MythBusters Globalization Support - Avoid Data Corruption

MythBusters Globalization Support
Avoid Data Corruption
Christian Gohmann
@CGohmannDE DOAG2018

About me…
MythBusters Globalization Support - Avoid Data Corruption15.09.2019
Christian Gohmann
Senior Consultant at Trivadis GmbH, Düsseldorf (Germany)
Instructor since 2014
– O-AI (Oracle Architecture and Internals)
– O-AI-DBA (Oracle Architecture and Internals for DBAs)
Tool Owner of TVD-Backup
Experiences with Oracle since 2006 (8i – 18c)
– Architecture, Installation & Configuration
– High Availability Solutions (RAC, Data Guard)
– Migration Projects
– Backup & Recovery (RMAN, Data Pump)
Blog: http://www.christian-gohmann.de

We help to generate added value from data

With over 650 specialists and IT experts in your region.
16 Trivadis branches and more than
650 employees
Experience from more than 1,900
projects per year at over 800
customers
250 Service Level Agreements
Over 4,000 training participants
Research and development budget:
CHF 5.0 million
Financially self-supporting and
sustainably profitable

Agenda
1. Fundamentals
2. Configuration
3. Conversion
4. Tools
5. Debugging
6. Migration

Fundamentals

Globalization Support
Store, process and retrieve data in native languages
– Allows the development of multilingual applications
Former known as National Language Support (NLS)
– Subset of the Globalization Support
Implemented with the Oracle NLS Runtime Library (NLSRTL)
– Functions for text and character processing and manipulations
Locale-specific data is stored in $ORACLE_HOME/nls/data
– Use ORA_NLS10 environment variable to choose a different location
– Loaded as independent set at runtime

Language Support
Refers to the additional language-dependent functionalities
– Display dates using translated month names
– Sort text according to cultural conventions
– Translated error messages and user interfaces
Only available for a subset of the languages
Parameter: NLS_LANGUAGE
“Language Support means the ability to store text of a specific language”
SQL> ALTER SESSION SET nls_language = 'FRENCH';
BUSTED

Territory Support
Has nothing to do with the ability to store characters in the database
Support of cultural conventions, specific to geographical locations
– Default local time format
– Date format
– Numeric and monetary conventions
Parameter: NLS_TERRITORY
“Territory support allows the database to store characters of different languages”
SQL> ALTER SESSION SET nls_territory = 'FRANCE';
BUSTED

Character Set Support 1/2
"A Character Set maps a number to a characters visual representation."
Two character sets are defined at database creation
National Character Set to store Unicode
– Limited to NCHAR, NVARCHAR2 and NCLOB data types
Oracle naming convention (except UTF8 and UTFE)
– <Region><Bit length><Standard Character Set Name>[S|C]
– Optional: S → Server, C → Client
– Example: WE8ISO8859P15
SQL> CREATE DATABASE TEST (
…
CHARACTER SET WE8MSWIN2152
NATIONAL CHARACTER SET AL16UTF16
);

Character Set Support 2/2
Windows character sets like WE8MSWIN1252 can be used on Unix
– EBCDIC character sets cannot be used on ASCII platforms and vice-versa
"Windows character sets are limited to the Windows platform"
BUSTED

Calendar Systems
Support of eight different calendar systems
Switch to a different calendar
NLS Calendar Utility can be used to register calendar modifications (*.nlt files)
Name Name
Gregorian Japanese Imperial
ROC Official (Republic of China) Thai Buddha
Persian English Hijrah
Arabic Hijrah Ethiopian
SQL> ALTER SESSION SET nls_calendar = 'Persian';
$> lxegen

Unicode Support
Support of the Unicode standard since Oracle 7
Two ways to store Unicode characters in the database
– Create a database with a Unicode character set (e.g. UTF8)
– Support multilingual data in specific columns (NCHAR, NVARCHAR, NCLOB)
Oracle recommends the usage of AL32UTF8 (default since Oracle 12c Release 2)
Database Migration Assistant for Unicode (DMU)
Oracle Version Unicode Version
10g 4.0
11g 5.0
12c 6.2 / 7.0
18c 9.0

