5.1.11 20 No limit 20 or higher Built-in 5.1.11 and newer 8 8 0 InnoDB before 1.0.3 (corresponding to Plugin) 8 8 0 InnoDB 1.0.3 and newer (corresponding to Plugin) 0 No limit 0 Note that InnoDB causes threads to sleep only when the number of concurrent threads is limited. When there is no limit on the number of threads, all contend equally to be scheduled. That is, if innodb_thread_concurrency is 0, the value of in- nodb_thread_sleep_delay is ignored. When there is a limit on the number of threads, InnoDB reduces context switching overhead by permitting multiple requests made during the execution of a single SQL statement to enter InnoDB without observing the limit set by in- nodb_thread_concurrency. Since a SQL statement (such as a join) may comprise multiple row operations within InnoDB, InnoDB assigns “tickets” that allow a thread to be scheduled repeatedly with minimal overhead. When a new SQL statement starts, a thread has no tickets, and it must observe innodb_thread_concurrency. Once the thread is entitled to enter InnoDB, it is assigned a number of tickets that it can use for subsequently entering InnoDB. If the tickets run out, innodb_thread_concurrency is observed again and further tickets are assigned. The number of tickets to assign is specified by the global option innodb_concurrency_tickets, which is 500 by default. A thread that is waiting for a lock is given one ticket once the lock becomes available. The correct values of these variables depend on your environment and workload. Try a range of different values to determine what value works for your applications. Before limiting the number of concurrently executing threads, review configuration options that may improve the performance of InnoDB on multi-core and multi-processor computers, such as innodb_use_sys_malloc and in- nodb_adaptive_hash_index. For general performance information about MySQL thread handling, see How MySQL Uses Threads for Client Connections. 7.7. Changes in the Read-Ahead Algorithm A read-ahead request is an I/O request to prefetch multiple pages in the buffer cache asynchronously, in anticipation that these pages will be needed soon. InnoDB has historically used two read-ahead algorithms to improve I/O performance. Random read-ahead is done if a certain number of pages from the same extent (64 consecutive pages) are found in the buffer cache. In such cases, InnoDB asynchronously issues a request to prefetch the remaining pages of the extent. Random read-ahead added unnecessary complexity to the InnoDB code and often resulted in performance degradation rather than improvement. Start- ing with InnoDB 1.0.4, this feature has been removed from InnoDB, and users should generally see equivalent or improved per- formance. Linear read-ahead is based on the access pattern of the pages in the buffer cache, not just their number. In releases before 1.0.4, if most pages belonging to some extent are accessed sequentially, InnoDB issues an asynchronous prefetch request for the entire next extent when it reads in the last page of the current extent. Beginning with InnoDB 1.0.4, users can control when InnoDB performs a read-ahead operation, by adjusting the number of sequential page accesses required to trigger an asynchronous read request using the new configuration parameter innodb_read_ahead_threshold. If the number of pages read from an extent of 64 pages is greater or equal to innodb_read_ahead_threshold, InnoDB initi- ates an asynchronous read-ahead operation of the entire following extent. Thus, this parameter controls how sensitive InnoDB is to the pattern of page accesses within an extent in deciding whether to read the following extent asynchronously. The higher the value, the more strict the access pattern check. For example, if you set the value to 48, InnoDB triggers a linear read-ahead request only when 48 pages in the current extent have been accessed sequentially. If the value is 8, InnoDB would trigger an asynchronous read-ahead even if as few as 8 pages in the extent were accessed sequentially. The new configuration parameter innodb_read_ahead_threshold may be set to any value from 0-64. The default value is 56, meaning that an asynchronous read-ahead is performed only when 56 of the 64 pages in the extent are accessed sequentially. You can set the value of this parameter in the MySQL option file (my.cnf or my.ini), or change it dynamically with the SET GLOBAL command, which requires the SUPER privilege. Starting with InnoDB 1.0.5 more statistics are provided through SHOW ENGINE INNODB STATUS command to measure the ef- fectiveness of the read-ahead algorithm. See Section 8.8, “More Read-Ahead Statistics” for more information. For more information about I/O performance, see Optimizing InnoDB Disk I/O and Optimizing Disk I/O. 7.8. Multiple Background I/O Threads InnoDB uses background threads to service various types of I/O requests. Starting from InnoDB 1.0.4, the number of background threads tasked with servicing read and write I/O on data pages is configurable. In previous versions of InnoDB, there was only one 28
- 37. Performance and Scalability Enhancements thread each for read and write on non-Windows platforms. On Windows, the number of background threads was controlled by in- nodb_file_io_threads. The configuration parameter innodb_file_io_threads has been removed in InnoDB 1.0.4. If you try to set a value for this parameter, a warning is written to the log file and the value is ignored. In place of innodb_file_io_threads, two new configuration parameters are introduced in InnoDB 1.0.4, which are effect- ive on all supported platforms. The two parameters innodb_read_io_threads and innodb_write_io_threads signify the number of background threads used for read and write requests respectively. You can set the value of these parameters in the MySQL option file (my.cnf or my.ini). These parameters cannot be changed dynamically. The default value for these paramet- ers is 4 and the permissible values range from 1-64. The purpose of this change is to make InnoDB more scalable on high end systems. Each background thread can handle up to 256 pending I/O requests. A major source of background I/O is the read-ahead requests. InnoDB tries to balance the load of incoming requests in such way that most of the background threads share work equally. InnoDB also attempts to allocate read requests from the same extent to the same thread to increase the chances of coalescing the requests together. If you have a high end I/O subsystem and you see more than 64 × innodb_read_io_threads pending read requests in SHOW ENGINE INNODB STATUS, you might gain by increasing the value of innodb_read_io_threads. For more information about InnoDB I/O performance, see Optimizing InnoDB Disk I/O. 7.9. Asynchronous I/O on Linux Starting in InnoDB 1.1 with MySQL 5.5, the asynchronous I/O capability that has been supported on Windows systems, is now available on Linux systems. (Other Unix-like systems continue to use synchronous I/O calls.) This feature improves the scalability of heavily I/O-bound systems, which typically show many pending reads/writes in the output of the command SHOW ENGINE INNODB STATUSG. If a problem with the asynchronous I/O subsystem in the OS prevents InnoDB from starting, set the option in- nodb_use_native_aio=0 in the configuration file. This new configuration option applies to Linux systems only, and cannot be changed once the server is running. For more information about InnoDB I/O performance, see Optimizing InnoDB Disk I/O. 7.10. Group Commit InnoDB, like any other ACID-compliant database engine, flushes the redo log of a transaction before it is committed. Historically, InnoDB used group commit functionality to group multiple such flush requests together to avoid one flush for each commit. With group commit, InnoDB can issue a single write to the log file to perform the commit action for multiple user transactions that com- mit at about the same time, significantly improving throughput. Group commit in InnoDB worked until MySQL 4.x. With the introduction of support for the distributed transactions and Two Phase Commit (2PC) in MySQL 5.0, group commit functionality inside InnoDB was broken. Beginning with InnoDB 1.0.4, the group commit functionality inside InnoDB works with the Two Phase Commit protocol in MySQL. Re-enabling of the group commit functionality fully ensures that the ordering of commit in the MySQL binlog and the In- noDB logfile is the same as it was before. It means it is totally safe to use MySQL Enterprise Backup with InnoDB 1.0.4 and above. When the binlog is enabled, you typically also set the configuration option sync_binlog=0, because group commit for the binary log is only supported if it is set to 0. Group commit is transparent to the user and nothing needs to be done by the user to take advantage of this significant performance improvement. For more information about performance of COMMIT and other transactional operations, see Optimizing InnoDB Transaction Management. 7.11. Controlling the Master Thread I/O Rate The master thread in InnoDB is a thread that performs various tasks in the background. Most of these tasks are I/O related, such as flushing dirty pages from the buffer cache or writing changes from the insert buffer to the appropriate secondary indexes. The mas- ter thread attempts to perform these tasks in a way that does not adversely affect the normal working of the server. It tries to estim- ate the free I/O bandwidth available and tune its activities to take advantage of this free capacity. Historically, InnoDB has used a hard coded value of 100 IOPs (input/output operations per second) as the total I/O capacity of the server. Beginning with InnoDB 1.0.4, a new configuration parameter indicates the overall I/O capacity available to InnoDB. The new para- meter innodb_io_capacity should be set to approximately the number of I/O operations that the system can perform per second. The value depends on your system configuration. When innodb_io_capacity is set, the master threads estimates the I/O bandwidth available for background tasks based on the set value. Setting the value to 100 reverts to the old behavior. You can set the value of innodb_io_capacity to any number 100 or greater, and the default value is 200. You can set the 29
- 38. Performance and Scalability Enhancements value of this parameter in the MySQL option file (my.cnf or my.ini) or change it dynamically with the SET GLOBAL com- mand, which requires the SUPER privilege. For more information about InnoDB I/O performance, see Optimizing InnoDB Disk I/O. 7.12. Controlling the Flushing Rate of Dirty Pages InnoDB performs certain tasks in the background, including flushing of dirty pages (those pages that have been changed but are not yet written to the database files) from the buffer cache, a task performed by the “master thread”. Currently, the master thread ag- gressively flushes buffer pool pages if the percentage of dirty pages in the buffer pool exceeds in- nodb_max_dirty_pages_pct. This behavior can cause temporary reductions in throughput when excessive buffer pool flushing takes place, limiting the I/O capa- city available for ordinary read and write activity. Beginning with release 1.0.4, InnoDB uses a new algorithm to estimate the re- quired rate of flushing based on the speed of redo log generation and the current rate of flushing. The intent of this change is to smooth overall performance, eliminating steep dips in throughput, by ensuring that buffer flush activity keeps up with the need to keep the buffer pool “clean”. InnoDB uses its log files in a circular fashion. Before reusing a portion of a log file, InnoDB flushes to disk all dirty buffer pool pages whose redo entries are contained in that portion of the log file, a process known as a “sharp checkpoint”. If a workload is write intensive, it generates a lot of redo information, all written to the log file. If all available space in the log files is used up, a sharp checkpoint occurs, causing a temporary reduction in throughput. This situation can happen even though in- nodb_max_dirty_pages_pct is not reached. Beginning with release 1.0.4, InnoDB uses a heuristic-based algorithm to avoid such a scenario, by measuring the number of dirty pages in the buffer cache and the rate at which redo is being generated. Based on these numbers, the master thread decides how many dirty pages to flush from the buffer cache each second. This self-adapting algorithm is able to deal with sudden changes in the workload. The primary aim of this feature is to smooth out I/O activity, avoiding sudden dips in throughput when flushing activity becomes high. Internal benchmarking has also shown that this algorithm not only maintains throughput over time, but can also improve overall throughput significantly. Because adaptive flushing is a new feature that can significantly affect the I/O pattern of a workload, a new configuration paramet- er lets you turn off this feature. The default value of the new boolean parameter innodb_adaptive_flushing is TRUE, en- abling the new algorithm. You can set the value of this parameter in the MySQL option file (my.cnf or my.ini) or change it dy- namically with the SET GLOBAL command, which requires the SUPER privilege. For more information about InnoDB I/O performance, see Optimizing InnoDB Disk I/O. 7.13. Using the PAUSE instruction in InnoDB spin loops Synchronization inside InnoDB frequently involves the use of spin loops: while waiting, InnoDB executes a tight loop of instruc- tions repeatedly to avoid having the InnoDB process and threads be rescheduled by the operating system. If the spin loops are ex- ecuted too quickly, system resources are wasted, imposing a performance penalty on transaction throughput. Most modern pro- cessors implement the PAUSE instruction for use in spin loops, so the processor can be more efficient. Beginning with 1.0.4, InnoDB uses a PAUSE instruction in its spin loops on all platforms where such an instruction is available. This technique increases overall performance with CPU-bound workloads, and has the added benefit of minimizing power con- sumption during the execution of the spin loops. You do not have to do anything to take advantage of this performance improvement. For performance considerations for InnoDB locking operations, see Optimizing Locking Operations. 7.14. Control of Spin Lock Polling Many InnoDB mutexes and rw-locks are reserved for a short time. On a multi-core system, it can be more efficient for a thread to continuously check if it can acquire a mutex or rw-lock for a while before sleeping. If the mutex or rw-lock becomes available dur- ing this polling period, the thread can continue immediately, in the same time slice. However, too-frequent polling by multiple threads of a shared object can cause “cache ping-pong”, the shipping of cache lines between processors. InnoDB minimizes this is- sue by waiting a random time between subsequent polls. The delay is implemented as a busy loop. Starting with InnoDB 1.0.4, you can control the maximum delay between testing a mutex or rw-lock using the parameter in- nodb_spin_wait_delay. In the 100 MHz Pentium era, the unit of delay was one microsecond. The duration of the delay loop depends on the C compiler and the target processor. On a system where all processor cores share a fast cache memory, you might reduce the maximum delay or disable the busy loop altogether by setting innodb_spin_wait_delay=0. On a system that consists of multiple processor chips, the shipping of cache lines can be slower and you might increase the maximum delay. 30
- 39. Performance and Scalability Enhancements The default value of innodb_spin_wait_delay is 6. The spin wait delay is a dynamic, global parameter that can be specified in the MySQL option file (my.cnf or my.ini) or changed at runtime with the command SET GLOBAL in- nodb_spin_wait_delay=delay, where delay is the desired maximum delay. Changing the setting requires the SUPER privilege. For performance considerations for InnoDB locking operations, see Optimizing Locking Operations. 7.15. Making Buffer Cache Scan Resistant Rather than using a strictly LRU algorithm, InnoDB uses a technique to minimize the amount of data that is brought into the buffer cache and never accessed again. The goal is to make sure that frequently accessed (“hot”) pages remain in the buffer cache, even as read-ahead and full table scans bring new blocks in that might or might not be accessed afterward. Newly read blocks are inserted into the middle of the list representing the buffer cache. of the LRU list. All newly read pages are inserted at a location that by default is 3/8 from the tail of the LRU list. The pages are moved to the front of the list (the most- recently used end) when they are accessed in the buffer cache for the first time. Thus pages that are never accessed never make it to the front portion of the LRU list, and “age out” sooner than with a strict LRU approach. This arrangement divides the LRU list into two segments, where the pages downstream of the insertion point are considered “old” and are desirable victims for LRU eviction. For an explanation of the inner workings of the InnoDB buffer pool and the specifics of its LRU replacement algorithm, see The InnoDB Buffer Pool. Starting with InnoDB 1.0.5, you can control the insertion point in the LRU list, and choose whether InnoDB applies the same op- timization to blocks brought into the buffer pool by table or index scans. The configuration parameter innodb_old_blocks_pct controls the percentage of “old” blocks in the LRU list. The default value of innodb_old_blocks_pct is 37, corresponding to the original fixed ratio of 3/8. The value range is 5 (new pages in the buffer pool age out very quickly) to 95 (only 5% of the buffer pool reserved for hot pages, making the algorithm close to the fa- miliar LRU strategy). The optimization that keeps the buffer cache from being churned by read-ahead can avoid similar problems due to table or index scans. In these scans, a data page is typically accessed a few times in quick succession and is never touched again. The configura- tion parameter innodb_old_blocks_time specifies the time window (in milliseconds) after the first access to a page during which it can be accessed without being moved to the front (most-recently used end) of the LRU list. The default value of in- nodb_old_blocks_time is 0, corresponding to the original behavior of moving a page to the most-recently used end of the buffer pool list when it is first accessed in the buffer pool. Increasing this value makes more and more blocks likely to age out faster from the buffer pool. Both the new parameters innodb_old_blocks_pct and innodb_old_blocks_time are dynamic, global and can be spe- cified in the MySQL option file (my.cnf or my.ini) or changed at runtime with the SET GLOBAL command. Changing the set- ting requires the SUPER privilege. To help you gauge the effect of setting these parameters, some additional statistics are reported by SHOW ENGINE INNODB STATUS command. The BUFFER POOL AND MEMORY section now looks like: Total memory allocated 1107296256; in additional pool allocated 0 Dictionary memory allocated 80360 Buffer pool size 65535 Free buffers 0 Database pages 63920 Old database pages 23600 Modified db pages 34969 Pending reads 32 Pending writes: LRU 0, flush list 0, single page 0 Pages made young 414946, not young 2930673 1274.75 youngs/s, 16521.90 non-youngs/s Pages read 486005, created 3178, written 160585 2132.37 reads/s, 3.40 creates/s, 323.74 writes/s Buffer pool hit rate 950 / 1000, young-making rate 30 / 1000 not 392 / 1000 Pages read ahead 1510.10/s, evicted without access 0.00/s LRU len: 63920, unzip_LRU len: 0 I/O sum[43690]:cur[221], unzip sum[0]:cur[0] • Old database pages is the number of pages in the “old” segment of the LRU list. • Pages made young and not young is the total number of “old” pages that have been made young or not respectively. • youngs/s and non-young/s is the rate at which page accesses to the “old” pages have resulted in making such pages young or otherwise respectively since the last invocation of the command. • young-making rate and not provides the same rate but in terms of overall buffer cache accesses instead of accesses just to the “old” pages. 31
- 40. Performance and Scalability Enhancements Because the effects of these parameters can vary widely based on your hardware configuration, your data, and the details of your workload, always benchmark to verify the effectiveness before changing these settings in any performance-critical or production environment. In mixed workloads where most of the activity is OLTP type with periodic batch reporting queries which result in large scans, set- ting the value of innodb_old_blocks_time during the batch runs can help keep the working set of the normal workload in the buffer cache. When scanning large tables that cannot fit entirely in the buffer pool, setting innodb_old_blocks_pct to a small value keeps the data that is only read once from consuming a significant portion of the buffer pool. For example, setting in- nodb_old_blocks_pct=5 restricts this data that is only read once to 5% of the buffer pool. When scanning small tables that do fit into memory, there is less overhead for moving pages around within the buffer pool, so you can leave innodb_old_blocks_pct at its default value, or even higher, such as innodb_old_blocks_pct=50. The effect of the innodb_old_blocks_time parameter is harder to predict than the innodb_old_blocks_pct paramet- er, is relatively small, and varies more with the workload. To arrive at an optimal value, conduct your own benchmarks if the per- formance improvement from adjusting innodb_old_blocks_pct is not sufficient. For more information about the InnoDB buffer pool, see The InnoDB Buffer Pool. 7.16. Improvements to Crash Recovery Performance Starting with InnoDB 1.0.7, a number of optimizations speed up certain steps of the recovery that happens on the next startup after a crash. In particular, scanning the redo log and applying the redo log are faster, due to improved algorithms for memory manage- ment. You do not need to take any actions to take advantage of this performance enhancement. If you kept the size of your logfiles artificially low because recovery took a long time, you can consider increasing the logfile size. For more information about InnoDB recovery, see The InnoDB Recovery Process. 7.17. Integration with MySQL PERFORMANCE_SCHEMA Starting with InnoDB 1.1 with MySQL 5.5, you can profile certain internal InnoDB operations using the MySQL Performance Schema feature. This type of tuning is primarily for expert users, those who push the limits of MySQL performance, read the MySQL source code, and evaluate optimization strategies to overcome performance bottlenecks. DBAs can also use this feature for capacity planning, to see whether their typical workload encounters any performance bottlenecks with a particular combination of CPU, RAM, and disk storage; and if so, to judge whether performance can be improved by increasing the capacity of some part of the system. To use this feature to examine InnoDB performance: • You must be running MySQL 5.5 or higher. You must build the database server from source, enabling the Performance Schema feature by building with the --with-perfschema option. Since the Performance Schema feature introduces some perform- ance overhead, you should use it on a test or development system rather than on a production system. • You must be running InnoDB 1.1 or higher. • You must be generally familiar with how to use the Performance Schema feature, for example to query tables in the per- formance_schema database. • Examine the following kinds of InnoDB objects by querying the appropriate performance_schema tables. The items asso- ciated with InnoDB all contain the substring innodb in the NAME column. For the definitions of the *_INSTANCES tables, see Performance Schema Instance Tables. For the definitions of the *_SUMMARY_* tables, see Performance Schema Summary Tables. For the definition of the PROCESSLIST table, see Per- formance Schema Miscellaneous Tables. For the definition of the *_CURRENT_* tables, see Performance Schema Events (Current) Table. For the definition of the *_HISTORY_* tables, see Performance Schema History Tables. • Mutexes in the mutex_instances table. (Mutexes and RW-locks related to the InnoDB buffer pool are not included in this coverage; the same applies to the output of the SHOW ENGINE INNODB MUTEX command.) • RW-locks in the rwlock_instances table. • RW-locks in the rwlock_instances table. • File I/O operations in the file_instances, file_summary_by_event_name, and file_summary_by_instance tables. 32
