SlideShare a Scribd company logo

Mining AWR V2 - Trend Analysis

Maris Elsins, profile picture
Maris Elsins

The document presents an analysis of AWR (Automatic Workload Repository) reports and trends, emphasizing their utility in monitoring database performance through snapshots that capture various metrics over time. It discusses the limitations of single data points ('dots') in understanding performance trends and explores methods for visualizing and interpreting statistical data to uncover insights about historical database behavior. Additionally, the author, Maris Elsins, provides a methodology for extracting and calculating deltas from AWR data to inform performance analysis.

Technology
Related topics:
Trend Analysis
1 of 84
Downloaded 88 times
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
© 2016 Pythian 1 Mining AWR V2 Trend Analysis MARIS ELSINS Lead Database Consultant Can you read the line below? Can you read the line below? This is the font size I use for listing code in slides!
© 2016 Pythian 2 Maris Elsins Lead Database Consultant At Pythian since 2011 Located in Riga, Latvia Oracle [Apps] DBA since 2005 Speaker at conferences since 2007 @MarisDBA elsins@pythian.com http://bit.ly/getMOSPatchV2
© 2016 Pythian 3 ABOUT PYTHIAN Pythian’s 400+ IT professionals help companies adopt and manage disruptive technologies to better compete
© 2016 Pythian 4 Systems currently managed by Pythian EXPERIENCED Pythian experts in 35 countries GLOBAL Millennia of experience gathered and shared over 19 years EXPERTS 11,800 2400
© 2016 Pythian 5 • “1-Dot” Philosophy • • “2-Dots” Philosophy • • • Data trends in AWR • Examples AGENDA
© 2016 Pythian 6 LET'S OVERSIMPLIFY WHAT IS AN AWR REPORT? It's a dot ! •
© 2016 Pythian 7 LET'S OVERSIMPLIFY SERIOUSLY? A DOT??? Are you gone totally nuts? Photo by herr_hartmann / CC BY-NC 2.0
© 2016 Pythian 8 • AWR Report gives you valuable drill-down capabilities (time interval): – Top Wait Events – Top SQL Statements in different categories – Top Segments by Logical Reads – ... LET'S OVERSIMPLIFY AWR REPORT AS A "DOT"! THINK ABOUT IT!
© 2016 Pythian 9 DOTS AND AWR - LET'S INVESTIGATE! HOW DID THE IO PERFORM ON SEPTEMBER 9? - AWR REPORT IS THE ANSWER!
© 2016 Pythian 10 LET'S OVERSIMPLIFY AWR REPORT AS A "DOT"! THINK ABOUT IT! • Lot of resource usage / performance information is aggregated down to a single number (a "dot" on a time/value graph) – How much time was spent waiting on User I/O? = 42189 s – How quickly did the db file sequential reads respond? = 0.84 ms – How many transactions per second are completed? = 47.09
© 2016 Pythian 11 LET'S OVERSIMPLIFY DB FILE SEQUENTIAL READS PERFORMANCE AS A "DOT" Time Resp. time 09-Sep-16 0.84 ms
© 2016 Pythian 12 "The FBI Academy teaches new agents that the best predictor of future behavior is past behavior." Ronald Kessler Who doesn't want to be an FBI agent specializing in database performance analysis? A “dot” doesn't reveal the past behaviour WHY THIS IS IMPORTANT?
© 2016 Pythian 13 1-DOT What does it tell us about how things develop over time? Time Value S1 V1
© 2016 Pythian 14 DOTS AND AWR COMPARING TWO PERIODS OF TIME (AWR DIFF REPORT)
© 2016 Pythian 15 What does it tell us about how things develop over time? 2-DOTS Time Value S1 V1 V2 S2
© 2016 Pythian 16 2-DOTS What does it tell us about how things develop over time? Time Value S1 V1 V2 S2
© 2016 Pythian 17 DOTS AND AWR • An AWR report describes the time interval between 2 snapshots i.e. – How much time was spent on CPU? – How many buffer gets did a particular query do? – How many Physical IOs did a top query do? – ... etc ... • A single aggregated value, that describes the reporting interval, which is insufficient for trend analysis