Enhancements in 12c and 18c
Oracle 12c
Support for Unicode 6.2 / 7.0 (12c Release 2)
Support for 12 new languages and 32 new territories
Support for the Unicode Collation Algorithm (UCA)
Support for Pluggable Databases with different Character Sets
Column Level Collation (Case-insensitive Database)
Desupport of CSSCAN and CSALTER for migration to Unicode
– Instead use Database Migration Assistant for Unicode (DMU)
Oracle 18c
Support for Unicode 9.0
Support for 3 new languages and 31 new territories

Configuration

General
A new session initializes the NLS environment from the NLS instance parameters
– If specific client-side NLS settings are given, an ALTER SESSION is executed
Check V$NLS_VALID_VALUES for valid values of NLS settings
Direct connection from the server acts like a client connection
Set NLS_% parameter on instance level requires restart of instance
SQL> ALTER SYSTEM SET nls_date_format = 'YYYY-MM-DD HH24:MI' SCOPE=SPFILE;

Client 1/3
NLS_LANG only represents the character set that the client use
– Example: German Windows client → GERMAN_GERMANY.WE8MSWIN1252
– Example: Unix client with Unicode locale → AMERICAN_AMERICA.UTF8
It can be the same, but it is not a must
It is required for proper conversion
If client and server character sets match, then no validation is enforced
“NLS_LANG must match the character set of the database server”
BUSTED

Client 2/3
If not set it defaults to the hardcoded value US7ASCII
Registry key: HKLMSOFTWAREORACLEKEY_<Oracle Home Name>
“If NLS_LANG is not set it uses the NLS_LANG of the database server”
SQL> ! echo $NLS_LANG
SQL> SELECT client_charset, network_service_banner
FROM v$session_connect_info
WHERE sid = SYS_CONTEXT('USERENV', 'SID');
CLIENT_CHARSET NETWORK_SERVICE_BANNER
-------------- ------------------------------------------------------------------------------
US7ASCII Oracle Bequeath NT Protocol Adapter for Linux: Version 12.1.0.1.0 - Production
US7ASCII Authentication service for Linux: Version 12.1.0.1.0 - Production
US7ASCII Encryption service for Linux: Version 12.1.0.1.0 - Production
US7ASCII Crypto-checksumming service for Linux: Version 12.1.0.1.0 - Production
BUSTED

Client 3/3
Setting NLS_LANG changes nothing on the client
The character set can only be changed with operating system tools
“Setting NLS_LANG changes the character set of the client”
# Change Locale on Linux
$> export LANG=de_DE.CP1252
$> locale
LANG=de_DE.CP1252
LC_CTYPE="de_DE.CP1252"
…
LC_IDENTIFICATION="de_DE.CP1252"
BUSTED

Miscellaneous
Different tools or input files require specific NLS_LANG settings
Tool NLS_LANG Comment
SQL Loader /
SQL Developer
Character set of the text/flat file
UTL_FILE Does not use it
Runs in the database. Assume encoding of the input
file is in the character set of the database. Otherwise
use UTL_RAW.CONVERT
Export Same as the database character set No character conversion
Import Character set of the source database Check with imp .. show=yes
Data Pump
Not affected by the NLS_LANG setting.
Only used for the parameter file.
SQLcl Does not evaluate NLS_LANG.
Because SQLcl is a Java program, it evaluates
environment variable LANG (except Windows)

Conversion

Character Conversion
Normally conversion is done at client side (OCI driver)
– 8.0.4 onwards
If character set is not known at client side, server side will do conversion
– 8.1.6 onwards
If server and client side character sets are the same, no validation will be
performed
– Increased performance
– Chance to write data garbage