- 41. Performance and Scalability Enhancements • Threads in the PROCESSLIST table. • During performance testing, examine the performance data in the events_waits_current and events_waits_history_long tables. If you are interested especially in InnoDB-related objects, use the clause where name like "%innodb%" to see just those entries; otherwise, examine the performance statistics for the overall MySQL server. • You must be running MySQL 5.5, with the Performance Schema enabled by building with the --with-perfschema build option. For more information about the MySQL Performance Schema, see MySQL Performance Schema. 7.18. Improvements to Performance from Multiple Buffer Pools This performance enhancement is primarily useful for people with a large buffer pool size, typically in the multi-gigabyte range. To take advantage of this speedup, you must set the new innodb_buffer_pool_instances configuration option, and you might also adjust the innodb_buffer_pool_size value. When the InnoDB buffer pool is large, many data requests can be satisfied by retrieving from memory. You might encounter bot- tlenecks from multiple threads trying to access the buffer pool at once. Starting in InnoDB 1.1 and MySQL 5.5, you can enable multiple buffer pools to minimize this contention. Each page that is stored in or read from the buffer pool is assigned to one of the buffer pools randomly, using a hashing function. Each buffer pool manages its own free lists, flush lists, LRUs, and all other data structures connected to a buffer pool, and is protected by its own buffer pool mutex. To enable this feature, set the innodb_buffer_pool_instances configuration option to a value from 1 (the default) to 64 (the maximum). This option only takes effect when you set the innodb_buffer_pool_size to a size of 1 gigabyte or more. The total size you specify is divided up among all the buffer pools. We recommend specifying a combination of in- nodb_buffer_pool_instances and innodb_buffer_pool_size so that each buffer pool instance is at least 1 giga- byte. For more information about the InnoDB buffer pool, see The InnoDB Buffer Pool. 7.19. Better Scalability with Multiple Rollback Segments The InnoDB rollback segment could be a bottleneck on high-capacity systems, because it can handle a maximum of 1023 concur- rent transactions that change any data. (Read-only transactions do not count against that maximum.) Starting in InnoDB 1.1 with MySQL 5.5, the limit on concurrent transactions is greatly expanded. The single rollback segment is divided into 128 segments, each of which can support up to 1023 transactions that perform writes, for a total of approximately 128K concurrent transactions. Each transaction is assigned to one of the rollback segments, and remains tied to that rollback segment for the duration. This en- hancement improves both scalability (higher number of concurrent transactions) and performance (less contention when different transactions access the rollback segments). To take advantage of this feature, you do not need to create any new database or tables, or reconfigure anything. You must do a slow shutdown before upgrading to InnoDB 1.1, or some time afterward. InnoDB makes the required changes inside the system ta- blespace automatically, the first time you restart after performing a slow shutdown. For more information about performance of InnoDB under high transactional load, see Optimizing InnoDB Transaction Manage- ment. 7.20. Better Scalability with Improved Purge Scheduling Starting in InnoDB 1.1 with MySQL 5.5, the purge operations (a type of garbage collection) that InnoDB performs automatically can be done in a separate thread, rather than as part of the master thread. This change improves scalability, because the main data- base operations run independently from maintenance work happening in the background. To enable this feature, set the configuration option innodb_purge_threads=1, as opposed to the default of 0, which com- bines the purge operation into the master thread. You might not notice a significant speedup, because the purge thread might encounter new types of contention; the single purge thread really lays the groundwork for further tuning and possibly multiple purge threads in the future. There is another new config- uration option, innodb_purge_batch_size with a default of 20 and maximum of 5000. This option is mainly intended for experimentation and tuning of purge operations, and should not be interesting to typical users. For more information about InnoDB I/O performance, see Optimizing InnoDB Disk I/O. 7.21. Improved Log Sys Mutex 33
- 42. Performance and Scalability Enhancements This is another performance improvement that comes for free, with no user action or configuration needed. The details here are in- tended for performance experts who delve into the InnoDB source code, or interpret reports with keywords such as “mutex” and “log_sys”. The mutex known as the log sys mutex has historically done double duty, controlling access to internal data structures related to log records and the LSN, as well as pages in the buffer pool that are changed when a mini-transaction is committed. Starting in InnoDB 1.1 with MySQL 5.5, these two kinds of operations are protected by separate mutexes, with a new log_buf mutex controlling writes to buffer pool pages due to mini-transactions. For performance considerations for InnoDB locking operations, see Optimizing Locking Operations. 7.22. Separate Flush List Mutex Starting with InnoDB 1.1 with MySQL 5.5, concurrent access to the buffer pool is faster. Operations involving the flush list, a data structure related to the buffer pool, are now controlled by a separate mutex and do not block access to the buffer pool. You do not need to configure anything to take advantage of this speedup; it is fully automatic. For more information about the InnoDB buffer pool, see The InnoDB Buffer Pool. 34
- 43. Chapter 8. Changes for Flexibility, Ease of Use and Reliability This chapter describes several features in InnoDB 1.1 that offer new flexibility and improve ease of use, reliability and perform- ance. The “Barracuda” file format improves efficiency for storing large variable-length columns, and enables table compression. Configuration options that once were unchangeable after startup, are now flexible and can be changed dynamically. Some improve- ments are automatic, such as faster and more efficient TRUNCATE TABLE. Others allow you the flexibility to control InnoDB be- havior; for example, you can control whether certain problems cause errors or just warnings. And informational messages and error reporting continue to be made more user-friendly. 8.1. Enabling New File Formats InnoDB supports named file formats to improve compatibility between database file formats and various InnoDB versions. To create new tables that require a new file format, you must enable the new “Barracuda” file format, using the configuration para- meter innodb_file_format. The value of this parameter determines whether a newly created table or index can use compres- sion or the new DYNAMIC row format. If you omit innodb_file_format or set it to “Antelope”, you preclude the use of new features that would make your database inaccessible to the built-in InnoDB in MySQL 5.1 and prior releases. You can set the value of innodb_file_format on the command line when you start mysqld, or in the option file my.cnf (Unix operating systems) or my.ini (Windows). You can also change it dynamically with the SET GLOBAL statement. Further information about managing file formats is presented in Chapter 4, InnoDB File Format Management. 8.2. Dynamic Control of System Configuration Parameters In InnoDB 1.1, it is possible to chance certain system configuration parameters dynamically, without shutting down and restarting the server as was previously necessary. This increases up-time and facilitates testing of various options. You can now set these parameters dynamically: • innodb_file_per_table • innodb_stats_on_metadata • innodb_lock_wait_timeout • innodb_adaptive_hash_index 8.2.1. Dynamically Changing innodb_file_per_table Since MySQL version 4.1, InnoDB has provided two options for how tables are stored on disk. You can choose to create a new ta- ble and its indexes in the shared system tablespace (corresponding to the set of files named ibdata files), along with other intern- al InnoDB system information. Or, you can chose to use a separate file (an .ibd file) to store a new table and its indexes. The tablespace style used for new tables is determined by the setting of the configuration parameter innodb_file_per_table at the time a table is created. Previously, the only way to set this parameter was in the MySQL option file (my.cnf or my.ini), and changing it required shutting down and restarting the server. In InnoDB 1.1, the configuration parameter in- nodb_file_per_table is dynamic, and can be set ON or OFF using the SET GLOBAL statement. The default setting is OFF, so new tables and indexes are created in the system tablespace. Dynamically changing the value of this parameter requires the SU- PER privilege and immediately affects the operation of all connections. Tables created when innodb_file_per_table is disabled cannot use the new compression capability, or use the new row format DYNAMIC. Tables created when innodb_file_per_table is enabled can use those new features, and each table and its indexes are stored in a new .ibd file. The ability to change the setting of innodb_file_per_table dynamically is useful for testing. As noted above, the parameter innodb_file_format is also dynamic, and must be set to “Barracuda” to create new compressed tables, or tables that use the new row format DYNAMIC. Since both parameters are dynamic, it is easy to experiment with these table formats without a system shutdown and restart. Note that InnoDB can add and drop a table's secondary indexes without re-creating the table, but must recreate the table when you change the clustered (primary key) index (see Chapter 2, Fast Index Creation in the InnoDB Storage Engine). When a table is re- created as a result of creating or dropping an index, the table and its indexes are stored in the shared system tablespace or in a sep- arate .ibd file just as if it were created using a CREATE TABLE statement (and depending on the setting of in- nodb_file_per_table). When an index is created without rebuilding the table, the index is stored in the same file as the clustered index, regardless of the setting of innodb_file_per_table. 35
- 44. Changes for Flexibility, Ease of Use and Reliability 8.2.2. Dynamically Changing innodb_stats_on_metadata As noted in Section 8.5, “Controlling Optimizer Statistics Estimation”, you can control how InnoDB gathers information about the number of distinct values in an index key. A related parameter, innodb_stats_on_metadata, has existed since MySQL re- lease 5.1.17 to control whether or not InnoDB performs statistics gathering when metadata statements are executed. See InnoDB Startup Options and System Variables for details. Beginning with release InnoDB 1.0.2, you can change the setting of innodb_stats_on_metadata dynamically at runtime with the statement SET GLOBAL innodb_stats_on_metadata=mode, where mode is either ON or OFF (or 1 or 0). Chan- ging this setting requires the SUPER privilege and immediately affects the operation of all connections. 8.2.3. Dynamically Changing innodb_lock_wait_timeout When a transaction is waiting for a resource, it will wait for the resource to become free, or stop waiting and return with the error ERROR HY000: Lock wait timeout exceeded; try restarting transaction The length of time a transaction will wait for a resource before “giving up” is determined by the value of the configuration para- meter innodb_lock_wait_timeout. The default setting for this parameter is 50 seconds. The minimum setting is 1 second, and values above 100,000,000 disable the timeout, so a transaction will wait “forever”. Following a timeout, the SQL statement that was executing will be rolled back. (In MySQL 5.0.12 and earlier, the transaction rolled back.) The user application may try the statement again (usually after waiting for a while), or rollback the entire transaction and restart. Before InnoDB 1.0.2, the only way to set this parameter was in the MySQL option file (my.cnf or my.ini), and changing it re- quired shutting down and restarting the server. Beginning with InnoDB 1.0.2, the configuration parameter in- nodb_lock_wait_timeout can be set at runtime with the SET GLOBAL or SET SESSION statement. Changing the GLOB- AL setting requires the SUPER privilege and affects the operation of all clients that subsequently connect. Any client can change the SESSION setting for innodb_lock_wait_timeout, which affects only that client. 8.2.4. Dynamically Changing innodb_adaptive_hash_index As described in Section 7.5, “Controlling Adaptive Hash Indexing”, it may be desirable, depending on your workload, to dynamic- ally enable or disable the adaptive hash indexing scheme InnoDB uses to improve query performance. Version 5.1.24 of MySQL introduced the start-up option innodb_adaptive_hash_index that allows the adaptive hash index to be disabled. It is enabled by default. Starting with InnoDB 1.0.3, the parameter can be modified by the SET GLOBAL statement, without restarting the server. Changing the setting requires the SUPER privilege. Disabling the adaptive hash index will empty the hash table immediately. Normal operations can continue while the hash table is emptied, and executing queries that have been using the hash table will access the index B-trees directly instead of attempting to utilize the hash index. When the adaptive hash index is enabled, the hash table will be populated during normal operation. 8.3. TRUNCATE TABLE Reclaims Space When you TRUNCATE a table that is stored in an .ibd file of its own (because innodb_file_per_table was enabled when the table was created), and if the table is not referenced in a FOREIGN KEY constraint, the table is dropped and re-created in a new .idb file. This operation is much faster than deleting the rows one by one. The operating system can reuse the disk space, in con- trast to tables within the InnoDB system tablespace, where only InnoDB can use the space after they are truncated. Page-level backups can also be smaller, without big blocks of unused space in the middle of the system tablespace. Previous versions of InnoDB would re-use the existing .idb file, thus releasing the space only to InnoDB for storage manage- ment, but not to the operating system. Note that when the table is truncated, the count of rows affected by the TRUNCATE state- ment is an arbitrary number. Note If there is a referential constraint between columns of the table being truncated, it still is truncated using this fast tech- nique. If there are referential constraints between the table being truncated and other tables, the truncate operation fails. This is a change to the previous behavior, which would transform the TRUNCATE operation to a DELETE operation that would remove all the rows and trigger ON DELETE operations on child tables. 8.4. InnoDB Strict Mode To guard against ignored typos and syntax errors in SQL, or other unintended consequences of various combinations of operational modes and SQL statements, InnoDB provides a “strict mode” of operations. In this mode, InnoDB raises error conditions in certain cases, rather than issuing a warning and processing the specified statement (perhaps with unintended defaults). This is analogous to sql_mode in MySQL, which controls what SQL syntax MySQL accepts, and determines whether it silently ignores errors, or val- 36
- 45. Changes for Flexibility, Ease of Use and Reliability idates input syntax and data values. Since strict mode is relatively new, some statements that execute without errors with earlier versions of InnoDB might generate errors unless you disable strict mode. The setting of InnoDB strict mode affects the handling of syntax errors on the CREATE TABLE, ALTER TABLE and CREATE INDEX statements. Starting with InnoDB 1.0.2, the strict mode also enables a record size check, so that an INSERT or UPDATE never fails due to the record being too large for the selected page size. Using the clauses and settings for ROW_FORMAT and KEY_BLOCK_SIZE on CREATE TABLE, ALTER TABLE, and CREATE INDEX statements can be confusing when not running in strict mode. Unless you run in strict mode, InnoDB ignores certain syntax errors and creates the table or index, with only a warning in the message log. However, if InnoDB strict mode is on, such problems generate an immediate error and the table or index is not created, thus saving time by catching the error at the time the statement is issued. InnoDB strict mode is set with the configuration parameter innodb_strict_mode, which can be specified as on or off. You can set the value on the command line when you start mysqld, or in the configuration file my.cnf (Unix operating systems) or my.ini (Windows). You can also enable or disable InnoDB strict mode at runtime with the statement SET [GLOBAL|SESSION] innodb_strict_mode=mode, where mode is either ON or OFF. Changing the GLOBAL setting re- quires the SUPER privilege and affects the operation of all clients that subsequently connect. Any client can change the SESSION setting for innodb_strict_mode, which affects only that client. 8.5. Controlling Optimizer Statistics Estimation The MySQL query optimizer uses estimated statistics about key distributions to select or avoid using an index in an execution plan, based on the relative selectivity of the index. Previously, InnoDB sampled 8 random pages from an index to get an estimate of the cardinality of (i.e., the number of distinct values in) the index. (This page sampling technique is frequently described as “index dives”.) This small number of page samples frequently was insufficient, and could give inaccurate estimates of an index's selectiv- ity and thus lead to poor choices by the query optimizer. To give users control over the quality of the statistics estimate (and thus better information for the query optimizer), the number of sampled pages now can be changed using the parameter innodb_stats_sample_pages. This feature addresses user requests such as that as expressed in MySQL Bug#25640: InnoDB Analyze Table Should Allow User Selection of Index Dives. You can change the number of sampled pages using the global parameter innodb_stats_sample_pages, which can be set at runtime. The default value for this parameter is 8, preserving the same behavior as in past releases. Note that the value of innodb_stats_sample_pages affects the index sampling for all tables and indexes. Note that there are the following potentially significant impacts when you change the index sample size: • Small values like 1 or 2 can result in very inaccurate estimates of cardinality. • Values much larger than 8 (say, 100), can cause a big slowdown in the time it takes to open a table or execute SHOW TABLE STATUS. • The optimizer might choose very different query plans based on different estimates of index selectivity. The cardinality estimation can be disabled for metadata statements such as SHOW TABLE STATUS by executing the statement SET GLOBAL innodb_stats_on_metadata=OFF (or 0). Before InnoDB 1.0.2, this variable could only be set in the MySQL option file (my.cnf or my.ini), and changing it required shutting down and restarting the server. The cardinality (the number of different key values) in every index of a table is calculated when a table is opened, at SHOW TABLE STATUS and ANALYZE TABLE and in other circumstances such as when the table has changed significantly. Note that all tables are opened, and the statistics are re-estimated, when the mysql client starts if the auto-rehash setting is set on (the default). The auto-rehash feature enables automatic name completion of database, table, and column names for interactive users. You can turn auto-rehash off to improve the start up time of the mysql client. Because the statistics are automatically recalculated at various times other than on execution of ANALYZE TABLE, it does not make sense to increase the index sample size, then run ANALYZE TABLE and decrease sample size to attempt to obtain better stat- istics. The “better” statistics calculated by ANALYZE running with a high value of innodb_stats_sample_pages will be wiped away. The estimated cardinality for an index is more accurate with a larger number of samples, but each sample might require a disk read, so you do not want to make the sample size too large. Choose a value for innodb_stats_sample_pages that results in reas- onably accurate estimates for all tables in your database without requiring excessive I/O. Although it is not possible to specify the sample size on a per-table basis, smaller tables generally require fewer index samples than larger tables do. If your database has many large tables, consider using a higher value for innodb_stats_sample_pages than if you have mostly smaller tables. 37
- 46. Changes for Flexibility, Ease of Use and Reliability 8.6. Better Error Handling when Dropping Indexes For efficiency, InnoDB requires an index to exist on foreign key columns so that UPDATE and DELETE operations on a “parent” table can easily check for the existence or non-existence of corresponding rows in the “child” table. To ensure that there is an ap- propriate index for such checks, MySQL sometimes implicitly creates or drops such indexes as a side-effect of CREATE TABLE, CREATE INDEX, and ALTER TABLE statements. When you DROP an index, InnoDB checks that this operation does not compromise referential integrity checking. That is, it is OK to drop the index if it is not used for checking foreign key constraints, or if there is another index that can also be used for that pur- pose. InnoDB prevents you from dropping the last usable index for enforcing any given referential constraint. Users have been con- fused by this behavior, as reported in MySQL Bug#21395. In releases prior to InnoDB 1.0.2, attempts to drop the only usable index would result in an error message such as: ERROR 1025 (HY000): Error on rename of './db2/#sql-18eb_3' to './db2/foo'(errno: 150) Beginning with InnoDB 1.0.2, this error condition is reported with a more friendly message: ERROR 1553 (HY000): Cannot drop index 'fooIdx': needed in a foreign key constraint A similar change in error reporting applies to an attempt to drop the primary key index. InnoDB ensures that there is a primary key index for every table, even if you do not declare an explicit PRIMARY KEY. In such cases, InnoDB creates an implicit clustered index using the first columns of the table that are declared UNIQUE and NOT NULL. When the InnoDB storage engine is used with a MySQL version earlier than 5.1.29, an attempt to drop an implicit clustered index (the first UNIQUE NOT NULL index) fails if the table does not contain a PRIMARY KEY. This has been reported as MySQL Bug#31233. Attempts to use the DROP INDEX or ALTER TABLE statement to drop such an index generate this error: ERROR 42000: This table type requires a primary key Beginning with MySQL 5.1.29 when using the InnoDB Plugin, attempts to drop such an index succeed. Behind the scenes, InnoDB copies the table, rebuilding the index using a different UNIQUE NOT NULL group of columns or a system-generated key. Since this operation changes the primary key, it uses the slow method of copying the table and re-creating the index, rather than the Fast Index Creation technique from Section 2.3, “Implementation Details of Fast Index Creation”. In those versions of MySQL that are affected by this bug, one way to change an index of this type is to create a new table and copy the data into it using INSERT INTO newtable SELECT * FROM oldtable, and then DROP the old table and rename the new table. However, if there are existing tables with references to the table whose index you are dropping, you must first use the ALTER TA- BLE statement to remove foreign key references from or to other tables. Because MySQL does not support dropping or creating FOREIGN KEY constraints, if you use ALTER TABLE to add or remove a REFERENCES constraint, the child table is copied, rather than using “Fast Index Creation”. 8.7. More Compact Output of SHOW ENGINE INNODB MUTEX The statement SHOW ENGINE INNODB MUTEX displays information about InnoDB mutexes and rw-locks. It can be a useful tuning aid on multi-core systems. However, with a big buffer pool, the size of the output may be overwhelming. There is a mutex and rw-lock in each 16K buffer pool block. It is highly improbable that an individual block mutex or rw-lock could become a per- formance bottleneck, and there are 65,536 blocks per gigabyte. Starting with InnoDB 1.0.4, SHOW ENGINE INNODB MUTEX skips the mutexes and rw-locks of buffer pool blocks. It also does not list any mutexes or rw-locks that have never been waited on (os_waits=0). Thus, SHOW ENGINE INNODB MUTEX only displays information about mutexes and rw-locks that do not belong to the buffer pool blocks, and that have caused at least one OS level wait. 8.8. More Read-Ahead Statistics As described in Section 7.7, “Changes in the Read-Ahead Algorithm” a read-ahead request is an asynchronous I/O request issued in anticipation that a page will be used in the near future. Knowing how many pages are read through this read-ahead mechanism, and how many of them are evicted from the buffer pool without ever being accessed, can be useful to help fine-tune the parameter in- nodb_read_ahead_threshold. Starting from InnoDB 1.0.5, SHOW STATUS output displays the global status variables Innodb_buffer_pool_read_ahead and Innodb_buffer_pool_read_ahead_evicted. These variables indicate the number of pages read in as part of read ahead, and the number of such pages evicted without ever being accessed respectively. These counters provide global values since 38
- 47. Changes for Flexibility, Ease of Use and Reliability the start of the server. Note that the status variables Innodb_buffer_pool_read_ahead_rnd and In- nodb_buffer_pool_read_ahead_seq have been removed from the SHOW STATUS output. In addition to the two counters mentioned above, SHOW INNODB STATUS also shows the rate at which the read-ahead pages are read in and the rate at which such pages are evicted without being accessed. The per-second averages are based on the statistics col- lected since the last invocation of SHOW INNODB STATUS and are displayed in the BUFFER POOL AND MEMORY section of the output. 39
- 48. Chapter 9. Installing the InnoDB Storage Engine When you use the InnoDB storage engine 1.1 and above, with MySQL 5.5 and above, you do not need to do anything special to in- stall: everything comes configured as part of the MySQL source and binary distributions. This is a change from earlier releases of the InnoDB Plugin, where you were required to match up MySQL and InnoDB version numbers and update your build and config- uration processes. The InnoDB storage engine is included in the MySQL distribution, starting from MySQL 5.1.38. From MySQL 5.1.46 and up, this is the only download location for the InnoDB storage engine; it is not available from the InnoDB web site. If you used any scripts or configuration files with the earlier InnoDB storage engine from the InnoDB web site, be aware that the fi- lename of the shared library as supplied by MySQL is ha_innodb_plugin.so or ha_innodb_plugin.dll, as opposed to ha_innodb.so or ha_innodb.dll in the older Plugin downloaded from the InnoDB web site. You might need to change the applicable file names in your startup or configuration scripts. Because the InnoDB storage engine has now replaced the built-in InnoDB, you no longer need to specify options like - -ignore-builtin-innodb and --plugin-load during startup. Prior to MySQL 5.5.5, we recommended specifying the following options in your configuration file. In MySQL 5.5.5 and higher, these settings are the default. default-storage-engine=InnoDB innodb_file_per_table=1 innodb_file_format=barracuda innodb_strict_mode=1 Tables created with the Barracuda file format are incompatible with InnoDB versions prior to the InnoDB Plugin. Tables in the Barracuda format can be compressed, and can store portions of long columns off-page, outside the B-tree nodes. The new InnoDB strict mode guards SQL or certain operational errors that otherwise generate warnings and possible unintended consequences of ignored or incorrect SQL commands or parameters. As described in Section 8.4, “InnoDB Strict Mode”, the GLOBAL parameter innodb_strict_mode can be set ON or OFF. You can also use the command SET SESSION in- nodb_strict_mode=mode (where mode is ON or OFF) to enable or disable InnoDB strict mode on a per-session basis. 40
- 49. Chapter 10. Upgrading the InnoDB Storage Engine Thanks to the pluggable storage engine architecture of MySQL, upgrading the InnoDB storage engine is usually a simple matter of shutting down MySQL, replacing a platform-specific executable file, and restarting the server. If you wish to upgrade and use your existing database, it is essential to perform a “slow” shutdown, or the new plugin may fail when merging buffered inserts or pur- ging deleted records. The slow shutdown also allows InnoDB to carry out the changes to the system tablespace needed for multiple rollback segments on the next startup. If your database does not contain any compressed tables, you should be able to use your database with the newest InnoDB storage engine without problems after a slow shutdown. However, if your database contains compressed tables, it may not be compatible with InnoDB storage engine 1.1. Because of an in- compatible change introduced in InnoDB storage engine version 1.0.2, some compressed tables may need to be rebuilt, as noted in Section 10.3, “Converting Compressed Tables Created Before Version 1.0.2”. Please follow these steps carefully. You may, of course, rebuild your database using mysqldump or other methods. This may be a preferable approach if your data- base is small or there are many referential constrains among tables. Note that once you have accessed your database with InnoDB storage engine 1.1, you should not try to use it with the Plugin prior to 1.0.2. 10.1. Upgrading the Dynamic InnoDB Storage Engine Before shutting down the MySQL server containing the InnoDB storage engine, you must enable “slow” shutdown: SET GLOBAL innodb_fast_shutdown=0; For the details of the shutdown procedure, see The Shutdown Process. In the directory where the MySQL server looks for plugins, rename the executable file of the old InnoDB storage engine (ha_innodb_plugin.so or ha_innodb_plugin.dll), so that you can restore it later if needed. You may remove the file later. The plugin directory is specified by the system variable plugin_dir. The default location is usually the lib/plugin subdirectory of the directory specified by basedir. Download a suitable package for your server platform, operating system and MySQL version. Extract the contents of the archive using tar or a similar tool for Linux and Unix, or Windows Explorer or WinZip or similar utility for Windows. Copy the file ha_innodb_plugin.so or ha_innodb_plugin.dll to the directory where the MySQL server looks for plugins. Start the MySQL server. Follow the procedure in Section 10.3, “Converting Compressed Tables Created Before Version 1.0.2” to convert any compressed tables if needed. 10.2. Upgrading a Statically Built InnoDB Storage Engine As with a dynamically installed InnoDB storage engine, you must perform a “slow” shutdown of the MySQL server. If you have built MySQL from source code and replaced the built-in InnoDB in MySQL with the InnoDB storage engine in the source tree, you will have a special version of the mysqld executable that contains the InnoDB storage engine. If you intend to upgrade a statically built InnoDB storage engine to another statically built plugin, rebuild the mysqld executable, shut down the server, and replace the mysqld executable before starting the server. Either way, please be sure to follow the instructions of Section 10.3, “Converting Compressed Tables Created Before Version 1.0.2” if any compressed tables were created. 10.3. Converting Compressed Tables Created Before Version 1.0.2 The InnoDB storage engine version 1.0.2 introduced an incompatible change to the format of compressed tables. Some com- pressed tables that were created with an earlier version of the InnoDB storage engine may need to be rebuilt with a bigger KEY_BLOCK_SIZE before they can be used. If you must keep your existing database when you upgrade to InnoDB storage engine 1.0.2 or newer, you must perform a “slow” shutdown of MySQL running the previous version of the InnoDB storage engine. Following such a shutdown, and using the newer release of the InnoDB storage engine, you must determine which compressed tables need conversion and then follow a procedure to upgrade these tables. Because most users do not have tables where this process is required, this manual does not detail the pro- cedures required. If you have created compressed tables with the InnoDB storage engine prior to release 1.0.2, review the relevant information on the InnoDB website. 41
- 50. Chapter 11. Downgrading the InnoDB Storage Engine 11.1. Overview There are times when you might want to use the InnoDB Plugin with a given database, and then downgrade to the built-in InnoDB in MySQL. One reason to do this is because you want to take advantage of a new InnoDB storage engine feature (such as “Fast In- dex Creation”), but revert to the standard built-in InnoDB in MySQL for production operation. If you have created new tables using the InnoDB storage engine, you might need to convert them to a format that the built-in In- noDB in MySQL can read. Specifically, if you have created tables that use ROW_FORMAT=COMPRESSED or ROW_FORMAT=DYNAMIC you must convert them to a different format, if you plan to access these tables with the built-in InnoDB in MySQL. If you do not do so, anomalous results may occur. Although InnoDB checks the format of tables and database files (specifically *.ibd files) for compatibility, it is unable to start if there are buffered changes for “too new format” tables in the redo log or in the system tablespace. Thus it is important to carefully follow these procedures when downgrading from the InnoDB storage engine to the built-in InnoDB in MySQL, version 5.1. This chapter describes the downgrade scenario, and the steps you should follow to ensure correct processing of your database. 11.2. The Built-in InnoDB, the Plugin and File Formats Starting with version 5.0.21, the built-in InnoDB in MySQL checks the table type before opening a table. Until now, all InnoDB tables have been tagged with the same type, although some changes to the format have been introduced in MySQL versions 4.0, 4.1, and 5.0. One of the important new features introduced with the InnoDB storage engine is support for identified file formats. This allows the InnoDB storage engine and versions of InnoDB since 5.0.21 to check for file compatibility. It also allows the user to preclude the use of features that would generate downward incompatibilities. By paying attention to the file format used, you can protect your database from corruptions, and ensure a smooth downgrade process. In general, before using a database file created with the InnoDB storage engine with the built-in InnoDB in MySQL you should verify that the tablespace files (the *.ibd files) are compatible with the built-in InnoDB in MySQL. The InnoDB storage engine can read and write tablespaces in both the formats “Antelope” and “Barracuda”. The built-in InnoDB can only read and write ta- blespaces in “Antelope” format. To make all tablespaces “legible” to the built-in InnoDB in MySQL, you should follow the instruc- tions in Section 11.3, “How to Downgrade” to reformat all tablespaces to be in the “Antelope” format. Generally, after a “slow” shutdown of the InnoDB storage engine (innodb_fast_shutdown=0), it should be safe to open the data files with the built-in InnoDB in MySQL. See Section 11.4, “Possible Problems” for a discussion of possible problems that can arise in this scenario and workarounds for them. 11.3. How to Downgrade 11.3.1. Converting Tables The built-in InnoDB in MySQL can access only tables in the “Antelope” file format, that is, in the REDUNDANT or COMPACT row format. If you have created tables in COMPRESSED or DYNAMIC format, the corresponding tablespaces in the new “Barracuda” file format, and it is necessary to downgrade these tables. First, identify the tables that require conversion, by executing this command: SELECT table_schema, table_name, row_format FROM information_schema.tables WHERE engine='innodb' AND row_format NOT IN ('Redundant', 'Compact'); Next, for each table that requires conversion, run the following command: ALTER TABLE table_name ROW_FORMAT=COMPACT; This command copies the table and its indexes to a new tablespace in the “Antelope” format. See Chapter 2, Fast Index Creation in the InnoDB Storage Engine for a discussion of exactly how such index creation operations are performed. 11.3.2. Adjusting the Configuration Before you shut down the InnoDB storage engine and start the basic built-in InnoDB in MySQL, review the configuration files. Changes to the startup options do not take effect until the server is restarted, or the InnoDB storage engine is uninstalled and rein- stalled. 42
- 51. Downgrading the InnoDB Storage Engine InnoDB 1.1 includes several configuration parameters that are not recognized by the built-in InnoDB prior to MySQL 5.5. For ex- ample, MySQL 5.1 and earlier (without the InnoDB Plugin) does not recognize innodb_file_format, in- nodb_file_format_check, and innodb_strict_mode. See Section B.1, “New Parameters” for a complete list of such configuration parameters that could cause downgrade issues. To include these parameters in the configuration file, use the loose_ form of the parameter names, so that the earlier version of InnoDB can start. In MySQL, configuration options can be specified in the mysqld command line or the option file (my.cnf or my.ini). See Us- ing Option Files for more information. 11.4. Possible Problems Failure to follow the downgrading procedure described in Section 11.3, “How to Downgrade” may lead to compatibility issues when files written by InnoDB 1.1 or the InnoDB Plugin are accessed by an earlier version of InnoDB. This section describes some internal recovery algorithms, to help explain why it is important to follow the downgrade procedure described above. It discusses the issues that may arise, and covers possible ways to fix them. A general fix is to install the plugin as described in Chapter 9, Installing the InnoDB Storage Engine and then follow the downgrad- ing procedure described in Section 11.3, “How to Downgrade”. In the future, the file format management features described in Chapter 4, InnoDB File Format Management will guard against the types of problems described in this section. 11.4.1. Accessing COMPRESSED or DYNAMIC Tables The built-in InnoDB in MySQL can only open tables that were created in REDUNDANT or COMPACT format. Starting with MySQL version 5.0.21, an attempt to open a table in some other format results in ERROR 1146 (42S02): Table 'test.t' doesn't exist. Also, a message “unknown table type” appears in the error log. In InnoDB 1.1, you can rebuild an incompatible table by issuing a statement ALTER TABLE table_name ROW_FORMAT=COMPACT. 11.4.2. Issues with UNDO and REDO As noted in Section 11.3, “How to Downgrade”, you should ensure a “slow” shutdown is done with the InnoDB storage engine, be- fore running with the built-in InnoDB in MySQL, to clean up all buffers. To initiate a slow shutdown, execute the command SET GLOBAL innodb_fast_shutdown=0 before initiating the shutdown of the InnoDB storage engine. We recommend “slow” shutdown (innodb_fast_shutdown=0) because the InnoDB storage engine may write special records to the transaction undo log that cause problems if the built-in InnoDB in MySQL attempts to read the log. Specifically, these spe- cial records are written when a record in a COMPRESSED or DYNAMIC table is updated or deleted and the record contains columns stored off-page. The built-in InnoDB in MySQL cannot read these undo log records. Also, the built-in InnoDB in MySQL cannot roll back incomplete transactions that affect tables that it is unable to read (tables in COMPRESSED or DYNAMIC format). Note that a “normal” shutdown does not necessarily empty the undo log. A normal shutdown occurs when in- nodb_fast_shutdown=1, the default. When InnoDB is shut down, some active transactions may have uncommitted modifica- tions, or they may be holding a read view that prevents the purging of some version information from the undo log. The next time InnoDB is started after a normal shutdown (innodb_fast_shutdown=1), it rolls back any incomplete transactions and purge old version information. Therefore, it is important to perform a “slow” shutdown (innodb_fast_shutdown=0) as part of the downgrade process. In case it is not possible to have the InnoDB storage engine clear the undo log, you can prevent the built-in InnoDB in MySQL from accessing the undo log by setting innodb_force_recovery=3. However, this is not a recommended approach, since in addition to preventing the purge of old versions, this recovery mode prevents the rollback of uncommitted transactions. For more information, see Forcing InnoDB Recovery. When it comes to downgrading, there are also considerations with respect to redo log information. For the purpose of crash recov- ery, InnoDB writes to the log files information about every modification to the data files. When recording changes to tables that were created in DYNAMIC or COMPRESSED format, the InnoDB storage engine writes redo log entries that cannot be recognized by the built-in InnoDB in MySQL. The built-in InnoDB in MySQL refuses to start if it sees any unknown entries in the redo log. When InnoDB is shut down cleanly, it flushes all unwritten changes from the buffer pool to the data files and makes a checkpoint in the redo log. When InnoDB is subsequently restarted, it scans the redo log starting from the last checkpoint. After a clean shut- down, InnoDB crash recovery only then sees the end-of-log marker in the redo log. In this case, the built-in InnoDB in MySQL would not see any unrecognizable redo log entries. This is a second reason why you should ensure a clean, slow shutdown of MySQL (innodb_fast_shutdown=0) before you attempt a downgrade. In an emergency, you may prevent the redo log scan and the crash recovery from the redo log by setting the parameter in- nodb_force_recovery=6. However, this is strongly discouraged, because it may lead into severe corruption. See Forcing 43
- 52. Downgrading the InnoDB Storage Engine InnoDB Recovery for more information. 11.4.3. Issues with the Doublewrite Buffer InnoDB uses a novel file flush technique called doublewrite. Before writing pages to a data file, InnoDB first writes them to a con- tiguous area called the doublewrite buffer. Only after the write and the flush to the doublewrite buffer have completed does InnoDB write the pages to their proper positions in the data file. If the operating system crashes in the middle of a page write, InnoDB can later find a good copy of the page from the doublewrite buffer during recovery. The doublewrite buffer may also contain compressed pages. However, the built-in InnoDB in MySQL cannot recognize such pages, and it assumes that compressed pages in the doublewrite buffer are corrupted. It also wrongly assumes that the tablespace (the .ibd file) consists of 16K byte pages. Thus, you may find InnoDB warnings in the error log of the form “a page in the doublewrite buffer is not within space bounds”. The doublewrite buffer is not scanned after a clean shutdown. In an emergency, you may prevent crash recovery by setting in- nodb_force_recovery=6. However, this is strongly discouraged, because it may lead into severe corruption. See Forcing InnoDB Recovery for more information. 11.4.4. Issues with the Insert Buffer Secondary indexes are usually non-unique, and DML operations on secondary indexes happen in a relatively random order. This would cause a lot of random disk I/O operations without a special mechanism used in InnoDB called the insert buffer. When a change is made to a non-unique secondary index page that is not in the buffer pool, InnoDB inserts the record into a special B-tree: the insert buffer. Periodically, the insert buffer is merged into the secondary index trees in the database. A merge also oc- curs whenever a secondary index page is loaded to the buffer pool. A “normal” shutdown does not clear the insert buffer. A normal shutdown occurs when innodb_fast_shutdown=1, the de- fault. If the insert buffer is not empty when the InnoDB storage engine is shut down, it may contain changes for tables in DYNAMIC or COMPRESSED format. Thus, starting the built-in InnoDB in MySQL on the data files may lead into a crash if the insert buffer is not empty. A “slow” shutdown merges all changes from the insert buffer. To initiate a slow shutdown, execute the command SET GLOBAL innodb_fast_shutdown=0 before initiating the shutdown of the InnoDB storage engine. To disable insert buffer merges, set innodb_force_recovery=4 so that you can back up the uncompressed tables with the built-in InnoDB in MySQL. Be sure not to use any WHERE conditions that would require access to secondary indexes. For more in- formation, see Forcing InnoDB Recovery. In the InnoDB storage engine 1.0.3 and later, you can disable the buffering of new operations by setting the parameter in- nodb_change_buffering. See Section 7.4, “Controlling InnoDB Change Buffering” for details. 44
- 53. Chapter 12. InnoDB Storage Engine Change History 12.1. Changes in InnoDB Storage Engine 1.x Since InnoDB 1.1 is tightly integrated with MySQL 5.5, for changes after the initial InnoDB 1.1 release, see the MySQL 5.5 Refer- ence Manual, Changes in Release 5.5.x (Production). 12.2. Changes in InnoDB Storage Engine 1.1 (April 13, 2010) For an overview of the changes, see this introduction article for MySQL 5.5 with InnoDB 1.1. The following is a condensed ver- sion of the change log. Fix for bug #52580: Crash in ha_innobase::open on executing INSERT with concurrent ALTER TABLE. Change in MySQL bug #51557 releases the mutex LOCK_open before ha_innobase::open(), causing racing condition for index translation table creation. Fix it by adding dict_sys mutex for the operation. Add support for multiple buffer pools. Fix Bug #26590: MySQL does not allow more than 1023 open transactions. Create additional rollback segments on startup. Reduce the upper limit of total rollback segments from 256 to 128. This is because we can't use the sign bit. It has not caused problems in the past because we only created one segment. InnoDB has always had the capability to use the additional rollback segments, there- fore this patch is backward compatible. The only requirement to maintain backward compatibility has been to ensure that the addi- tional segments are created after the double write buffer. This is to avoid breaking assumptions in the existing code. Implement Performance Schema in InnoDB. Objects in four different modules in InnoDB have been performance instrumented, these modules are: mutexes, rwlocks, file I/O, and threads We mostly preserved the existing APIs, but APIs would point to instru- mented function wrappers if performance schema is defined. There are 4 different defines that controls the instrumentation of each module. The feature is off by default, and will be compiled in with special build option, and requre configure option to turn it on when server boots. Implement the buf_pool_watch for DeleteBuffering in the page hash table. This serves two purposes. It allows multiple watches to be set at the same time (by multiple purge threads) and it removes a race condition when the read of a block completes about the time the buffer pool watch is being set. Introduce a new mutex to protect flush_list. Redesign mtr_commit() in a way that log_sys mutex is not held while all mtr_memos are popped and is released just after the modified blocks are inserted into the flush_list. This should reduce contention on log_sys mutex. Implement the global variable innodb_change_buffering, with the following values: • none: buffer nothing • inserts: buffer inserts (like InnoDB so far) • deletes: buffer delete-marks • changes: buffer inserts and delete-marks • purges: buffer delete-marks and deletes • all: buffer all operations (insert, delete-mark, delete) The default is all. All values except none and inserts will make InnoDB write new-format records to the insert buffer, even for inserts. Provide support for native AIO on Linux. 12.3. Changes in InnoDB Plugin 1.0.x The InnoDB 1.0.x releases that accompany MySQL 5.1 have their own change history. Changes up to InnoDB 1.0.8 are listed at ht- tp://dev.mysql.com/doc/innodb-plugin/1.0/en/innodb-changes.html. Changes from InnoDB 1.0.9 and up are listed in Changes in Release 5.1.x (Production), incorporated into the main MySQL change log. 45
- 54. Appendix A. Third-Party Software Oracle acknowledges that certain Third Party and Open Source software has been used to develop or is incorporated in the InnoDB storage engine. This appendix includes required third-party license information. A.1. Performance Patches from Google Innobase Oy gratefully acknowledges the following contributions from Google, Inc. to improve InnoDB performance: • Replacing InnoDB’s use of Pthreads mutexes with calls to GCC atomic builtins, as discussed in Section 7.2, “Faster Locking for Improved Scalability”. This change means that InnoDB mutex and and rw-lock operations take less CPU time, and im- proves throughput on those platforms where the atomic operations are available. • Controlling master thread I/O rate, as discussed in Section 7.11, “Controlling the Master Thread I/O Rate”. The master thread in InnoDB is a thread that performs various tasks in the background. Historically, InnoDB has used a hard coded value as the total I/O capacity of the server. With this change, user can control the number of I/O operations that can be performed per second based on their own workload. Changes from the Google contributions were incorporated in the following source code files: btr0cur.c, btr0sea.c, buf0buf.c, buf0buf.ic, ha_innodb.cc, log0log.c, log0log.h, os0sync.h, row0sel.c, srv0srv.c, srv0srv.h, srv0start.c, sync0arr.c, sync0rw.c, sync0rw.h, sync0rw.ic, sync0sync.c, sync0sync.h, sync0sync.ic, and univ.i. These contributions are incorporated subject to the conditions contained in the file COPYING.Google, which are reproduced here. Copyright (c) 2008, 2009, Google Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the Google Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. A.2. Multiple Background I/O Threads Patch from Percona Innobase Oy gratefully acknowledges the contribution of Percona, Inc. to improve InnoDB performance by implementing configur- able background threads, as discussed in Section 7.8, “Multiple Background I/O Threads”. InnoDB uses background threads to ser- vice various types of I/O requests. The change provides another way to make InnoDB more scalable on high end systems. Changes from the Percona, Inc. contribution were incorporated in the following source code files: ha_innodb.cc, os0file.c, os0file.h, srv0srv.c, srv0srv.h, and srv0start.c. This contribution is incorporated subject to the conditions contained in the file COPYING.Percona, which are reproduced here. Copyright (c) 2008, 2009, Percona Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above 46
- 55. Third-Party Software copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the Percona Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. A.3. Performance Patches from Sun Microsystems Innobase Oy gratefully acknowledges the following contributions from Sun Microsystems, Inc. to improve InnoDB performance: • Introducing the PAUSE instruction inside spin loops, as discussed in Section 7.13, “Using the PAUSE instruction in InnoDB spin loops”. This change increases performance in high concurrency, CPU-bound workloads. • Enabling inlining of functions and prefetch with Sun Studio. Changes from the Sun Microsystems, Inc. contribution were incorporated in the following source code files: univ.i, ut0ut.c, and ut0ut.h. This contribution is incorporated subject to the conditions contained in the file COPYING.Sun_Microsystems, which are re- produced here. Copyright (c) 2009, Sun Microsystems, Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of Sun Microsystems, Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 47
- 56. Appendix B. List of Parameters Changed in InnoDB 1.1 and InnoDB Plugin 1.0 B.1. New Parameters Throughout the course of development, InnoDB 1.1 and its predecessor the InnoDB Plugin introduced new configuration paramet- ers. The following table summarizes those parameters: Table B.1. InnoDB 1.1 New Parameter Summary Name Cmd- Option System Scope Dynamic Default Line File Var innodb_adaptive_flushing YES YES YES GLOBAL YES TRUE innodb_buffer_pool_instances YES YES YES GLOBAL YES TRUE innodb_change_buffering YES YES YES GLOBAL YES inserts innodb_file_format YES YES YES GLOBAL YES Antelope innodb_file_format_check YES YES YES GLOBAL NO 1 innodb_file_format_max YES YES YES GLOBAL YES Antelope for a new data- base; Bar- racuda if any tables using that file format exist in the database innodb_io_capacity YES YES YES GLOBAL YES 200 innodb_old_blocks_pct YES YES YES GLOBAL YES 37 innodb_old_blocks_time YES YES YES GLOBAL YES 0 innodb_purge_batch_size YES YES YES GLOBAL YES 0 innodb_purge_threads YES YES YES GLOBAL YES 0 innodb_read_ahead_threshold YES YES YES GLOBAL YES 56 innodb_read_io_threads YES YES YES GLOBAL NO 4 innodb_spin_wait_delay YES YES YES GLOBAL YES 6 innodb_stats_sample_pages YES YES YES GLOBAL YES 8 innodb_strict_mode YES YES YES GLOB- YES FALSE AL|SESSI ON innodb_use_native_aio YES YES YES GLOBAL NO TRUE innodb_use_sys_malloc YES YES YES GLOBAL NO TRUE innodb_write_io_threads YES YES YES GLOBAL NO 4 • innodb_adaptive_flushing Whether InnoDB uses a new algorithm to estimate the required rate of flushing. The default value is TRUE. This parameter was added in InnoDB storage engine 1.0.4. See Section 7.12, “Controlling the Flushing Rate of Dirty Pages” for more information. • innodb_change_buffering Whether InnoDB performs insert buffering. The default value is "inserts" (buffer insert operations). This parameter was added in InnoDB storage engine 1.0.3. See Section 7.4, “Controlling InnoDB Change Buffering” for more information. • innodb_file_format The default file format for new InnoDB tables. The default is “Antelope”. To enable support for table compression, change it to “Barracuda”. This parameter was added in InnoDB storage engine 1.0.1. See Section 4.1, “Enabling File Formats” for more in- formation. 48