Character Lifecycle
Database
WEISO8859P15
Client / Tool
WE8MSWIN1252
Codepage 1252
Insert
"€"
Translate
"€" → 128 (20AC)
Convert codepage
128 → 164
Send codepage
164
User

Special Characters
Some character sets offers mappings from unknown characters to known
characters
– E.g. é → e, ä → a
Character Description
□ The used font has no graphical representation for the codepage
¿ Replacement character that is used for unknown characters in the character set

Tools

SQL*Plus 1/2
SQL*Plus in a DOS box does not use the regular Windows codepage
– Instead it uses the OEM environment (e.g. codepage 850)
– Character set part of NLS_LANG must be set to WE8PC850
– Alternatively change codepage using chcp and set different font
“The codepage of running SQL*Plus in a DOS box differs from the regular operating system
codepage"
$> chcp
Active Codepage: 850
$> chcp 1252
CONFIRMED

SQL*Plus 2/2
Until Oracle 10g an ANSI version of SQL*Plus (sqlplusw.exe) exists
– Uses the configured Windows codepage (e.g. 1252)
– Officially replaced by the SQL Developer
On Windows only: SQL*Plus is not UTF8/Unicode aware

SQL Developer 1/2
SQL Developer needs no client side NLS configuration
– Unicode client
Download from Oracle Technology Network (OTN)
– http://www.oracle.com/technetwork/developer-tools/sql-developer
Change font if □ characters appear in the output
– Reason: The used font cannot display the character
– Tools > Preferences > Code-Editor > Fonts

SQL Developer 2/2
Change specific NLS settings of the SQL Developer session
– Tools > Preferences > Database > NLS
– Changes will be applied immediately

Telnet/SSH Client
Responsible for conversion of the Unix locale to the client's environment
– Translation settings must match the encoding of the Unix Shell
Example: Connection to an en_EN.ISO8859-15 Unix Shell with PuTTY

Locale Builder 1/2
View, modify or create Language, Territory, Character Set or Linguistic Sort
definitions
Available since Oracle 9i
– Windows: %ORACLE_HOME%nlslbuilderlbuilder.exe
– Unix: $ORACLE_HOME/nls/lbuilder/lbuilder
Open NLS locale data definitions using file or object name

Locale Builder 2/2

Debugging

General 1/2
Check data with a Unicode client like SQL Developer
Check NLS settings on different levels
Client character set is visible in the V$SESSION_CONNECT_INFO view
– Available since 11g Release 1
SQL> SELECT * FROM nls_database_parameters;
SQL> SELECT * FROM nls_instance_parameters;
SQL> SELECT * FROM nls_session_parameters; -- Current session
SQL> SELECT client_charset, network_service_banner
FROM v$session_connect_info
WHERE sid = SYS_CONTEXT('USERENV', 'SID');

General 2/2
Get NLS_LANG value set in Registry only
SQL> host echo %NLS_LANG%
%NLS_LANG%
SQL> @.[%NLS_LANG%].
SP2-0310: Unable to open file ".[GERMAN_GERMANY.WE8MSWIN1252]..sql"

Character Set Health Check
Use CSSCAN utility to search for incompatible characters (< 12c)
– Requires a schema called CSMIG to store internal data
Run CSSCAN as SYSDBA
– Set FROMCHAR and TOCHAR to the same value
Check cs_healthceck.txt for "Convertible" or "Lossy" entries
– If no entries exist, the current data is fine
SQL> @?/rdbms/admin/csminst.sql
$> csscan FULL=Y FROMCHAR=WE8ISO8859P1 TOCHAR=WE8ISO8859P1
ARRAY=1024000 LOG=cs_healthcheck CAPTURE=Y PROCESS=4