- 57. List of Parameters Changed in InnoDB 1.1 and InnoDB Plugin 1.0 • innodb_file_format_check and innodb_file_format_max Controls whether InnoDB performs file format compatibility checking when opening a database. The default value is in- nodb-file-format-check=1, with innodb_file_format_max set to the highest format that is used in the database (either “Barracuda” or “Antelope”). See Section 4.2.1, “Compatibility Check When InnoDB Is Started” for more information. • innodb_io_capacity The number of I/O operations that can be performed per second. The allowable value range is any number 100 or greater, and the default value is 200. This parameter was added in InnoDB storage engine 1.0.4. To reproduce the earlier behavior, use a value of 100. See Section 7.11, “Controlling the Master Thread I/O Rate” for more information. • innodb_old_blocks_pct Controls the desired percentage of “old” blocks in the LRU list of the buffer pool. The default value is 37 and the allowable value range is 5 to 95. This parameter was added in InnoDB storage engine 1.0.5. See Section 7.15, “Making Buffer Cache Scan Resistant” for more information. • innodb_old_blocks_time The time in milliseconds since the first access to a block during which it can be accessed again without being made “young”. The default value is 0 which means that blocks are moved to the “young” end of the LRU list at the first access. This parameter was added in InnoDB storage engine 1.0.5. See Section 7.15, “Making Buffer Cache Scan Resistant” for more information. • innodb_read_ahead_threshold Control the sensitivity of the linear read ahead. The allowable value range is 0 to 64 and the default value is 56. This paramet- er was added in InnoDB storage engine 1.0.4. See Section 7.7, “Changes in the Read-Ahead Algorithm” for more information. • innodb_read_io_threads The number of background I/O threads used for reads. The allowable value range is 1 to 64 and the default value is 4. This parameter was added in InnoDB storage engine 1.0.4. See Section 7.8, “Multiple Background I/O Threads” for more informa- tion. • innodb_spin_wait_delay Maximum delay between polling for a spin lock. The allowable value range is 0 (meaning unlimited) or positive integers and the default value is 6. This parameter was added in InnoDB storage engine 1.0.4. See Section 7.14, “Control of Spin Lock Polling” for more information. • innodb_stats_sample_pages The number of index pages to sample when calculating statistics. The allowable value range is 1-unlimited and the default value is 8. This parameter was added in InnoDB storage engine 1.0.2. See Section 8.5, “Controlling Optimizer Statistics Estim- ation” for more information. • innodb_strict_mode Whether InnoDB raises error conditions in certain cases, rather than issuing a warning. This parameter was added in InnoDB storage engine 1.0.2. See Section 8.4, “InnoDB Strict Mode” for more information. • innodb_use_sys_malloc Whether InnoDB uses its own memory allocator or an allocator of the operating system. The default value is ON (use an allocat- or of the underlying system). This parameter was added in InnoDB storage engine 1.0.3. See Section 7.3, “Using Operating System Memory Allocators” for more information. • innodb_write_io_threads The number of background I/O threads used for writes. The allowable value range is 1 to 64 and the default value is 4. This parameter was added in InnoDB storage engine 1.0.4. See Section 7.8, “Multiple Background I/O Threads” for more informa- tion. B.2. Deprecated Parameters Beginning in InnoDB storage engine 1.0.4, the following configuration parameter has been removed: • innodb_file_io_threads 49
- 58. List of Parameters Changed in InnoDB 1.1 and InnoDB Plugin 1.0 This parameter has been replaced by two new parameters innodb_read_io_threads and in- nodb_write_io_threads. See Section 7.8, “Multiple Background I/O Threads” for more information. B.3. Parameters with New Defaults For better out-of-the-box performance, the following InnoDB configuration parameters have new default values since MySQL 5.1: Table B.2. InnoDB Parameters with New Defaults Name Old Default New Default innodb_additional_mem_pool_size 1MB 8MB innodb_buffer_pool_size 8MB 128MB innodb_change_buffering inserts all innodb_file_format_check ON 1 innodb_log_buffer_size 1MB 8MB innodb_max_dirty_pages_pct 90 75 innodb_sync_spin_loops 20 30 innodb_thread_concurrency 8 0 50
- 59. Appendix C. InnoDB Glossary These terms are commonly used in information about the InnoDB storage engine. A ACID An acronym standing for atomicity, consistency, isolation, and durability. These properties are all desirable in a database system, and are all closely tied to the notion of a transaction. The transactional features of InnoDB adhere to the ACID principles. Transactions are atomic units of work that can be committed or rolled back. When a transac- tion makes multiple changes to the database, either all the changes succeed when the transac- tion is committed, or all the changes are undone when the transaction is rolled back. The database remains in a consistent state at all times -- after each commit or rollback, and while transactions are in progress. Transactions are protected (isolated) from each other while they are in progress; they cannot interfere with each other or see each other's uncommitted data. This isolation is achieved through the notion of locking. (Expert users can adjust the isolation level, trading off less protection in favor of increased performance, when they can be sure that the transactions really do not interfere with each other.) The results of transactions are durable: once a commit operation succeeds, the changes made by that transaction are safe from power failures, system crashes, race conditions, or other po- tential dangers that many non-database applications are vulnerable to. Durability typically in- volves writing to disk storage, with a certain amount of redundancy to protect against power failures or software crashes during write operations. (In InnoDB, the doublewrite buffer as- sists with durability.) See also commit. See also doublewrite buffer. See also isolation level. See also locking. See also rollback. See also transaction. adaptive flushing An algorithm, first introduced in the InnoDB Plugin, that smooths out the I/O overhead in- troduced by checkpoints. Instead of flushing all modified pages from the buffer pool to the data files at once, the InnoDB storage engine periodically flushes small sets of modified pages. The adaptive flushing algorithm extends this process by estimating the optimal rate to perform these periodic flushes, based on the rate of flushing and how fast redo information is generated. See also buffer pool. See also checkpoint. See also data files. See also flush. See also plugin. adaptive hash index If a table fits almost entirely in main memory, the fastest way to perform queries on it is to use hash indexes. InnoDB has a mechanism that monitors index searches made to the in- dexes defined for a table. If InnoDB notices that queries could benefit from a hash index, it builds one automatically. The hash index is always built based on an existing B-tree index on the table. InnoDB can build a hash index on a prefix of any length of the key defined for the B-tree, depending on the pattern of searches that InnoDB observes for the B-tree index. A hash index can be par- tial; the whole B-tree index does not need to be cached in the buffer pool. InnoDB builds hash indexes on demand for those pages of the index that are often accessed. 51
- 60. InnoDB Glossary In a sense, the adaptive hash index configures InnoDB at runtime to take advantage of ample main memory, coming closer to the architecture of main-memory databases. See also B-tree. AIO Acronym for asynchronous I/O. You might see this acronym in InnoDB messages or keywords. See also asynchronous I/O. Antelope The code name for the original InnoDB file format. It supports the redundant and compact row formats, but not the newer dynamic and compressed row formats available in the Bar- racuda file format. You can select the file format to use through the innodb_file_format option. See also Barracuda. See also compact row format. See also compressed row format. See also dynamic row format. See also file format. See also ibdata file. See also innodb_file_format. See also redundant row format. See also row format. application programming interface A set of functions or procedures. An API provides a stable set of names and types for func- (API) tions, procedures, parameters, and return values. apply When a backup produced by the MySQL Enterprise Backup product does not include the most recent changes that occurred while the backup was underway, the process of updating the backup files to include those changes is known as the apply step. It is specified by the - -apply-log option of the ibbackup or innobackup commands. Before the changes are applied, we refer to the files as a raw backup. After the changes are applied, we refer to the files as a prepared backup. The changes are recorded in the ib- backup_logfile file; once the apply step is finished, this file is no longer necessary. See also hot backup. See also ibbackup_logfile. See also MySQL Enterprise Backup. See also prepared backup. See also raw backup. .ARM file Archive storage engine metadata. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also .ARZ file. See also innobackup command. See also MySQL Enterprise Backup. .ARZ file Archive storage engine data. Files with this extension are always included in backups pro- duced by the innobackup command of the MySQL Enterprise Backup product. See also .ARM file. See also innobackup command. 52
- 61. InnoDB Glossary See also MySQL Enterprise Backup. asynchronous I/O A type of I/O operation that allows other processing to proceed before the I/O is completed. Also known as non-blocking I/O and abbreviated as AIO. InnoDB uses this type of I/O for certain operations that can run in parallel without affecting the reliability of the database, such as reading pages into the buffer pool that have not actually been requested, but might be needed soon. Historically, InnoDB has used asynchronous I/O on Windows systems only. Starting with the InnoDB Plugin 1.1, InnoDB uses asynchronous I/O on Linux systems. This change intro- duces a dependency on libaio. On other Unix-like systems, InnoDB uses synchronous I/O only. See also buffer pool. See also non-blocking I/O. auto-increment A property of a table column (specified by the AUTO_INCREMENT keyword) that automat- ically adds an ascending sequence of values in the column. InnoDB supports auto-increment only for primary key columns. It saves work for the developer, not to have to produce new unique values when inserting new rows. It provides useful information for the query optimizer, because the column is known to be not null and with unique values. The values from such a column can be used as lookup keys in various contexts, and because they are auto-generated there is no reason to ever change them; for this reason, primary key columns are often specified as auto- incrementing. Auto-increment columns can be problematic with statement-based replication, because re- playing the statements on a slave might not produce the same set of column values as on the master, due to timing issues. When you have an auto-incrementing primary key, you can use statement-based replication only with the setting innodb_autoinc_lock_mode=1. If you have innodb_autoinc_lock_mode=2, which allows higher concurrency for insert operations, use row-based replication rather than statement-based replication. The setting innodb_autoinc_lock_mode=0 is the previous (traditional) default setting and should not be used except for compatibility purposes. See also auto-increment locking. See also innodb_autoinc_lock_mode. See also primary key. See also row-based replication. See also statement-based replication. auto-increment locking The convenience of an auto-increment primary key involves some tradeoff with concur- rency. In the simplest case, if one transaction is inserting values into the table, any other transactions must wait to do their own inserts into that table, so that rows inserted by the first transaction receive consecutive primary key values. InnoDB includes optimizations, and the innodb_autoinc_lock_mode option, so that you can choose how to trade off between predictable sequences of auto-increment values and maximum concurrency for insert opera- tions. See also auto-increment. See also concurrency. See also innodb_autoinc_lock_mode. autocommit A setting that causes a commit operation after each SQL statement. This mode is not usually recommended for working with InnoDB tables, where transactions typically span several statements. It is more appropriate for working with MyISAM tables where transactions are not applicable, or in specialized situations with InnoDB to minimize overhead from locking and generation of undo data. See also commit. See also locking. 53
- 62. InnoDB Glossary See also SQL. See also transaction. See also undo. B B-tree A tree data structure that is popular for use in database indexes. The structure is kept sorted at all times, en- abling fast lookup for exact matches (equals operator) and ranges (for example, greater than, less than, and BETWEEN operators). This type of index is available for most storage engines, such as InnoDB and MyIS- AM. Because B-tree nodes can have many children, a B-tree is not the same as a binary tree, which is limited to 2 children per node. Contrast with hash index, which is only available in the MEMORY storage engine. The MEMORY storage engine can also use B-tree indexes, and you should choose B-tree indexes for MEMORY tables if some queries use range operators. See also hash_index. backticks Identifiers within MySQL SQL statements must be quoted using the backtick character (`) if they contain special characters or reserved words. For example, to refer to a table named FOO#BAR or a column named SELECT, you would specify the identifiers as `FOO#BAR` and `SELECT`. Since the backticks provide an extra level of safety, they are used extensively in program-generated SQL statements, where the identifi- er names might not be known in advance. Many other database systems use double quotation marks (") around such special names. For portability, you can enable ANSI_QUOTES mode in MySQL and use double quotation marks instead of backticks to qualify identifier names. See also SQL. Barracuda The code name for an InnoDB file format that supports compression for table data. This file format was first introduced in the InnoDB Plugin. It supports the dynamic and compressed row formats. You can se- lect it through the innodb_file_format option. The MySQL Enterprise Backup product version 3.5 and above supports backing up tablespaces that use the Barracuda file format. See also Antelope. See also compact row format. See also compressed row format. See also dynamic row format. See also file format. See also ibdata file. See also innodb_file_format. See also row format. beta An early stage in the life of a software product, when it is available only for evaluation, typically without a definite release number or a number less than 1. InnoDB does not use the beta designation, preferring an early adopter phase that can extend over several point releases, leading to a GA release. See also early adopter. See also GA. binary log A file containing a record of all statements that attempt to change table data. These statements can be re- played to bring slave servers up to date in a replication scenario, or to bring a database up to date after restoring table data from a backup. The binary logging feature can be turned on and off, although Oracle re- 54
- 63. InnoDB Glossary commends always enabling it if you use replication or perform backups. You can examine the contents of the binary log, or replay those statements during replication or recovery, by using the mysqlbinlog command. For the MySQL Enterprise Backup product, the file name of the binary log and the current position within the file are important details. To record this information for the master server when taking a backup in a replication context, you can specify the --slave-info option. Prior to MySQL 5.0, a similar capability was available, known as the update log. In MySQL 5.0 and higher, the binary log replaces the update log. See also binlog. See also MySQL Enterprise Backup. binlog An informal name for the binary log file. For example, you might see this abbreviation used in e-mail mes- sages or forum discussions. See also binary log. bottleneck A portion of a system that is constrained in size or capacity, that has the effect of limiting overall through- put. For example, a memory area might be smaller than necessary; access to a single required resource might prevent multiple CPU cores from running simultaneously; or waiting for disk I/O to complete might prevent the CPU from running at full capacity. Removing bottlenecks tends to improve concurrency. For example, the ability to have multiple InnoDB buffer pool instances reduces contention when multiple ses- sions read from and write to the buffer pool simultaneously. See also buffer pool. See also concurrency. buffer A memory or disk area used for temporary storage. Data is buffered in memory so that it can be written to disk efficiently, with a few large I/O operations rather than many small ones. Data is buffered on disk for greater reliability, so that it can be recovered even when a crash or other failure occurs at the worst possible time. The main types of buffers used by InnoDB are the buffer pool, the doublewrite buffer, and the in- sert buffer. See also buffer pool. See also doublewrite buffer. See also insert buffer. buffer pool The memory area that holds cached InnoDB data for both tables and indexes. For efficiency of high-volume read operations, the buffer pool is divided into pages that can potentially hold multiple rows. For efficiency of cache management, the buffer pool is implemented as a linked list of pages; data that is rarely used is aged out of the cache, using a variation of the LRU algorithm. On systems with large memory, you can im- prove concurrency by dividing the buffer pool into multiple instances. See also LRU. See also page. built-in The built-in InnoDB storage engine within MySQL is the original form of distribution for the storage en- gine. Contrast with the InnoDB Plugin. Starting with MySQL 5.5, the InnoDB Plugin is merged back into the MySQL code base as the built-in InnoDB storage engine (known as InnoDB 1.1). This distinction is important mainly in MySQL 5.1, where a feature or bug fix might apply to the InnoDB Plugin but not the built-in InnoDB, or vice versa. See also InnoDB. See also plugin. business rules The relationships and sequences of actions that form the basis of business software, used to run a commer- cial company. Sometimes these rules are dictated by law, other times by company policy. Careful planning ensures that the relationships encoded and enforced by the database, and the actions performed through ap- plication logic, accurately reflect the real policies of the company and can handle real-life situations. For example, an employee leaving a company might trigger a sequence of actions from the human resources 55
- 64. InnoDB Glossary department. The human resources database might also need the flexibility to represent data about a person who has been hired, but not yet started work. Closing an account at an online service might result in data being removed from a database, or the data might be moved or flagged so that it could be recovered if the account is re-opened. A company might establish policies regarding salary maximums, minimums, and ad- justments, in addition to basic sanity checks such as the salary not being a negative number. A retail data- base might not allow a purchase with the same serial number to be returned more than once, or might not al- low credit card purchases above a certain value, while a database used to detect fraud might allow these kinds of things. See also relational. See also trigger. C cache The general term for any memory area that stores copies of data for frequent or high-speed re- trieval. In InnoDB, the primary kind of cache structure is the buffer pool. See also buffer. See also buffer pool. cardinality The number of different values in a table column. When queries refer to columns that have an as- sociated index, the cardinality of each column influences which access method is most efficient. If the values in a column have a very uneven distribution, the cardinality might not be a good way to determine the best query plan. (http://bugs.mysql.com/bug.php?id=36513) Cardinality can also apply to the number of distinct values present in multiple columns, as in a composite index. See also composite index. See also index. See also random dive. See also selectivity. change buffering The general term for the features involving the insert buffer, consisting of insert buffering, de- lete buffering, and purge buffering. Index changes resulting from SQL statements, which could normally involve random I/O operations, are held back and performed periodically by a back- ground thread. This sequence of operations can write the disk blocks for a series of index values more efficiently than if each value were written to disk immediately. See also delete buffering. See also insert buffer. See also insert buffering. See also purge buffering. checkpoint As changes are made to data pages that are cached in the buffer pool, those changes are written to the data files sometime later, a process known as flushing. The checkpoint is a record of the latest changes (represented by an LSN value) that have been successfully written to the data files. See also buffer pool. See also data files. See also flush. See also LSN. child table In a foreign key relationship, a child table is one whose rows refer (or point) to rows in another table with an identical value for a specific column. The corresponding row in the parent table must exist before the row can be created in the child table. The values in the child table can pre- vent delete or update operations on the parent table, or can cause automatic deletion or updates in 56
- 65. InnoDB Glossary the child table, based on the ON CASCADE option used when creating the foreign key. See also foreign key. See also parent table. client A type of program that sends requests to a server, and interprets or processes the results. The cli- ent software might run only some of the time (such as a mail or chat program), and might run in- teractively (such as the mysql command processor). See also mysql. See also server. clustered index The InnoDB term for a primary key index. InnoDB table storage is organized based on the val- ues of the primary key columns, to speed up queries and sorts involving the primary key columns. For best performance, choose the primary key columns carefully based on the most per- formance-critical queries. Because modifying the columns of the clustered index is an expensive operation, choose primary columns that are rarely or never updated. In the Oracle Database product, this type of table is known as an index-organized table. See also index. See also primary key. See also secondary index. cold backup A backup taken while the database is shut down. For busy applications and web sites, this might not be practical, and you might prefer a warm backup or a hot backup. See also hot backup. See also warm backup. column index An index on a single column. See also composite index. See also index. column prefix When an index is created with a length specification, such as CREATE INDEX idx ON t1 (c1(N)), only the first N characters of the column value are stored in the index. Keeping the in- dex prefix small makes the index compact, and the memory and disk I/O savings help perform- ance. (Although making the index prefix too small can hinder query optimization by making rows with different values appear to the query optimizer to be duplicates.) For columns containing binary values or long text strings, where sorting is not a major considera- tion and storing the entire value in the index would waste space, the index automatically uses the first N (typically 768) characters of the value to do lookups and sorts. See also index. commit A SQL statement that ends a transaction, making permanent any changes made by the transac- tion. It is the opposite of rollback, which undoes any changes made in the transaction. InnoDB uses an optimistic mechanism for commits, so that changes can be written to the data files before the commit actually occurs. This technique makes the commit itself faster, with the tradeoff that more work is required in case of a rollback. By default, MySQL uses the autocommit setting, which automatically issues a commit following each SQL statement. See also autocommit. See also optimistic. See also rollback. See also SQL. 57
- 66. InnoDB Glossary See also transaction. compact row format The default InnoDB row format since MySQL 5.0.3. It has a more compact representation for nulls and variable-length fields than the prior default (redundant row format). Because of the B-tree indexes that make row lookups so fast in InnoDB, there is little if any per- formance benefit to keeping all rows the same size. See also redundant row format. See also row format. composite index An index that includes multiple columns. See also index. See also index prefix. compressed backup A set of files produced by the MySQL Enterprise Backup product, where each tablespace is compressed. The compressed files are renamed with a .ibz file extension. Applying compression right at the start of the backup process helps to avoid storage overhead during the compression process, and to avoid network overhead when transferring the backup files to another server. The process of applying the binary log takes longer, and requires uncom- pressing the backup files. See also apply. See also binary log. See also compression. See also hot backup. See also MySQL Enterprise Backup. See also tablespace. compressed row format A row format introduced in the InnoDB Plugin, available as part of the Barracuda file format. Large fields are stored away from the page that holds the rest of the row data, as in dynamic row format. Both index pages and the large fields are compressed, yielding memory and disk savings. Depending on the structure of the data, the decrease in memory and disk usage might or might not outweigh the performance overhead of uncompressing the data as it is used. See also Barracuda. See also dynamic row format. See also row format. compression A feature with wide-ranging benefits from using less disk space, performing less I/O, and using less cache memory. InnoDB products include more than one feature related to compression: table data can be kept in a compressed format during database operation; data can be compressed as part of a backup operation with the MySQL Enterprise Backup product. When InnoDB table data is compressed, the compression applies to the table itself, any associ- ated index data, and the pages loaded into the buffer pool. Compression does not apply to pages in the undo buffer. The table compression feature requires using MySQL 5.5 or higher, or the InnoDB Plugin in MySQL 5.1 or earlier, and creating the table using the Barracuda file format and compressed row format, with the innodb_file_per_table setting turned on. The compression feature of the MySQL Enterprise Backup product makes a compressed copy of each tablespace, changing the extension from .ibd to .ibz. Compressing the backup data al- lows you to keep more backups on hand, and reduces the time to transfer backups to a different server. The data is uncompressed during the restore operation. When a compressed backup opera- tion processes a table that is already compressed, it skips the compression step for that table, be- cause compressing again would result in little or no space savings. 58
- 67. InnoDB Glossary See also Barracuda. See also buffer pool. See also compressed row format. See also hot backup. See also index. See also plugin. See also table. See also undo buffer. concatenated index See composite index. concurrency The ability of multiple operations (in database terminology, transactions) to run simultaneously, without interfering with each other. Concurrency is also involved with performance, because ideally the protection for multiple simultaneous transactions works with a minimum of perform- ance overhead, using efficient mechanisms for locking. See also ACID. See also locking. See also transaction. configuration file The file that holds the option values used by MySQL at startup. Traditionally, on Linux and UNIX this file is named my.cnf, and on Windows it is named my.ini. You can set a number of options related to InnoDB under the [mysqld] section of the file. When you use the MySQL Enterprise Backup product, you typically use two configuration files: one that specifies where the data comes from and how it is structured (which could be the original configuration file for your real server), and a stripped-down one containing only a small set of options that specify where the backup data goes and how it is structured. The configuration files used with the MySQL Enterprise Backup product must contain certain options that are typically left out of regular configuration files, so you might need to add some options to your ex- isting configuration file for use with MySQL Enterprise Backup. See also my.cnf. See also option file. consistent read A read operation that uses snapshot information to present query results based on a point in time, regardless of changes performed by other transactions running at the same time. If queried data has been changed by another transaction, the original data is reconstructed based on the contents of the undo log. This technique avoids some of the locking issues that can reduce concurrency by forcing transactions to wait for other transactions to finish. With the repeatable read isolation level, the snapshot is based on the time when the first read operation is performed. With the read committed isolation level, the snapshot is reset to the time of each consistent read operation. Consistent read is the default mode in which InnoDB processes SELECT statements in READ COMMITTED and REPEATABLE READ isolation levels. Because a consistent read does not set any locks on the tables it accesses, other sessions are free to modify those tables while a con- sistent read is being performed on the table. See also ACID. See also concurrency. See also isolation level. See also locking. See also MVCC. 59
- 68. InnoDB Glossary See also read committed. See also read uncommitted. See also repeatable read. See also serializable read. See also transaction. See also undo log. constraint An automatic test that can block database changes to prevent data from becoming inconsistent. (In computer science terms, a kind of assertion related to an invariant condition.) Constraints are a crucial component of the ACID philosophy, to maintain data consistency. Constraints suppor- ted by MySQL include foreign key constraints and unique constraints. See also ACID. See also foreign key. See also relational. See also unique constraint. covering index An index that includes all the columns retrieved by a query. Instead of using the index values as pointers to find the full table rows, the query returns values from the index structure, saving disk I/O. InnoDB can apply this optimization technique to more indexes than MyISAM can, because InnoDB secondary indexes also include the primary key columns. InnoDB cannot apply this technique for queries against tables modified by a transactions, until that transaction ends. Any column index or composite index could act as a covering index, given the right query. Design your indexes and queries to take advantage of this optimization technique wherever pos- sible. See also column index. See also composite index. See also index. See also secondary index. crash recovery The cleanup activities that occur when InnoDB is started again after a crash. Changes that were committed before the crash, but not yet written into the tablespace files, are reconstructed from the doublewrite buffer. When the database is shut down normally, this type of activity is per- formed during shutdown by the purge operation. During normal operation, committed data can be stored in the insert buffer for a period of time before being written to the tablespace files. There is always a tradeoff between keeping the ta- blespace files up-to-date, which introduces performance overhead during normal operation, and buffering the data, which can make shutdown and crash recovery take longer. See also doublewrite buffer. See also insert buffer. See also purge. cursor An internal data structure that is used to represent the result set of a query, or other operation that performs a search using a SQL WHERE clause. It works like an iterator in other high-level lan- guages, producing each value from the result set as requested. Although usually SQL handles the processing of cursors for you, you might delve into the inner workings when dealing with performance-critical code. See also query. D 60
- 69. InnoDB Glossary data definition language See DDL. data dictionary A set of tables, controlled by the InnoDB storage engine, that keeps track of InnoDB-related objects such as tables, indexes, and table columns. These tables are part of the InnoDB sys- tem tablespace. Because the MySQL Enterprise Backup product always backs up the system tablespace, all backups include the contents of the data dictionary. See also hot backup. See also MySQL Enterprise Backup. See also system tablespace. data files The files that physically contain the InnoDB table and index data. There can be a one- to-many relationship between data files and tables, as in the case of the system tablespace, which can hold multiple InnoDB tables as well as the data dictionary. There can also be a one-to-one relationship between data files and tables, as when the file-per-table setting is enabled, causing each newly created table to be stored in a separate tablespace. See also data dictionary. See also file-per-table. See also system tablespace. See also tablespace. data manipulation language See DML. data warehouse A database system or application that primarily runs large queries. The read-only or read- mostly data might be organized in denormalized form for query efficiency. Contrast with OLTP. See also denormalized. See also OLTP. See also query. database An InnoDB database is largely defined by its data files. For long-time MySQL users, a database is a familiar notion. Users coming from an Oracle Database background will find that the MySQL meaning of a database is closer to what Or- acle Database calls a schema. See also data files. DDL Data definition language, a set of SQL statements for manipulating the database itself rather than individual table rows. Includes all forms of the CREATE, ALTER, and DROP statements. Also includes the TRUNCATE statement, because it works differently than a DELETE FROM table_name statement, even though the ultimate effect is similar. DDL statements automatically commit the current transaction; they cannot be rolled back. See also commit. See also rollback. See also SQL. See also transaction. deadlock A situation where different transactions are unable to proceed, because each holds a lock that the other needs. Because both transactions are waiting for a resource to become avail- able, neither will ever release the locks it holds. A deadlock can occur when the transactions acquire locks on multiple tables, but in the op- posite order. A deadlock can also occur when statements such as UPDATE or SELECT ... FOR UPDATE lock ranges of index records and gaps, with each transaction acquiring some 61
- 70. InnoDB Glossary locks but not others due to a timing issue. To reduce the possibility of deadlocks, use transactions rather than LOCK TABLE state- ments; keep transactions that insert or update data small enough that they do not stay open for long periods of time; when different transactions update multiple tables or large ranges of rows, use the same order of operations (such as SELECT ... FOR UPDATE) in each transaction; create indexes on the columns used in SELECT ... FOR UPDATE and UP- DATE ... WHERE statements. The possibility of deadlocks is not affected by the isolation level, because the isolation level changes the behavior of read operations, while deadlocks occur because of write operations. If a deadlock does occur, InnoDB detects the condition and rolls back one of the transac- tions (the victim). Thus, even if your application logic is perfectly correct, you must still handle the case where a transaction must be retried. To monitor how frequently deadlocks occur, use the command SHOW ENGINE INNODB STATUS. See also concurrency. See also gap. See also isolation level. See also lock. See also locking. See also rollback. See also transaction. See also victim. deadlock detection A mechanism that automatically detects when a deadlock occurs, and automatically rolls back one of the transactions involved (the victim). See also deadlock. See also rollback. See also transaction. See also victim. delete When InnoDB processes a DELETE statement, the rows are immediately marked for deletion and no longer are returned by queries. The storage is reclaimed sometime later, during the periodic garbage collection known as the purge operation, performed by a separate thread. See also purge. delete buffering The technique of storing index changes due to DELETE operations in the insert buffer rather than writing them immediately, so that the physical writes can be performed to minim- ize random I/O. (Because delete operations are a two-step process, this operation buffers the write that normally marks an index record for deletion.) It is one of the types of change buf- fering; the others are insert buffering and purge buffering. See also change buffering. See also insert buffer. See also insert buffering. See also purge buffering. denormalized A data storage strategy that duplicates data across different tables, rather than linking the tables with foreign keys and join queries. Typically used in data warehouse applications, where the data is not updated after loading. In such applications, query performance is more important than making it simple to maintain consistent data during updates. Contrast with normalized. See also normalized. 62
- 71. InnoDB Glossary descending index See also index. dirty page A page in the InnoDB buffer pool that has been updated in memory, where the changes are not yet written (flushed) to the data files. See also buffer pool. See also flush. See also page. See also write combining. dirty read An operation that retrieves unreliable data, data that was updated by another transaction but not yet committed. It is only possible with the isolation level known as read uncommitted. This kind of operation does not adhere to the ACID principle of database design. It is con- sidered very risky, because the data could be rolled back, or updated further before being committed; then, the transaction doing the dirty read would be using data that was never con- firmed as accurate. Its polar opposite is consistent read, where InnoDB goes to great lengths to ensure that a transaction does not read information updated by another transaction, even if the other trans- action commits in the meantime. See also ACID. See also commit. See also consistent read. See also isolation level. See also read committed. See also read uncommitted. See also rollback. disk-based A kind of database that primarily organizes data on disk storage (hard drives or equivalent). Data is brought back and forth between disk and memory to be operated upon. It is the op- posite of an in-memory database. InnoDB is disk-based. See also in-memory database. DML Data manipulation language, a set of SQL statements for performing insert, update, and de- lete operations. The SELECT statement is sometimes considered as a DML statement, be- cause the SELECT ... FOR UPDATE form is subject to the same considerations for lock- ing as INSERT, UPDATE, and DELETE. DML statements operate in the context of a transaction, so their effects can be committed or rolled back as a single unit. See also commit. See also locking. See also rollback. See also SQL. See also transaction. doublewrite buffer InnoDB uses a novel file flush technique called doublewrite. Before writing pages to a data file, InnoDB first writes them to a contiguous area called the doublewrite buffer. Only after the write and the flush to the doublewrite buffer have completed, does InnoDB write the pages to their proper positions in the data file. If the operating system crashes in the middle of a page write, InnoDB can later find a good copy of the page from the doublewrite buffer during crash recovery. 63
- 72. InnoDB Glossary Although data is always written twice, the doublewrite buffer does not require twice as much I/O overhead or twice as many I/O operations. Data is written to the buffer itself as a large sequential chunk, with a single fsync call to the operating system. The doublewrite buffer can be turned off by specifying the option in- nodb_doublewrite=0. See also crash recovery. See also purge. dynamic row format A row format introduced in the InnoDB Plugin, available as part of the Barracuda file format. Because TEXT and BLOB fields are stored outside of the rest of the page that holds the row data, it is very efficient for rows that include large objects, resulting in fewer I/O op- erations when the large fields do not need to be accessed. See also Barracuda. See also row format. E early adopter A stage similar to beta, when a software product is typically evaluated for performance, functionality, and compatibility in a non-mission-critical setting. InnoDB uses the early adopter designation rather than beta, through a succession of point releases leading up to a GA release. See also beta. See also GA. Embedded InnoDB A product that allows the capabilities of the InnoDB storage engine to be embedded within an applica- tion program, the same as any other library, without the need for a full MySQL installation. Instead of using SQL, the application manipulates data using C API calls. It is intended for applications that re- quire maximal performance and minimal configuration and setup, such as in the embedded systems market. See also cursor. eviction The process of removing an item from a cache or other temporary storage area. Often, but not always, uses the LRU algorithm to determine which item to remove. See also LRU. exclusive lock A kind of lock that prevents any other transaction from locking the same row. Depending on the trans- action isolation level, this kind of lock might block other transactions from writing to the same row, or might also block other transactions from reading the same row. The default InnoDB isolation level, REPEATABLE READ, enables higher concurrency by allowing transactions to read rows that have exclusive locks, a technique known as consistent read. See also concurrency. See also consistent read. See also isolation level. See also lock. See also repeatable read. See also shared lock. See also transaction. explicit row lock F 64
- 73. InnoDB Glossary fast index creation A capability first introduced in the InnoDB Plugin, that speeds up creation of secondary indexes by avoiding the need to completely rewrite the associated table. The speedup applies to drop- ping secondary indexes also. Because index maintenance can add performance overhead to many data transfer operations, consider doing operations such as ALTER TABLE ... ENGINE=INNODB or INSERT IN- TO ... SELECT * FROM ... without any secondary indexes in place, and creating the indexes afterward. Even if you do not use the InnoDB Plugin as your primary storage engine, you can take advant- age of this capability by enabling the Plugin temporarily, just to create or drop indexes, and then switch back to the built-in InnoDB storage engine for normal use. See also index. See also secondary index. fast shutdown A shutdown procedure that is required before installation of the InnoDB Plugin. From the MySQL command line, issue the following command before performing the shutdown: SET GLOBAL innodb_fast_shutdown=0; To make this type of shutdown the default, specify by the configuration parameter in- nodb_fast_shutdown=0. See also shutdown. See also slow shutdown. file format The format used by InnoDB for its data files named ibdata1, ibdata2, and so on. Each file format supports one or more row formats. See also Antelope. See also Barracuda. See also ibdata file. See also row format. file-per-table A general name for the setting controlled by the innodb_file_per_table option. For each table created while this setting is in effect, the data is stored in a separate file rather than in the system tablespace. When table data is stored in a separate file, you have more flexibility to choose non-default file formats and row formats, which are required for features such as data compression. The TRUNCATE TABLE operation is also much faster, and the reclaimed space can be used by the operating system rather than remaining reserved for InnoDB. This mode is the default setting in MySQL 5.5.5 and higher. The MySQL Enterprise Backup product is more flexible for tables that are in their own files. For example, tables can be excluded from a backup, but only if they are in separate files. Thus, this setting is suitable for tables that are backed up less frequently or on a different schedule. See also file format. See also ibdata file. See also innodb_file_per_table. See also row format. See also system tablespace. fixed row format This row format is used by the MyISAM storage engine, not by InnoDB. If you create an In- noDB table with the option row_format=fixed, InnoDB translates this option to use the compact row format instead, although the fixed value might still show up in output such as SHOW TABLE STATUS reports. See also compact row format. See also row format. 65
- 74. InnoDB Glossary flush To write changes to the database files, that had been buffered in a memory area or a temporary disk storage area. The InnoDB storage structures that are periodically flushed include the redo log, the undo log, and the buffer pool. Flushing can happen because a memory area becomes full and the system needs to free some space, because a commit operation means the changes from a transaction can be finalized, or because a shutdown operation means that all outstanding work should be finalized. See also buffer pool. See also commit. See also redo log. See also undo log. flush list An internal InnoDB data structure that tracks pages in the buffer pool that have been changed and need to be written back out to disk. This data structure is updated frequently by InnoDB's internal mini-transactions, and so is protected by its own mutex to allow concurrent access to the buffer pool. See also buffer pool. See also mini-transaction. See also mutex. See also page. foreign key A type of pointer relationship, between rows in separate InnoDB tables. The foreign key rela- tionship is defined on one column in both the parent table and the child table. In addition to enabling fast lookup of related information, foreign keys help to enforce referen- tial integrity, by preventing any of these pointers from becoming invalid as data is inserted, up- dated, and deleted. This enforcement mechanism is a type of constraint. A row that points to another table cannot be inserted if the associated foreign key value does not exist in the other ta- ble. If a row is deleted or its foreign key value changed, and rows in another table point to that foreign key value, the foreign key can be set up to prevent the deletion, cause the corresponding column values in the other table to become null, or automatically delete the corresponding rows in the other table. One of the stages in designing a normalized database is to identify data that is duplicated, sep- arate that data into a new table, and set up a foreign key relationship so that the multiple tables can be queried like a single table, using a join operation. See also child table. See also foreign key constraint. See also join. See also normalized. See also NULL. See also parent table. See also relational. foreign key constraint The type of constraint that maintains database consistency through a foreign key relationship. Unlike other kinds of constraints that prevent data from being inserted or updated, foreign key constraints can also cause data in child rows to be deleted, changed to different values, or set to null, based on the ON CASCADE option specified when creating the foreign key. See also child table. See also constraint. See also foreign key. See also NULL. 66
- 75. InnoDB Glossary .FRM file The file containing the metadata, such as the table definition, of a MySQL table. For backups, you must always keep the full set of .FRM files along with the backup data, to be able to restore tables that are altered or dropped after the backup. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. If you use the ibbackup command instead, you must copy the .FRM files yourself. Although each InnoDB table has a .FRM file, InnoDB maintains its own table metadata in the system tablespace; the .FRM files are not needed for InnoDB to operate on InnoDB tables. These files are backed up by the MySQL Enterprise Backup product. These files must not be modified by an ALTER TABLE operation while the backup is taking place, which is why backups that include non-InnoDB tables perform a FLUSH TABLES WITH READ LOCK op- eration to freeze such activity while backing up the .FRM files. Restoring a backup can result in .FRM files being created, changed, or removed to match the state of the database at the time of the backup. See also ibbackup command. See also innobackup command. See also MySQL Enterprise Backup. full table scan An operation that requires reading the entire contents of a table, rather than just selected por- tions using an index. Typically performed either with small lookup tables, or in data warehous- ing situations with large tables where all available data is aggregated and analyzed. How fre- quently these operations occur, and the sizes of the tables relative to available memory, have implications for the algorithms used in query optimization and managing the buffer pool. The purpose of indexes is to allow lookups for specific values or ranges of values within a large table, thus avoiding full table scans when practical. See also buffer pool. See also index. See also LRU. G GA "Generally available", the stage when a software product leaves beta and is available for sale, official support, and production use. See also beta. See also early adopter. gap A place in an InnoDB index data structure where new values could be inserted. When you lock a set of rows with a statement such as SELECT ... FOR UPDATE, InnoDB can create locks that apply to the gaps as well as the actual values in the index. For example, if you select all values greater than 10 for update, a gap lock prevents another transaction from inserting a new value that is greater than 10. Gap locks are part of the tradeoff between performance and concurrency, and are used in some transaction isolation levels and not others. See also concurrency. See also index. See also infimum record. See also isolation level. See also supremum record. group commit An InnoDB optimization that performs some low-level I/O operations (log write) once for a set of commit operations, rather than flushing and syncing separately for each commit. 67
- 76. InnoDB Glossary When the binlog is enabled, you typically also set the configuration option sync_binlog=0, because group commit for the binary log is only supported if it is set to 0. See also commit. See also log write. See also plugin. See also XA. H hash_index A type of index intended for queries that use equality operators rather than range operators such as greater-than or BETWEEN. It is available in only the MEMORY storage engine. Although hash indexes are the default for MEMORY tables for historic reasons, for general-purpose queries, B-tree indexes are often a better choice. The InnoDB storage engine includes a variant of this index type, the adaptive hash index, that is constructed automatically if needed based on runtime conditions. See also adaptive hash index. See also B-tree. See also index. heartbeat A periodic message that is sent to indicate that a system is functioning properly. In a replication context, if the master stops sending such messages, one of the slaves can take its place. Similar techniques can be used between the servers in a cluster environment, to confirm that all of them are operating properly. See also master database. See also replication. See also slave database. hot A condition where a row, table, or internal data structure is accessed so frequently, in a context that requires some form of locking or mutual exclusion, that it results in a performance or scalability issue. Although “hot” typically indicates an undesirable condition, a hot backup is the preferred kind of backup. See also hot backup. hot backup A backup taken while the database and is running and applications are reading and writing to it. The backup in- volves more than simply copying data files: it must include any data that was inserted or updated while the backup was in process; it must exclude any data that was deleted while the backup was in process; and it must ig- nore any changes that were not committed. The Oracle product that performs hot backups, of InnoDB tables especially but also tables from MyISAM and other storage engines, is known as MySQL Enterprise Backup. The hot backup process consists of two stages. The initial copying of the data files produces a raw backup. The apply step incorporates any changes to the database that happened while the backup was running. Applying the changes produces a prepared backup; these files are ready to be restored whenever necessary. See also apply. See also MySQL Enterprise Backup. See also prepared backup. See also raw backup. I ib-file set The set of files managed by InnoDB within a MySQL database: the system tablespace, any 68