DUMP Function
Returns datatype code, length (bytes) and internal representation
Supports CHAR, VARCHAR2, NCHAR, NVARCHAR2, DATE, NUMBER
– Cannot be used on CLOB or NCLOB columns
Useful to check if a specific character is supported
– Requires the usage of an Unicode client like SQL Developer
– If an insert in a table succeeded the Character Set supports the character
SQL> SELECT DUMP(col1, 10) FROM tab1;
SQL> SELECT DUMP('Text with € sign', 10, 1, 10) FROM dual;

Migration

Superset vs. Subset
Character set can be a Superset or a Subset of the new character set
Type Description
Superset All characters from character set A are included in character set B.
Binary Superset
Same as Superset, but the binary representations of the character are the
same in character set A and B
Subset
Only a subset of the characters in character set A are included in character
set B.
Binary Subset
Same as Subset, but the binary representations of the character are the
same in character set A and B

Database Migration Assistant for Unicode (DMU) 1/2
Migrates databases from legacy character sets to Unicode
– A GUI guides through all steps of the migration process
Validation mode for existing Unicode databases
– Searches for not correctly encoded characters
Version 2.2 was released in November 2017
– Support for JDK 8
– New command-line scanner (DMU-CLS) for scanning databases and generating reports
Since version 2.1.1 only database versions 11.2.0.4 and higher are supported
– For older database versions, you have to use version 2.1.0 (Patch 21138450)
Only supported tool for character set migration in 12c and higher
– Replaces CSSCAN and CSALTER
– Can only migrate databases to Unicode

Database Migration Assistant for Unicode (DMU) 2/2
Download from Oracle Technology Network or My Oracle Support
– OTN version: OTN Developer License (Evaluation)
http://www.oracle.com/technetwork/database/database-
technologies/globalization/dmu
– MOS version: Program Update under the Database Support Contract (Migration) -
Patch #27084436
Requires the installation of the Java Development Kit 8

Requirements
Oracle Database must be 10.2.0.4 (11.2.0.4) or higher
– Installed XML Database feature
– DBMS_DUMA_INTERNAL package
ASCII-based character set
Oracle Database Vault must be disabled
Data Dictionary must contain ASCII characters only
No Flashback archives can exist
Data that requires conversions …
– … cannot reside in read-only or offline tablespaces
– … can be present in tables in the recycle bin
– … can be present in reference partitioning key column
SQL> @?/rdbms/admin/prvtdumi.plb

Repository Configuration Wizard
MythBusters Globalization Support - Avoid Data Corruption15.09.201915.09.2019 MythBusters Globalization Support - Avoid Data Corruption

Migration
MythBusters Globalization Support - Avoid Data Corruption15.09.201915.09.2019 MythBusters Globalization Support - Avoid Data Corruption

Database Globalization Support Guide 12c Release 2
https://docs.oracle.com/en/database/oracle/oracle-database/12.2/nlspg/index.html
Database Globalization Support Guide 18c
https://docs.oracle.com/en/database/oracle/oracle-database/18/nlspg/index.html
My Oracle Support Notes - https://support.oracle.com
264157.1 The Correct NLS_LANG Setting in Unix Environments
179133.1 The Correct NLS_LANG in a Microsoft Windows Environment
179133.1 NLS_LANG Explained (How does Client-Server Character Conversion Work?)
241047.1 The Priority of NLS Parameters Explained (Where To Define NLS Parameters)

Christian Gohmann
Senior Consultant
Tel. +49-211-58 6664 702
christian.gohmann@trivadis.com
15.09.2019 MythBusters Globalization Support - Avoid Data Corruption

Trivadis @ DOAG 2018
#opencompany
Booth: 3rd Floor – next to the escalator
We share our Know how!
Just come across, Live-Presentations
and documents archive
T-Shirts, Contest and much more
We look forward to your visit
15.09.2019 MythBusters Globalization Support - Avoid Data Corruption

MythBusters Globalization Support - Avoid Data Corruption

More Related Content

What's hot (20)

Similar to MythBusters Globalization Support - Avoid Data Corruption (15)

Recently uploaded (20)

MythBusters Globalization Support - Avoid Data Corruption