- 77. InnoDB Glossary file-per-table tablespaces, and the (typically 2) log files. Used sometimes in detailed discus- sions of InnoDB file structures and formats, to avoid ambiguity between the meanings of database between different DBMS products, and the non-InnoDB files that may be part of a MySQL database. See also database. See also schema. ib_logfile A set of files, typically named ib_logfile0 and ib_logfile1, that form the redo log. These files record statements that attempt to change data in InnoDB tables. These statements are replayed automatically to correct data written by incomplete transactions, on startup fol- lowing a crash. This data can not be used for manual recovery; for that type of operation, use the binary log. See also binary log. See also redo log. ibbackup command The fundamental command-line tool of the MySQL Enterprise Backup product. It per- forms a hot backup operation for InnoDB tables. You use this command directly if all your data is in InnoDB tables, if all the important data that you need to back up is in InnoDB tables, or if you are running on the Windows platform. If you also need to back up tables from MyISAM or other storage engines, you use the innobackup command instead (available on UNIX and Linux systems only). See also hot backup. See also innobackup command. See also MySQL Enterprise Backup. ibbackup_logfile A supplemental backup file created by the MySQL Enterprise Backup product during a hot backup operation. It contains information about any data changes that occurred while the backup was running. The initial backup files, including ibbackup_logfile, are known as a raw backup, because the changes that occurred during the backup operation are not yet incorporated. After you perform the apply step to the raw backup files, the resulting files do include those final data changes, and are known as a prepared backup. At this stage, the ibbackup_logfile file is no longer necessary. See also apply. See also hot backup. See also MySQL Enterprise Backup. See also prepared backup. See also raw backup. ibd file Each InnoDB tablespace created using the file-per-table setting has a filename with a .ibd extension. This extension does not apply to the system tablespace, which is made up of files named ibdata1, ibdata2, and so on. See also file-per-table. See also ibz file. See also system tablespace. See also tablespace. ibdata file A set of files with names such as ibdata1, ibdata2, and so on, that make up the InnoDB system tablespace. These files contain metadata about InnoDB tables, and can contain some or all of the table data also (depending on whether the file-per-table option is in effect when each table is created). See also Antelope. See also Barracuda. 69
- 78. InnoDB Glossary See also file format. See also file-per-table. See also system tablespace. ibz file When the MySQL Enterprise Backup product performs a compressed backup, it trans- forms each tablespace file that is created using the file-per-table setting from a .ibd exten- sion to a .ibz extension. The compression applied during backup is distinct from the compressed row format that keeps table data compressed during normal operation. A compressed backup operation skips the compression step for a tablespace that is already in compressed row format, as compress- ing a second time would slow down the backup but produce little or no space savings. See also compressed backup. See also compressed row format. See also file-per-table. See also ibd file. See also MySQL Enterprise Backup. See also tablespace. implicit row lock See also locking. in-memory database A type of database system that maintains data in memory, to avoid overhead due to disk I/O and translation between disk blocks and memory areas. Some in-memory databases sacrifice durability (the "D" in the ACID design philosophy) and are vulnerable to hardware, power, and other types of failures, making them more suitable for read-only operations. Other in- memory databases do use durability mechanisms such as logging changes to disk or using non-volatile memory. See also ACID. See also disk-based. incremental backup A type of hot backup, performed by the MySQL Enterprise Backup product, that only saves data changed since some point in time. Having a full backup and a succession of incre- mental backups lets you reconstruct backup data over a long period, without the storage overhead of keeping several full backups on hand. You can restore the full backup and then apply each of the incremental backups in succession, or you can keep the full backup up- to-date by applying each incremental backup to it, then perform a single restore operation. The granularity of changed data is at the page level. A page might actually cover more than one row. Each changed page is included in the backup. See also hot backup. See also MySQL Enterprise Backup. See also page. index A data structure that provides a fast lookup capability for rows of a table, typically by form- ing a tree structure representing all the values of a particular column or set of columns. See also B-tree. See also clustered index. See also column index. See also composite index. See also covering index. See also partial index. 70
- 79. InnoDB Glossary See also primary key. See also secondary index. index prefix In an index that applies to multiple columns (known as a composite index), the initial or leading columns of the index. A query that references the first 1, 2, 3, and so on columns of a composite index can use the index, even if the query does not reference all the columns in the index. See also composite index. See also index. infimum record A pseudo-record in an index, representing the gap below the smallest value in that index. If a transaction has a statement such as SELECT ... FOR UPDATE ... WHERE col < 10;, and the smallest value in the column is 5, it is a lock on the infimum record that pre- vents other transactions from inserting even smaller values such as 0, -10, and so on. See also gap. See also pseudo-record. See also supremum record. information schema The actual name of the schema that contains the data dictionary. To examine information (metadata) about the database, you can query tables such as INFORMA- TION_SCHEMA.TABLES and INFORMATION_SCHEMA.COLUMNS. The InnoDB Plugin introduced some tables into the information schema that are specific to InnoDB, such as INNODB_LOCKS and INNODB_TRX. You use these tables not to see how the database is structured, but to get real-time information about the workings of the InnoDB storage engine that can help with performance monitoring, tuning, and troubleshooting. In particular, these tables provide information about the InnoDB features related to compres- sion, and transactions and their associated locks. See also compression. See also data dictionary. See also locking. See also schema. See also transaction. innobackup command A command-line tool of the MySQL Enterprise Backup product. It performs a hot backup operation for tables from InnoDB, MyISAM, and other storage engines. It calls the ib- backup command to back up the InnoDB tables; if you only need to back up InnoDB tables, it can be more efficient to call ibbackup directly. This command is not available on the Windows platform. See also hot backup. See also ibbackup command. See also MySQL Enterprise Backup. InnoDB A storage engine for MySQL that combines high performance with transactional capability (that is, reliability, robustness, and concurrent access). Tables created under the InnoDB stor- age engine are ideally suited for hot backups. See also ACID. See also hot backup. See also storage engine. See also transaction. InnoDB Hot Backup A licensed backup product, superceded in MySQL 5.1 and above by MySQL Enterprise 71
- 80. InnoDB Glossary Backup. See also MySQL Enterprise Backup. innodb_autoinc_lock_mode The option that controls the algorithm used for auto-increment locking. When you have an auto-incrementing primary key, you can use statement-based replication only with the setting innodb_autoinc_lock_mode=1. This setting is known as consecutive lock mode, be- cause multi-row inserts within a transaction receive consecutive auto-increment values. If you have innodb_autoinc_lock_mode=2, which allows higher concurrency for insert operations, use row-based replication rather than statement-based replication. This setting is known as interleaved lock mode, because multiple multi-row insert statements running at the same time can receive autoincrement values that are interleaved. The setting in- nodb_autoinc_lock_mode=0 is the previous (traditional) default setting and should not be used except for compatibility purposes. See also auto-increment locking. innodb_file_format A setting that determines the file format for all tablespaces created after you specify a value for this option. To create tablespaces other than the system tablespace, you must use the file-per-table option also. See also Antelope. See also Barracuda. See also file format. See also file-per-table. See also innodb_file_per_table. See also system tablespace. See also tablespace. innodb_file_per_table The option that allows you to use the file-per-table setting, which stores newly created tables in their own data files, outside the system tablespace. This option is needed to take full advantage of many other features, such as such as table compression in the InnoDB Plu- gin, or backups of named tables in MySQL Enterprise Backup. This option was once static, but can now be set using the SET GLOBAL command. See also data files. See also file-per-table. See also system tablespace. innodb_lock_wait_timeout A setting that provides an alternative to automatic deadlock detection. Rolls back any trans- action that waits more than a specified time to acquire a lock. Especially useful if deadlocks are caused by updates to multiple tables controlled by different storage engines; such dead- locks are not detected automatically. See also deadlock. See also deadlock detection. See also victim. innodb_locks_unsafe_for_binlog See also locking. innodb_strict_mode The option that controls whether InnoDB operates in strict mode, where conditions that are normally treated as warnings, cause errors instead (and the underlying statements fail). This mode is the default setting in MySQL 5.5.5 and higher. See also strict mode. innodb_table_locks See also locking. 72
- 81. InnoDB Glossary insert One of the primary DML operations in SQL. The performance of inserts is a key factor in data warehouse systems that load millions of rows into tables, and OLTP systems where many concurrent connections might insert rows into the same table, in arbitrary order. If in- sert performance is important to you, you should learn about InnoDB features such as the in- sert buffer used in change buffering, and auto-increment columns. See also auto-increment. See also change buffering. See also DML. See also insert buffer. See also SQL. insert buffer A special index data structure that records changes to pages in secondary indexes. These val- ues could result from SQL INSERT UPDATE, or DELETE statements. The set of features in- volving the insert buffer is known collectively as change buffering, consisting of insert buffering, delete buffering, and purge buffering. Changes are only recorded in the insert buffer when the relevant page from the secondary in- dex is not in the buffer pool. When the relevant index page is brought into the buffer pool while associated changes are still in the insert buffer, the changes for that page are applied in the buffer pool using the data from the insert buffer. Periodically, the purge operation that runs during times when the system is mostly idle, or during a slow shutdown, writes the new index pages to disk. The purge operation can write the disk blocks for a series of index val- ues more efficiently than if each value were written to disk immediately. Physically, the insert buffer is part of the system tablespace, so that the index changes re- main buffered across database restarts. The changes are only applied when the pages are brought into the buffer pool due to some other read operation. To see information about the current data in the insert buffer, issue the SHOW INNODB STATUS command. See also buffer pool. See also change buffering. See also delete buffering. See also insert buffering. See also page. See also purge. See also purge buffering. See also system tablespace. insert buffering The technique of storing secondary index changes due to INSERT operations in the insert buffer rather than writing them immediately, so that the physical writes can be performed to minimize random I/O. It is one of the types of change buffering; the others are delete buf- fering and purge buffering. Insert buffering is not used if the secondary index is unique, because the uniqueness of new values cannot be verified before the new entries are written out. Other kinds of change buf- fering do work for unique indexes. See also change buffering. See also delete buffering. See also insert buffer. See also purge buffering. insert-intention gap lock See also gap. 73
- 82. InnoDB Glossary See also intention lock. See also lock. See also locking. intention exclusive lock See intention lock. intention lock A kind of lock that applies to the table level, used to indicate what kind of lock the transac- tion intends to acquire on rows in the table. Different transactions can acquire different kinds of intention locks on the same table, but the first transaction to acquire an intention exclus- ive (IX) lock on a table prevents other transactions from acquiring any S or X locks on the table. Conversely, the first transaction to acquire an intention shared (IS) lock on a table prevents other transactions from acquiring any X locks on the table. The two-phase process allows the lock requests to be resolved in order, without blocking locks and corresponding operations that are compatible. See also insert-intention gap lock. See also lock. See also lock mode. See also locking. intention shared lock See intention lock. isolation level One of the foundations of database processing. Isolation is the I in the acronym ACID; the isolation level is the setting that fine-tunes the balance between performance and reliability, consistency, and reproducibility of results when multiple transactions are making changes and performing queries at the same time. From highest amount of consistency and protection to the least, the isolation levels supported by InnoDB are: serializable read, repeatable read, consistent read, and read uncommit- ted. With the built-in InnoDB storage engine and the InnoDB Plugin, many users can keep the default isolation level (repeatable read) for all operations. Expert users might choose the read committed level as they push the boundaries of scalability with OLTP processing, or during data warehousing operations where minor inconsistencies do not affect the aggregate results of large amounts of data. The levels on the edges (serializable read and read un- committed) change the processing behavior to such an extent that they are rarely used. See also ACID. See also consistent read. See also read uncommitted. See also repeatable read. See also serializable read. See also transaction. J join A query that retrieves data from more than one table, by referencing columns in the tables that hold identical values. These columns are frequently part of a foreign key relationship. See also foreign key. See also query. K 74
- 83. InnoDB Glossary key_block_size An option to specify the size of data pages within a table that uses compressed row format. The default is 8 kilobytes. Lower values risk hitting internal limits that depend on the combination of row size and com- pression percentage. See also compressed row format. L latch A lightweight structure used by InnoDB to implement a lock for its own internal memory structures, typic- ally held for a brief time measured in milliseconds or microseconds. A general term that includes both mutexes (for exclusive access) and rw-locks (for shared access). Certain latches are the focus of perform- ance tuning within the InnoDB storage engine, such as the data dictionary mutex. Statistics about latch use and contention are available through the Performance Schema interface. See also data dictionary. See also lock. See also locking. See also mutex. See also Performance Schema. See also rw-lock. list The InnoDB buffer pool is represented as a list of memory pages. The list is reordered as new pages are accessed and enter the buffer pool, as pages within the buffer pool are accessed again and are considered newer, and as pages that are not accessed for a long time are replaced in the buffer pool. See also buffer pool. See also eviction. See also LRU. See also sublist. lock The high-level notion of an object that controls access to a resource, such as a table, row, or internal data structure, as part of a locking strategy. For intensive performance tuning, you might delve into the actual structures that implement locks, such as mutexes and latches. See also latch. See also lock mode. See also locking. See also mutex. See also Pthreads. lock escalation An operation used in some database systems that converts many row locks into a single table lock, saving memory space but reducing concurrent access to the table. InnoDB uses a space-efficient representation for row locks, so that lock escalation is not needed. See also locking. See also row lock. See also table lock. lock mode A shared (S) lock allows a transaction to read a row. Multiple transactions can acquire an S lock on that same row at the same time. An exclusive (X) lock allows a transaction to update or delete a row. No other transaction can acquire any kind of lock on that same row at the same time. Intention locks apply to the table level, and are used to indicate what kind of lock the transaction intends 75
- 84. InnoDB Glossary to acquire on rows in the table. Different transactions can acquire different kinds of intention locks on the same table, but the first transaction to acquire an intention exclusive (IX) lock on a table prevents other transactions from acquiring any S or X locks on the table. Conversely, the first transaction to acquire an intention shared (IS) lock on a table prevents other transactions from acquiring any X locks on the table. The two-phase process allows the lock requests to be resolved in order, without blocking locks and corres- ponding operations that are compatible. See also intention lock. See also lock. See also locking. lock monitor See also deadlock. See also lock. See also locking. locking The system of protecting a transaction from seeing or changing data that is being queried or changed by other transactions. The locking strategy must balance reliability and consistency of database operations (the principles of the ACID philosophy) against the performance needed for good concurrency. Fine- tuning the locking strategy often involves choosing an isolation level and ensuring all your database oper- ations are safe and reliable for that isolation level. See also ACID. See also concurrency. See also isolation level. See also latch. See also lock. See also mutex. See also transaction. log write See also group commit. logical backup A backup that reproduces table structure and data, without copying the actual data files. For example, the mysqldump command produces a logical backup, because its output contains statements such as CRE- ATE TABLE and INSERT that can re-create the data. Contrast with physical backup. See also physical backup. loose_ A prefix added to InnoDB configuration options when installing the InnoDB Plugin after server startup, so any new configuration options not recognized by the current level of MySQL do not cause a startup fail- ure. MySQL processes configuration options that start with this prefix, but gives a warning rather than a failure if the part after the prefix is not a recognized option. For details, see the MySQL documentation. LRU An acronym meaning “least recently used”, a common method for managing storage areas. The items that have not been used recently are removed, when space is needed to cache newer items. InnoDB uses the LRU mechanism by default to manage the pages within the buffer pool, but makes exceptions in cases where a page might be read only a single time, such as during a full-table scan. The amount that the buffer cache policy differs from the strict LRU algorithm is governed by the options in- nodb_old_blocks_pct and innodb_old_blocks_time. See also buffer pool. See also eviction. LSN Acronym for log sequence number. This arbitrary, ever-increasing value represents a point in time cor- responding to operations recorded in the redo log. (This point in time is regardless of transaction boundar- ies; it can fall in the middle of one or more transactions.) It is used internally by InnoDB during crash re- covery and for managing the buffer pool. In the MySQL Enterprise Backup product, you can specify an LSN to represent the point in time from 76
- 85. InnoDB Glossary which to take an incremental backup. The relevant LSN is displayed by the output of the ibbackup command. Once you have the LSN corresponding to the time of a full backup, you can specify that value to take a subsequent incremental backup, whose output contains another LSN for the next incremental backup. See also hot backup. See also incremental backup. See also redo log. M master database In a replication scenario, the database that processes the initial insert, update, and delete re- quests for data. These changes are propagated to, and repeated on, other databases known as slave databases. See also replication. See also slave database. master thread See also thread. mid-point insertion strategy The technique of initially bringing blocks into the buffer pool not at the "newest" end of the list, but instead somewhere in the middle. The exact location of the mid-point can vary, based on the setting of the innodb_old_blocks_pct option. The intent is that blocks that are only read once, such as during a full table scan, can be aged out of the buffer pool sooner than with a strict LRU algorithm. See also buffer pool. See also full table scan. See also LRU. mini-transaction An internal phase of InnoDB processing, when making changes to the data dictionary dur- ing a DDL operation. The changes to the data dictionary are made without affecting the transaction that is being processed by MySQL. See also data dictionary. See also DDL. See also transaction. monitor See also lock monitor. .MRG file A file containing references to other tables, used by the MERGE storage engine. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also MySQL Enterprise Backup. multi-core A type of processor that can take advantage of multi-threaded programs, such as the MySQL server. multiversion concurrency control See MVCC. mutex Informal abbreviation for “mutex variable”. (Mutex itself is short for “mutual exclusion”.) The low-level object that InnoDB uses to represent and enforce exclusive-access locks to in- ternal in-memory data structures. Once the lock is acquired, any other process, thread, and so on is prevented from acquiring the same lock. Contrast with rw-locks, which allow shared access. Mutexes and rw-locks are known collectively as latches. See also latch. 77
- 86. InnoDB Glossary See also lock. See also Performance Schema. See also Pthreads. See also rw-lock. MVCC Acronym for multiversion concurrency control. This technique allows InnoDB transactions with certain isolation levels perform consistent read operations; that is, to query rows that are being updated by other transactions, and see the values from before those updates oc- curred. This is a powerful technique to increase concurrency, by allowing queries to proceed without waiting due to locks held by the other transactions. This technique is not universal in the database world. Some other database products, and some other storage engines within MySQL, do not support it. See also ACID. See also concurrency. See also consistent read. See also isolation level. See also lock. my.cnf The name, on UNIX or Linux systems, of the MySQL option file. See also my.ini. See also option file. my.ini The name, on Windows systems, of the MySQL option file. See also my.cnf. See also option file. .MYD files The files that MySQL uses to store the data for MyISAM tables. Files with this extension are always included in backups produced by the innobackup command of the MySQL En- terprise Backup product. See also innobackup command. See also .MYI files. See also MySQL Enterprise Backup. .MYI files The files that MySQL uses to store the indexes for MyISAM tables. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also .MYD files. See also MySQL Enterprise Backup. mysql The mysql program is the command-line interpreter for the MySQL database. It processes SQL statements, and also MySQL-specific commands such as SHOW TABLES. See also mysqld. MySQL Enterprise Backup A licensed product, superceding InnoDB Hot Backup, that performs hot backups of MySQL databases. It offers the most efficiency and flexibility when backing up InnoDB tables, but can also back up MyISAM and other kinds of tables. See also hot backup. See also InnoDB. 78
- 87. InnoDB Glossary mysqld The mysqld program is the database engine for the MySQL database. It runs as a UNIX daemon or Windows service, constantly waiting for requests and performing maintenance work in the background. See also mysql. mysqldump See also hot backup. N non-blocking I/O An industry term that means the same as asynchronous I/O. See also asynchronous I/O. non-repeatable read The occurrence when a query retrieves data, and a later query within the same transaction retrieves what should be the same data, but the queries return different results (changed by another transaction committing in the meantime). This kind of operation goes against the ACID principle of database design. Within a transaction, data should be consistent, with predictable and stable relationships. Among different isolation levels, non-repeatable reads are prevented by the serializable read and repeatable read levels, and allowed by the consistent read, and read uncommitted levels. See also ACID. See also consistent read. See also isolation level. See also locking. See also read uncommitted. See also repeatable read. See also serializable read. See also transaction. normalized A database design strategy where data is split into multiple tables, and duplicate values condensed into single rows represented by an ID, to avoid storing, querying, and updating redundant or lengthy values. It is typically used in OLTP applications. For example, an address might be given a unique ID, so that a census database could represent the relationship lives at this address by associating that ID with each member of a family, rather than storing multiple copies of a complex value such as 123 Main Street, Anytown, USA. For another example, although a simple address book application might store each phone number in the same table as a person's name and address, a phone company database might give each phone number a special ID, and store the numbers and IDs in a separate table. This normalized representa- tion could simplify large-scale updates when area codes split apart. Normalization is not always recommended. Data that is primarily queried, and only updated by de- leting entirely and reloading, is often kept in fewer, larger tables with redundant copies of duplicate values. This data representation is referred to as denormalized, and is frequently found in data warehousing applications. See also denormalized. See also foreign key. See also OLTP. See also relational. NULL A special value in SQL, indicating the absence of data. Any arithmetic operation or equality test in- volving a NULL value, in turn produces a NULL result. (Thus it is similar to the IEEE floating-point 79
- 88. InnoDB Glossary concept of NaN, not a number.) Any aggregate calculation such as AVG() ignores rows with NULL values, when determining how many rows to divide by. The only test that works with NULL values uses the SQL idioms IS NULL or IS NOT NULL. NULL values play a part in index operations, because for performance a database must minimize the overhead of keeping track of missing data values, Typically, NULL values are not represented in an index, because a query that tests an indexed column using a standard comparison operator could never match a row with a NULL value for that column. For the same reason, unique indexes do not prevent NULL values; those values simply are not represented in the index. Declaring a NOT NULL constraint on a column provides reassurance that there are no rows left out of the index, allowing for better query optimization (accurate counting of rows and estimation of whether to use the index). Because the primary key must be able to uniquely identify every row in the table, a single-column primary key cannot contain any NULL values, and a multi-column primary key cannot contain any rows with NULL values in all columns. Although the Oracle database allows a NULL value to be concatenated with a string, InnoDB treats the result of such an operation as NULL. See also index. See also primary key. See also SQL. O off-page column A column containing variable-length data (such as BLOB and VARCHAR) that are too long to fit on a B- tree page. The data is stored in overflow pages. See also B-tree. See also overflow page. OLTP Acronym for "Online Transaction Processing". A database system, or a database application, that runs a workload with many transactions, with frequent writes as well as reads, typically affecting small amounts of data at a time. For example, an airline reservation system or an application that processes bank deposits. The data might be organized in normalized form for a balance between DML (insert/update/delete) effi- ciency and query efficiency. Contrast with data warehouse. See also data warehouse. See also DML. See also query. See also transaction. .OPT file A file containing database configuration information. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also MySQL Enterprise Backup. optimistic A methodology that guides low-level implementation decisions for a relational database system. The re- quirements of performance and concurrency in a relational database mean that operations must be started or dispatched quickly. The requirements of consistency and referential integrity mean that any operation could fail: a transaction might be rolled back, a DML operation could violate a constraint, a request for a lock could cause a deadlock, a network error could cause a timeout. An optimistic strategy is one that as- sumes most requests or attempts will succeed, so that relatively little work is done to prepare for the fail- ure case. When this assumption is true, the database does little unnecessary work; when requests do fail, extra work must be done to clean up and undo changes. InnoDB uses optimistic strategies for operations such as locking and commits. For example, data changed by a transaction can be written to the data files before the commit occurs, making the commit itself very fast, but requiring more work to undo the changes if the transaction is rolled back. The opposite of an optimistic strategy is a pessimistic one, where a system is optimized to deal with oper- 80
- 89. InnoDB Glossary ations that are unreliable and frequently unsuccessful. This methodology is rare in a database system, be- cause so much care goes into choosing reliable hardware, networks, and algorithms. See also commit. See also locking. See also pessimistic. option A configuration parameter for MySQL. In the context of the InnoDB storage engine (either the built-in InnoDB or the InnoDB Plugin), each op- tion name starts with the prefix innodb_. See also built-in. See also option file. See also plugin. option file Traditionally, on Linux and UNIX this file is named my.cnf, and on Windows it is named my.ini. See also configuration file. See also my.cnf. overflow page Separately allocated disk pages that hold variable-length columns (such as BLOB and VARCHAR) that are too long to fit on a B-tree page. The associated columns are known as off-page columns. See also B-tree. See also off-page column. See also page. P page A unit representing how much data InnoDB transfers at any one time between disk (the data files) and memory (the buffer pool). A page can contain one or more rows, depending on how much data is in each row. If a row does not fit entirely into a single page, InnoDB sets up addi- tional pointer-style data structures so that the information about the row can be stored in one page. One way to fit more data in each page is to use compressed row format. See also buffer pool. See also compressed row format. See also data files. See also page size. page size Currently, this value is fixed at 16 kilobytes. This is considered a reasonable compromise: large enough to hold the data for most rows, yet small enough to minimize the performance overhead of transferring unneeded data to memory. Other values are not tested or supported. See also page. .PAR file A file containing partition definitions. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also MySQL Enterprise Backup. parent table The table in a foreign key relationship that holds the initial column values pointed to from the child table. The consequences of deleting, or updating rows in the parent table depend on the ON UPDATE and ON DELETE clauses in the foreign key definition. Rows with corresponding 81
- 90. InnoDB Glossary values in the child table could be automatically deleted or updated in turn, or those columns could be set to NULL, or the operation could be prevented. See also child table. See also foreign key. partial index An index that represents only part of a column value, typically the first N characters (the prefix) of a long VARCHAR value. See also index. See also index prefix. Performance Schema The performance_schema schema, in MySQL 5.5 and up, presents a set of tables that you can query to get detailed information about the performance characteristics of many internal parts of the MySQL server. See also latch. See also mutex. See also rw-lock. pessimistic A methodology that sacrifices performance or concurrency in favor of safety. It is appropriate if a high proportion of requests or attempts might fail, or if the consequences of a failed request are severe. InnoDB uses what is known as a pessimistic locking strategy, to minimize the chance of deadlocks. At the application level, you might avoid deadlocks by usiing a pessimist- ic strategy of acquiring all locks needed by a transaction at the very beginning. Many built-in database mechanisms use the opposite optimistic methodology. See also deadlock. See also locking. See also optimistic. phantom A row that appears in the result set of a query, but not in the result set of an earlier query. For example, if a query is run twice within a transaction, and in the meantime, another transaction commits after inserting a new row or updating a row so that it matches the WHERE clause of the query. This occurrence is known as a phantom read. It is harder to guard against than a non-re- peatable read, because locking all the rows from the first query result set does not prevent the changes that cause the phantom to appear. Among different isolation levels, phantom reads are prevented by the serializable read level, and allowed by the repeatable read, consistent read, and read uncommitted levels. See also consistent read. See also isolation level. See also non-repeatable read. See also read uncommitted. See also repeatable read. See also serializable read. See also transaction. physical backup A backup that copies the actual data files. For example, the MySQL Enterprise Backup com- mand produces a physical backup, because its output contains data files that can be used directly by the mysqld server. Contrast with logical backup. See also logical backup. See also MySQL Enterprise Backup. 82
- 91. InnoDB Glossary PITR backup Acronym for point-in-time recovery. See also point-in-time recovery. plugin In MySQL 5.1 and earlier, a separately installable form of the InnoDB storage engine that in- cludes features and performance enhancements not included in the built-in InnoDB for those releases. For MySQL 5.5 and higher, the MySQL distribution includes the very latest InnoDB features and performance enhancements, known as InnoDB 1.1, and there is no longer a separate In- noDB Plugin. This distinction is important mainly in MySQL 5.1, where a feature or bug fix might apply to the InnoDB Plugin but not the built-in InnoDB, or vice versa. See also built-in. See also InnoDB. point-in-time recovery The process of restoring a backup to recreate the state of the database at a specific date and time. Commonly abbreviated PITR. Because it is unlikely that the specified time corresponds exactly to the time of a backup, this technique usually requires a combination of a physical backup and a logical backup. For example, with the MySQL Enterprise Backup product, you restore the last backup that you took before the specified point in time, then replay changes from the binary log between the time of the backup and the PITR time. See also logical backup. See also MySQL Enterprise Backup. See also physical backup. See also PITR backup. predicate locking See also locking. prefix See index prefix. prepared backup A set of backup files, produced by the MySQL Enterprise Backup product, after all the stages of applying binary logs and incremental backups are finished. The resulting files are ready to be restored. Prior to the apply steps, the files are known as a raw backup. See also binary log. See also hot backup. See also incremental backup. See also MySQL Enterprise Backup. See also raw backup. See also restore. primary key A set of columns -- and by implication, the index based on this set of columns -- that can uniquely identify every row in a table. As such, it must be a unique index that does not contain any NULL values. InnoDB requires that every table has such an index (also called the clustered index or cluster index), and organizes the table storage based on the column values of the primary key. See also clustered index. See also index. pseudo-record An artificial record in an index, used for locking key values or ranges that do not currently ex- ist. See also infimum record. See also locking. 83
- 92. InnoDB Glossary See also supremum record. Pthreads The POSIX threads standard, which defines an API for threading and locking operations on UNIX and Linux systems. On UNIX and Linux systems, InnoDB uses this implementation for mutexes. See also mutex. purge A type of garbage collection performed by a separate thread, running on a periodic schedule. The purge includes these actions: removing obsolete values from indexes; physically removing rows that were marked for deletion by previous DELETE statements. See also crash recovery. See also delete. See also doublewrite buffer. purge buffering The technique of storing index changes due to DELETE operations in the insert buffer rather than writing them immediately, so that the physical writes can be performed to minimize ran- dom I/O. (Because delete operations are a two-step process, this operation buffers the write that normally purges an index record that was previously marked for deletion.) It is one of the types of change buffering; the others are insert buffering. and delete buffering See also change buffering. See also delete buffering. See also insert buffer. See also insert buffering. purge thread A thread within the InnoDB process that is dedicated to performing the periodic purge opera- tion. See also purge. See also thread. Q query An operation that reads information from one or more tables in a database. Depending on the organization of data and the parameters of the query, the lookup might be optimized by consulting an index. Normally with MySQL, queries are coded using SQL. (Even if you are using some other language for the main applica- tion logic.) See also database. See also index. See also SQL. See also table. R RAID Acronym for "Redundant Array of Inexpensive Drives". Spreading I/O operations across multiple drives enables greater concurrency at the hardware level, and improves the efficiency of low- level write operations that otherwise would be performed in sequence. See also concurrency. random dive A technique for quickly estimating the number of different values in a column (the column's car- dinality). InnoDB samples pages at random from the index and uses that data to estimate the number of different values. This operation occurs when each table is first opened. 84
- 93. InnoDB Glossary Originally, the number of sampled pages was fixed at 8; now, it is determined by the setting of the innodb_stats_sample_pages parameter. The way the random pages are picked depends on the setting of the in- nodb_use_legacy_cardinality_algorithm parameter. The default setting (OFF) has better random- ness than in older releases. See also cardinality. raw backup The initial set of backup files produced by the MySQL Enterprise Backup product, before the changes reflected in the binary log and any incremental backups are applied. At this stage, the files are not ready to restore. After these changes are applied, the files are known as a prepared backup. See also binary log. See also hot backup. See also ibbackup_logfile. See also incremental backup. See also MySQL Enterprise Backup. See also prepared backup. See also restore. read committed An isolation level that uses a locking strategy that relaxes some of the protection between trans- actions, in the interest of performance. Transactions cannot see uncommitted data from other transactions, but they can see data that is committed by another transaction after the current trans- action started. Thus, a transaction never sees any bad data, but the data that it does see may de- pend to some extent on the timing of other transactions. When a transaction with this isolation level performs UPDATE ... WHERE or DELETE ... WHERE operations, other transactions might have to wait. The transaction can perform SELECT ... FOR UPDATE, and LOCK IN SHARE MODE operations without making other transac- tions wait. See also ACID. See also isolation level. See also locking. See also repeatable read. See also serializable read. See also transaction. read uncommitted The isolation level that provides the least amount of protection between transactions. Queries employ a locking strategy that allows them to proceed in situations where they would normally wait for another transaction. However, this extra performance comes at the cost of less reliable results, including data that has been changed by other transactions and not committed yet (known as dirty read). Use this isolation level only with great caution, and be aware that the results might not be consistent or reproducible, depending on what other transactions are doing at the same time. Typically, transactions with this isolation level do only queries, not insert, update, or delete operations. See also ACID. See also dirty read. See also isolation level. See also locking. See also transaction. 85
- 94. InnoDB Glossary read-ahead A type of I/O request that prefetches pages into the buffer cache asynchronously, in anticipation that these pages will be needed soon. record-only lock See also locking. redo log A set of files, typically named ib_logfile0 and ib_logfile1, that record statements that attempt to change data in InnoDB tables. These statements are replayed automatically to correct data written by incomplete transactions, on startup following a crash. See also ib_logfile. redundant row format The oldest InnoDB row format. Prior to MySQL 5.0.3, it was the only row format available in In- noDB. In My SQL 5.0.3 and later, the default is compact row format. You can still specify re- dundant row format for compatibility with older InnoDB tables. See also compact row format. See also row format. referential integrity The technique of maintaining data always in a consistent format, part of the ACID philosophy. In particular, data in different tables is kept in synch through the use of foreign keys, which can pre- vent changes from happening or automatically propagate those changes to all related tables. See also ACID. See also foreign key. relational An important aspect of modern database systems. In a mathematical context, the relations within a database are derived from set theory. For ex- ample, the OR and AND operators of a WHERE clause represent the notions of union and intersec- tion. In everyday terms, the database encodes and enforces relationships such as one-to-one, one- to-many, many-to-one, and uniqueness. For example, a person might have zero, one, or many phone numbers in an address database; a single phone number might be associated with several family members. In a financial database, a person might be required to have exactly one taxpayer ID, and any taxpayer ID could only be associated with one person. At the database level, these relationships are expressed through SQL features such as columns within a table, unique and NOT NULL constraints, foreign keys, and different kinds of join oper- ations. Complex relationships typically involve data split between more than one table. Often, the data is normalized, so that duplicate values in one-to-many relationships are stored only once. See also ACID. See also foreign key. See also normalized. repeatable read The default isolation level for InnoDB. It prevents any rows that are queried from being changed by other transactions, thus blocking non-repeatable reads but not phantom reads. It uses a mod- erately strict locking strategy so that all queries within a transaction see data from the same snap- shot, that is, the data as it was at the time the transaction started. When a transaction with this isolation level performs UPDATE ... WHERE, DELETE ... WHERE, SELECT ... FOR UPDATE, and LOCK IN SHARE MODE operations, other trans- actions might have to wait. See also ACID. See also consistent read. See also isolation level. See also locking. See also serializable read. See also transaction. 86
- 95. InnoDB Glossary replication The practice of sending changes from a master database, to one or more slave databases, so that all databases have the same data. This technique has a wide range of uses, such as load- balancing for better scalability, disaster recovery, and testing software upgrades and configura- tion changes. The changes can be sent between the database by methods called row-based rep- lication and statement-based replication. See also master database. See also row-based replication. See also slave database. See also statement-based replication. restore The process of putting a set of backup files from the MySQL Enterprise Backup product in place for use by MySQL. This operation can be performed to fix a corrupted database, to return to some earlier point in time, or (in a replication context) to set up a new slave database. In the MySQL Enterprise Backup product, this operation is performed by the --copy-back option of the innobackup command. See also hot backup. See also innobackup command. See also MySQL Enterprise Backup. See also prepared backup. See also replication. See also slave database. rollback A SQL statement that ends a transaction, undoing any changes made by the transaction. It is the opposite of commit, which makes permanent any changes made in the transaction. By default, MySQL uses the autocommit setting, which automatically issues a commit following each SQL statement. You must change this setting before you can use the rollback technique. See also ACID. See also commit. See also transaction. rollback segment The storage area containing the undo log, part of the system tablespace. See also system tablespace. See also undo log. row format The disk storage format for a row from an InnoDB table. As InnoDB gains new capabilities such as compression, new row formats are introduced to support the resulting improvements in storage efficiency and performance. Each table has its own row format, specified through the ROW_FORMAT option. To see the row format for each InnoDB table, issue the command SHOW TABLE STATUS. Because all the tables in the system tablespace share the same row format, to take advantage of other row formats typically requires setting the innodb_file_per_table option, so that each table is stored in a separate tablespace. See also compact row format. See also compressed row format. See also dynamic row format. See also fixed row format. See also redundant row format. row lock A lock that prevents a row from being accessed in an incompatible way by another transaction. 87
- 96. InnoDB Glossary See also lock. See also lock mode. See also transaction. row-based replication This form of replication is safe to use for all settings of the innodb_autoinc_lock_mode option. See also auto-increment locking. See also innodb_autoinc_lock_mode. See also replication. See also statement-based replication. rw-lock The low-level object that InnoDB uses to represent and enforce shared-access locks to internal in-memory data structures. Once the lock is acquired, any other process, thread, and so on can read the data structure, but no one else can write to it. Contrast with mutexes, which enforce ex- clusive access. Mutexes and rw-locks are known collectively as latches. See also latch. See also lock. See also mutex. See also Performance Schema. S savepoint Savepoints help to implement nested transactions. They can be used to provide scope to op- erations on tables that are part of a larger transaction. For example, scheduling a trip in a re- servation system might involve booking several different flights; if a desired flight is un- available, you might roll back the changes involved in booking that one leg, without rolling back the earlier flights that were successfully booked. See also rollback. See also transaction. scalability The ability to add more work and issue more simultaneous requests to a system, without a sudden drop in performance due to exceeding the limits of system capacity. Software archi- tecture, hardware configuration, application coding, and type of workload all play a part in scalability. When the system reaches its maximum capacity, popular techniques for increas- ing scalability are scale up (increasing the capacity of existing hardware or software) and scale out (adding new hardware or more instances of server software). schema Conceptually, a schema is a set of interrelated database objects, such as tables, table columns, data types of the columns, indexes, foreign keys, and so on. These objects are con- nected through SQL syntax, because the columns make up the tables, the foreign keys refer to tables and columns, and so on. Ideally, they are also connected logically, working together as part of a unified application or flexible framework. In MySQL, physically, a schema is analogous to a database. Typically, all the objects for a MySQL application are contained within a single database. Some other database products draw a distinction. For example, in the Oracle Database product, a schema represents only a part of a database: the tables and other objects owned by a single user. See also database. See also ib-file set. See also information schema. secondary index A type of InnoDB index that represents a subset of table columns. An InnoDB table can 88
- 97. InnoDB Glossary have zero, one, or many secondary indexes. (Contrast with the clustered index, which is re- quired for each InnoDB table, and stores the data for all the table columns.) A secondary index can be used to satisfy queries that only require values from the indexed columns. For more complex queries, it can be used to identify the relevant rows in the table, which are then retrieved through lookups using the clustered index. Creating and dropping secondary indexes has traditionally involved significant overhead from copying all the data in the InnoDB table. The fast index creation feature of the In- noDB Plugin makes both CREATE INDEX and DROP INDEX statements much faster for InnoDB secondary indexes. See also clustered index. See also fast index creation. See also index. selectivity A property of data distribution, the number of distinct values in a column divided by the number of records in the table. High selectivity means that the column values are relatively unique, and can retrieved efficiently through an index. If you (or the query optimizer) can predict that a test in a WHERE clause only matches a small number (or proportion) of rows in a table, the overall query tends to be efficient if it evaluates that test first, using an index. See also cardinality. semi-consistent read A type of read operation used for UPDATE statements, that is a combination of read com- mitted and consistent read. When an UPDATE statement examines a row that is already locked, InnoDB returns the latest committed version to MySQL so that MySQL can determ- ine whether the row matches the WHERE condition of the UPDATE. If the row matches (must be updated), MySQL reads the row again, and this time InnoDB either locks it or waits for a lock on it. This type of read operation can only happen when the transaction has the read committed isolation level, or when the innodb_locks_unsafe_for_binlog option is enabled. See also consistent read. See also isolation level. See also read committed. serializable read The isolation level that uses the most conservative locking strategy, to prevent any other transactions from inserting or changing data that was read by this transaction, until it is fin- ished. This way, the same query can be run over and over within a transaction, and be certain to retrieve the same set of results each time. Any attempt to change data that was committed by another transaction since the start of the current transaction, cause the current transaction to wait. This is the default isolation level specified by the SQL standard. In practice, this degree of strictness is rarely needed, so the default isolation level for InnoDB is the next most strict, repeatable read. See also ACID. See also consistent read. See also isolation level. See also locking. See also repeatable read. See also transaction. server A type of program that runs continuously, waiting to receive and act upon requests from an- other program (the client). Because often an entire computer is dedicated to running one or more server programs (such as a database server, a web server, an application server, or some combination of these), the term server can also refer to the computer that runs the serv- er software. See also client. 89
- 98. InnoDB Glossary See also mysqld. shared lock A kind of lock that allows other transactions to read the locked object, and to also acquire other shared locks on it, but not to write to it. The opposite of exclusive lock. See also exclusive lock. See also lock. See also transaction. shared tablespace Another way of referring to the system tablespace. See also system tablespace. shutdown The process of stopping the InnoDB storage engine. This process can do various cleanup op- erations, so it is slow to shut down but fast to start up later; or it can skip the cleanup opera- tions, so it is fast to shut down but must do the cleanup the next time it starts. The shutdown mode is controlled by the innodb_fast_shutdown option. See also fast shutdown. See also slow shutdown. slave database A slave database, frequently shortened to "slave", is a database in a replication scenario that receives changes from another database (the master database) and applies those same changes. Thus it maintains the same contents as the master, although it might lag somewhat behind. In MySQL, slave databases are commonly used in disaster recovery, to take the place of a master database that fails. They are also commonly used for testing software upgrades and new settings, to ensure that database configuration changes do not cause problems with per- formance or reliability. See also master database. See also replication. --sleep An option for the Hot Backup product that reduces the I/O overhead of the backup operation. (The backup takes longer as a result.) slow shutdown A type of shutdown that does additional flushing operations before completing. Specified by the configuration parameter innodb_fast_shutdown=0. Although the shutdown itself can take longer, that time will be saved on the subsequent startup. See also fast shutdown. See also shutdown. space ID See also tablespace. spin A type of wait operation that continuously tests whether a resource becomes available. This technique is used for resources that are typically held only for brief periods, where it is more efficient to wait in a “busy loop” than to put the thread to sleep and perform a context switch. If the resource does not become available within a short time, the spin loop ceases and anoth- er wait technique is used. See also latch. See also lock. See also mutex. See also wait. SQL The Structured Query Language that is standard for performing database operations. Often divided into the categories DDL, DML, and queries. See also DDL. 90
- 99. InnoDB Glossary See also DML. See also query. SSD Acronym for “solid-state drive”. A type of storage device with different performance charac- teristics than a traditional hard drive: smaller storage capacity, faster for random reads, no moving parts, and with a number of considerations affecting write performance. See also disk-based. statement-based replication This form of replication requires some care with the setting for the in- nodb_autoinc_lock_mode option, to avoid potential timing problems with auto-increment locking. See also auto-increment locking. See also innodb_autoinc_lock_mode. See also replication. See also row-based replication. storage engine A component of the MySQL database that performs the low-level work of storing, updating, and querying data. InnoDB is one such storage engine. Different storage engines are de- signed with different tradeoffs between factors such as memory usage versus disk usage, read speed versus write speed, and speed versus robustness. The MySQL Enterprise Backup product is optimized for backing up tables produced by the InnoDB storage engine. It can also back up tables produced by MyISAM and other stor- age engines. See also InnoDB. See also MySQL Enterprise Backup. stored procedure strict mode The general name for the setting controlled by the innodb_strict_mode option. Turn- ing on this setting causes certain conditions that are normally treated as warnings, to be con- sidered errors. For example, certain invalid combinations of options related to file format and row format, that normally produce a warning and continue with default values, now cause the CREATE TABLE operation to fail. MySQL also has something called strict mode. See also innodb_strict_mode. sublist Within the list structure that represents the buffer pool, pages that are relatively old and relat- ively new are represented by different portions of the list. A set of parameters control the size of these portions and the dividing point between the new and old pages. See also buffer pool. See also eviction. See also list. See also LRU. supremum record A pseudo-record in an index, representing the gap above the largest value in that index. If a transaction has a statement such as SELECT ... FOR UPDATE ... WHERE col > 10;, and the largest value in the column is 20, it is a lock on the supremum record that pre- vents other transactions from inserting even larger values such as 50, 100, and so on. See also gap. See also infimum record. See also pseudo-record. system tablespace By default, this single data file stores all the table data for a database, as well as all the 91
- 100. InnoDB Glossary metadata for InnoDB-related objects (the data dictionary). Turning on the innodb_file_per_table option causes each newly created table to be stored in its own tablespace, reducing the size of, and dependencies on, the system tablespace. Keeping all table data in the system tablespace has implications for the MySQL Enterprise Backup product (backing up one large file rather than several smaller files), and prevents you from using certain InnoDB features that require the newer Barracuda file format. on the See also Barracuda. See also data dictionary. See also file format. See also file-per-table. See also ibdata file. See also innodb_file_per_table. See also tablespace. T table Although a table is a distinct, addressable object in the context of SQL, for backup purposes we are often concerned with whether the table is part of the system tablespace, or was cre- ated under the file-per-table setting and so resides in its own tablespace. See also clustered index. See also file-per-table. See also system tablespace. See also table. table lock A lock that prevents any other transaction from accessing a table. InnoDB makes consider- able effort to make such locks unnecessary, by using techniques such as row locks and con- sistent reads for processing DML statements and queries. You can create such a lock through SQL using the LOCK TABLE statement; one of the steps in migrating from other database systems or MySQL storage engines is to remove such statements wherever practic- al. See also consistent read. See also DML. See also lock. See also locking. See also query. See also row lock. See also table. See also transaction. table monitor See also monitor. tablespace A data file that can hold data for one or more tables. The system tablespace contains the tables that make up the data dictionary, and by default holds all the other InnoDB tables. Turning on the innodb_file_per_table option allows newly created tables to each have their own tablespace, with a separate data file for each table. Tablespaces created by the built-in InnoDB storage engine are upward compatible with the 92
- 101. InnoDB Glossary InnoDB Plugin. Tablespaces created by the InnoDB Plugin are downward compatible with the built-in InnoDB storage engine, if they use the Antelope file format. See also Antelope. See also Barracuda. See also compressed row format. See also data dictionary. See also file-per-table. See also ibdata file. See also innodb_file_per_table. See also system tablespace. tablespace dictionary A representation of the data dictionary metadata for a table, within the InnoDB tablespace. This metadata can be checked against the .frm file for consistency when the table is opened, to diagnose errors resulting from out-of-date .frm files. This information is present for In- noDB tables that are part of the system tablespace, as well as for tables that have their own .ibd file because of the file-per-table option. See also data dictionary. See also file-per-table. See also .FRM file. See also ibd file. See also system tablespace. See also tablespace. tablespace monitor See also monitor. temporary table A table whose data does not need to be truly permanent. For example, temporary tables might be used as storage areas for intermediate results in complicated calculations or trans- formations; this intermediate data would not need to be recovered after a crash. Database products can take various shortcuts to improve the performance of operations on temporary tables, by being less scrupulous about writing data to disk and other measures to protect the data across restarts. Sometimes, the data itself is removed automatically at a set time, such as when the transac- tion ends or when the session ends. With some database products, the table itself is removed automatically too. See also table. thread See also master thread. See also Pthreads. torn page An error condition that can occur due to a combination of I/O device configuration and hard- ware failure. If data is written out in chunks smaller than the InnoDB page size (typically 16KB), a hardware failure while writing could result in only part of a page being stored to disk. The InnoDB doublewrite buffer guards against this possibility. See also doublewrite buffer. transaction Transactions are atomic units of work that can be committed or rolled back. When a transac- tion makes multiple changes to the database, either all the changes succeed when the transac- tion is committed, or all the changes are undone when the transaction is rolled back. Database transactions, as implemented by InnoDB, have properties that are collectively known by the acronym ACID, for atomicity, consistency, isolation, and durability. 93
- 102. InnoDB Glossary See also ACID. See also commit. See also isolation level. See also lock. See also rollback. transaction ID An internal field associated with each row. This field is physically changed by INSERT, UP- DATE, and DELETE operations to record which transaction has locked the row. See also implicit row lock. transportable tablespace A feature that allows a tablespace to be moved from one database to another. Traditionally, this has not been possible for InnoDB tablespaces because all table data was part of the sys- tem tablespace. Now, with the file-per-table setting, an InnoDB table can reside in its own .ibd file that you can copy individually. Also, the original .ibd file contains information such as the space ID that ties the tablespace to its original database. The transportable ta- blespace feature involves rewriting some information inside the tablespace to make it com- patible with the destination database. See also file-per-table. See also ibd file. See also space ID. See also system tablespace. See also tablespace. .TRG file A file containing trigger parameters. Files with this extension are always included in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also MySQL Enterprise Backup. See also .TRN file. trigger See also stored procedure. .TRN file A file containing trigger namespace information. Files with this extension are always in- cluded in backups produced by the innobackup command of the MySQL Enterprise Backup product. See also innobackup command. See also MySQL Enterprise Backup. See also .TRG file. troubleshooting Resources for troubleshooting InnoDB reliability and performance issues include: the In- formation Schema tables. truncate A DDL operation that removes the entire contents of a table. Although conceptually it has the same result as a DELETE statement with no WHERE clause, it operates differently behind the scenes: InnoDB creates a new empty table, drops the old table, then renames the new ta- ble to take the place of the old one. Because this is a DDL operation, it cannot be rolled back. If the table being truncated contains foreign keys that reference another table, the truncation operation uses a slower method of operation, deleting one row at a time so that correspond- ing rows in the referenced table can be deleted as needed by any ON DELETE CASCADE clause. (MySQL 5.5 and higher do not allow this slower form of truncate, and return an error instead if foreign keys are involved. In this case, use a DELETE statement instead. 94
- 103. InnoDB Glossary See also DDL. See also foreign key. See also rollback. tuple A technical term designating an ordered set of elements. It is an abstract notion, used in formal discussions of database theory. In the database field, tuples are usually represented by the columns of a table row. They could also be represented by the result sets of queries, for example, queries that retrieved only some columns of a table, or columns from joined tables. See also cursor. two-phase commit An operation that is part of a distributed transaction, under the XA specification. (Sometimes abbreviated as 2PC.) When multiple databases participate in the transaction, either all databases commit the changes, or all databases roll back the changes. See also commit. See also rollback. See also transaction. See also XA. two-phase locking See also XA. U undo Data that is maintained throughout the life of a transaction, recording all changes so that they can be undone in case of a rollback operation. See also rollback. See also transaction. See also undo log. undo buffer See undo log. undo log A storage area that holds copies of data modified by active transactions. If another transaction needs to see the original data (as part of a consistent read operation), the unmodified data is retrieved from this storage area. This area is physically part of the system tablespace. It is split into separate portions, the insert undo buffer and the update undo buffer. Collectively, these parts are also known as the roll- back segment, a familiar term for Oracle DBAs. See also consistent read. See also rollback segment. See also system tablespace. See also transaction. unique constraint A kind of constraint that asserts that a column can not contain any duplicate values. In terms of rela- tional algebra, it is used to specify 1-to-1 relationships. For efficiency in checking whether a value can be inserted, (that is, the value does not already exist in the column), a unique constraint is supported by an underlying unique index. See also constraint. See also relational. See also unique index. unique index An index on a column or set of columns that have a unique constraint. Because the index is known not to contain any duplicate values, certain kinds of lookups and count operations are more efficient than in 95
- 104. InnoDB Glossary the normal kind of index. Most of the lookups against this type of index are simply to determine if a certain value exists or not. The number of values in the index is the same as the number of rows in the table, or at least the number of rows with non-null values for the associated columns. See also cardinality. See also unique constraint. See also unique key. unique key The set of columns (one or more) comprising a unique index. When you can define a WHERE condition that matches exactly one row, and the query can use an associated unique index, the lookup and error handling can be performed very efficiently. See also cardinality. See also unique constraint. See also unique index. V victim The transaction that is automatically chosen to be rolled back when a deadlock is detected. InnoDB rolls back the trans- action that has updated the fewest rows. See also deadlock. See also deadlock detection. See also innodb_lock_wait_timeout. W wait When an operation, such as acquiring a lock, mutex, or latch, cannot be completed immediately, InnoDB pauses and tries again. The mechanism for pausing is elaborate enough that this operation has its own name, the wait. Individual threads are paused using a combination of internal InnoDB scheduling, operat- ing system wait() calls, and short-duration spin loops. On systems with heavy load and many transactions, you might use the output from the SHOW INNODB STATUS command to determine whether threads are spending too much time waiting, and if so, how you can improve concurrency. See also concurrency. See also latch. See also lock. See also mutex. See also spin. warm backup A backup taken while the database is running, but that restricts some database operations during the backup process. For example, tables might become read-only. For busy applications and web sites, you might prefer a hot backup. See also cold backup. See also hot backup. warm up To run a system under a typical workload for some time after startup, so that the buffer pool and other memory regions are filled as they would be under normal conditions. Typically done during performance testing, to ensure consistent results across multiple runs; otherwise, performance might be artificially low during the first run. See also buffer pool. 96
- 105. InnoDB Glossary Windows The built-in InnoDB storage engine and the InnoDB Plugin are supported on all the same Microsoft Win- dows versions as MySQL. The MySQL Enterprise Backup product is available on Windows, although the innobackup com- mand is not part of the Windows edition of the product. workload The combination and volume of SQL and other database operations, performed by a database application during typical or peak usage. You can subject the database to a particular workload during performance testing to identify bottlenecks, or during capacity planning. See also bottleneck. See also SQL. write combining An optimization technique that reduces write operations when dirty pages are flushed from the InnoDB buffer pool. If a row in a page is updated multiple times, or multiple rows on the same page are updated, all of those changes are stored to the data files in a single write operation rather than one write for each change. See also buffer pool. See also dirty page. See also flush. X X A standard interface for coordinating distributed transactions, allowing multiple databases to participate in a transaction A while maintaining ACID compliance. XA Distributed Transaction support is turned on by default. If you are not using this feature, note that it adds an extra fsync for each transaction, which may adversely affect performance. See also commit. See also transaction. See also two-phase commit. See also two-phase locking. 97
- 106. compressed row format, ndex compression, algorithms, 10 application and schema design, 9 Symbols BLOBs, VARCHAR and TEXT, 11 buffer pool, 12 --sleep, data and indexes, 11 .ARM file, data characteristics, 8 .ARZ file, enabling for a table, 6 .FRM file, implementation, 10 .MRG file, log files, 12 .MYD files, modification log, 11 .MYI files, overview, 6 .OPT file, tuning, 8 .PAR file, Compression .TRG file, compressed page size, 10 .TRN file, configuration characteristics, 9 information schema, 18, 18 A innodb_strict_mode, 36 ACID, KEY_BLOCK_SIZE, 10 adaptive flushing, monitoring, 10 adaptive hash index, 27, 36, overflow pages, 11 AIO, workload characteristics, 9 ALTER TABLE concatenated index, ROW_FORMAT, 17 concurrency, Antelope, configuration file, Antelope file format, 13 consistent read, application programming interface (API), constraint, apply, covering index, asynchronous I/O, 29, crash recovery, auto-increment, CREATE INDEX, 3 auto-increment locking, CREATE TABLE autocommit, KEY_BLOCK_SIZE, 10 options for table compression, 6 B ROW_FORMAT, 17 cursor, B-tree, background threads master, 29, 30 D read, 28 data definition language, write, 28 data dictionary, backticks, data files, Barracuda, data manipulation language, Barracuda file format, 13 data warehouse, beta, database, binary log, DDL, binlog, deadlock, bottleneck, deadlock detection, buffer, delete, buffer cache, 31 delete buffering, buffer pool, denormalized, built-in, descending index, business rules, dirty page, dirty read, C disk-based, DML, cache, doublewrite buffer, cardinality, downgrading, 42 change buffering, 44, DROP INDEX, 3 disabling, 26 dynamic row format, checkpoint, child table, client, E clustered index, early adopter, cold backup, Embedded InnoDB, column index, eviction, column prefix, exclusive lock, commit, explicit row lock, compact row format, composite index, compressed backup, F 98
- 107. Index Fast Index Creation innobackup command, concurrency, 4 InnoDB, crash recovery, 4 troubleshooting examples, 3 fast index creation, 4 implementation, 3 InnoDB Hot Backup, limitations, 4 InnoDB parameters, deprecated, 49 overview, 3 innodb_file_io_threads, 28 fast index creation, InnoDB parameters, new, 48 fast shutdown, innodb_adaptive_flushing, 30 file format, 13, innodb_change_buffering, 26 Antelope, 11 innodb_file_format, 35 Barracuda, 6 innodb_file_format_check, 14 downgrading, 16 innodb_io_capacity, 29 identifying, 16 innodb_read_ahead_threshold, 28 file format management innodb_read_io_threads, 28 downgrading, 42 innodb_spin_wait_delay, 30 enabling new file formats, 35 innodb_stats_sample_pages, 37 file per table, 35 innodb_strict_mode, 36 file-per-table, innodb_use_sys_malloc, 25 fixed row format, innodb_write_io_threads, 28 flush, InnoDB parameters, with new defaults, 50 flush list, innodb_additional_mem_pool_size, 50 flush list mutex, 34 innodb_buffer_pool_size, 50 foreign key, innodb_change_buffering, 50 foreign key constraint, innodb_file_format_check, 50 FOREIGN KEY constraints innodb_log_buffer_size, 50 and fast index creation, 4 innodb_max_dirty_pages_pct, 30, 50 and TRUNCATE TABLE, 36 innodb_sync_spin_loops, 50 full table scan, innodb_thread_concurrency, 50 InnoDB storage engine G compatibility, 1 downloading, 1 GA, features, 1 gap, installing, 1 group commit, 29, restrictions, 2 innodb_adaptive_flushing, 30 H innodb_adaptive_hash_index, 27 hash_index, and innodb_thread_concurrency, 27 heartbeat, dynamically changing, 36 hot, innodb_additional_mem_pool_size hot backup, and innodb_use_sys_malloc, 25 innodb_autoinc_lock_mode, innodb_change_buffering, 26 I innodb_concurrency_tickets, 27 ib-file set, 13, innodb_file_format, 13, ibbackup command, Antelope, 11 ibbackup_logfile, Barracuda, 6 ibd file, downgrading, 42 ibdata file, enabling new file formats, 35 ibz file, identifying, 16 ib_logfile, innodb_file_format_check, 14 implicit row lock, innodb_file_io_threads, 28 in-memory database, innodb_file_per_table, 6, incremental backup, dynamically changing, 35 index, innodb_io_capacity, 29 Index dives (for statistics estimation), 37 innodb_locks_unsafe_for_binlog, index prefix, innodb_lock_wait_timeout, Indexes dynamically changing, 36 creating and dropping, 3 innodb_max_dirty_pages_pct, 30 primary (clustered) and secondary, 3 innodb_old_blocks_pct, 31 infimum record, innodb_old_blocks_time, 31 information schema, innodb_read_ahead_threshold, 28 information schema tables, 18 innodb_read_io_threads, 28 INNODB_CMP, 18 innodb_spin_wait_delay, 30 INNODB_CMPMEM, 18 innodb_stats_on_metadata INNODB_CMPMEM_RESET, 18 dynamically changing, 36 INNODB_CMP_RESET, 18 innodb_stats_sample_pages, 37 INNODB_LOCKS, 19 innodb_strict_mode, 36, INNODB_LOCK_WAITS, 19 innodb_table_locks, INNODB_TRX, 19 innodb_thread_concurrency, 27 99
- 108. Index innodb_thread_sleep_delay, 27 non-repeatable read, innodb_use_sys_malloc, 25 normalized, and innodb_thread_concurrency, 27 NULL, innodb_write_io_threads, 28 insert, insert buffer, O off-page column, insert buffering, 44, OLTP, disabling, 26 optimistic, insert-intention gap lock, Optimizer Statistics Estimation, 36, 37 intention exclusive lock, option, intention lock, option file, intention shared lock, overflow page, internal memory allocator disabling, 25 isolation level, P page, J page size, parent table, join, partial index, Performance Schema, 32, K pessimistic, KEY_BLOCK_SIZE, 6, 10 phantom, key_block_size, physical backup, PITR backup, L plugin, point-in-time recovery, latch, predicate locking, list, prefix, lock, prepared backup, lock escalation, primary key, lock mode, PROCESSLIST lock monitor, possible inconsistency with INFORMATION_SCHEMA tables, 24 lock wait timeout, 36 pseudo-record, Locking Pthreads, information schema, 18, 19, 23 purge, locking, purge buffering, log buf mutex, 33 purge scheduling, 33 log sys mutex, 33 purge thread, log write, logical backup, loose_, Q LRU, query, LRU page replacement, 31 LSN, R RAID, M random dive, master database, raw backup, master thread, read ahead, 38 memory allocator read committed, innodb_use_sys_malloc, 25 read uncommitted, mid-point insertion strategy, read-ahead, midpoint insertion, 31 linear, 28 mini-transaction, random, 28 monitor, record-only lock, multi-core, redo log, multiple buffer pools, 33 redundant row format, multiple rollback segments, 33 referential integrity, multiversion concurrency control, relational, mutex, repeatable read, MVCC, replication, my.cnf, restore, my.ini, rollback, mysql, rollback segment, MySQL Enterprise Backup, row format, mysqld, row lock, mysqldump, row-based replication, ROW_FORMAT N COMPACT, 17 COMPRESSED, 6, 17 non-blocking I/O, DYNAMIC, 17 100
- 109. Index REDUNDANT, 17 victim, rw-lock, W S wait, savepoint, warm backup, scalability, warm up, schema, Windows, secondary index, workload, selectivity, write combining, semi-consistent read, serializable read, server, X XA, shared lock, shared tablespace, SHOW ENGINE INNODB MUTEX, 38 SHOW ENGINE INNODB STATUS and innodb_adaptive_hash_index, 27 and innodb_use_sys_malloc, 25 shutdown, slave database, slow shutdown, space ID, spin, SQL, SSD, statement-based replication, storage engine, stored procedure, strict mode, 36, sublist, supremum record, system tablespace, T table, table lock, table monitor, table scan, 31 tablespace, tablespace dictionary, tablespace monitor, temporary table, thread, torn page, transaction, transaction ID, transportable tablespace, trigger, troubleshooting, truncate, TRUNCATE TABLE, 36 tuple, two-phase commit, two-phase locking, U undo, undo buffer, undo log, unique constraint, unique index, unique key, upgrading converting compressed tables, 41 dynamic plugin, 41 static plugin, 41 V 101