© 2016 Pythian 18 IS IT WORTH MINING AWR? Trends in AWR
© 2016 Pythian 19 MINING DATA TRENDS IN AWR • Each AWR snapshot represents a moment in time. • Multitude of statistics are saved for each snapshot • It's an obvious thing to graph these values! • And the best about it is you pick the metric to graph!
© 2016 Pythian 20 SQL> @awr_wait_trend.sql "db file sequential read" 10 1 TIME (SNAP) EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- .... 09.09.2016 00:00:00 db file sequential read 55302 50.670 .916 09.09.2016 01:00:00 db file sequential read 45372 31.303 .690 09.09.2016 02:00:00 db file sequential read 80171 68.995 .861 09.09.2016 03:00:00 db file sequential read 215704 235.555 1.092 09.09.2016 04:00:00 db file sequential read 67104 46.547 .694 09.09.2016 05:00:00 db file sequential read 172241 104.396 .606 09.09.2016 06:00:00 db file sequential read 65497 36.598 .559 09.09.2016 07:00:00 db file sequential read 76487 46.231 .604 09.09.2016 08:00:00 db file sequential read 53193 27.399 .515 09.09.2016 09:00:00 db file sequential read 78568 45.846 .584 09.09.2016 10:00:00 db file sequential read 88612 48.192 .544 09.09.2016 11:00:00 db file sequential read 71483 42.597 .596 09.09.2016 12:00:00 db file sequential read 79741 53.432 .670 09.09.2016 13:00:00 db file sequential read 89707 65.399 .729 09.09.2016 14:00:00 db file sequential read 85208 49.717 .583 09.09.2016 15:00:00 db file sequential read 125878 99.274 .789 09.09.2016 16:00:00 db file sequential read 71224 39.754 .558 09.09.2016 17:00:00 db file sequential read 78324 46.832 .598 09.09.2016 18:00:00 db file sequential read 72536 67.021 .924 09.09.2016 19:00:00 db file sequential read 49199 37.167 .755 09.09.2016 20:00:00 db file sequential read 47840 26.727 .559 ... DOTS AND AWR HOW DID THE IO PERFORM ON SEPTEMBER 9?
© 2016 Pythian 21 DOTS AND AWR HOW DID THE IO PERFORM ON SEPTEMBER 9?
© 2016 Pythian 22 DOTS AND AWR DO YOU STILL REMEMBER THIS PICTURE?
© 2016 Pythian 23 DOTS AND AWR HOW DID THE IO PERFORM BETWEEN SEPTEMBER 2 AND 9?
© 2016 Pythian 24 DOTS AND AWR TOTAL WAITS COMPARISON BETWEEN SEPTEMBER 2 AND 9?
© 2016 Pythian 25 DOTS AND AWR RESPONSE TIME COMPARISON BETWEEN SEPTEMBER 2 AND 9?
© 2016 Pythian 26 HOW DID I DO IT? Extracting the “Past Behaviour” from AWR
© 2016 Pythian 27 Information on the DBA_HIST_views can be found in Oracle® Database Reference 12c Release 2 (12.2) http://docs.oracle.com/database/122/REFRN/toc.htm SELECT v.snap_id, v.value FROM dba_hist_sysstat v WHERE v.stat_name = 'user commits' ORDER BY v.snap_id; SNAP_ID VALUE ---------- ---------- 12621 625766 12622 626608 12623 627853 12624 629348 12625 630130 12626 630904 .... HOW DID I DO IT? STEP 1: PICK THE METRIC AND FIND THE CORRECT DBA_HIST_% VIEW
© 2016 Pythian 28 SELECT Lag(snap_id) over (ORDER BY v.snap_id) from_snap, v.snap_id to_snap, v.value - Lag(v.value) over (ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v WHERE v.stat_name = 'user commits' ORDER BY v.snap_id; FROM_SNAP TO_SNAP DELTA_VALUE ---------- ---------- ----------- 12621 12621 12622 842 12622 12623 1245 12623 12624 1495 12624 12625 782 12625 12626 774 ... HOW DID I DO IT? STEP 2: CALCULATE DELTAS
© 2016 Pythian 29 SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id ORDER BY s.end_interval_time; FROM_TIME TO_TIME DELTA_VALUE ------------------------------- ------------------------------- ----------- 09-SEP-16 04.00.44.539000000 PM 09-SEP-16 04.00.44.539000000 PM 09-SEP-16 05.00.58.620000000 PM 842 09-SEP-16 05.00.58.620000000 PM 09-SEP-16 06.00.22.386000000 PM 1245 09-SEP-16 06.00.22.386000000 PM 09-SEP-16 07.00.35.070000000 PM 1495 09-SEP-16 07.00.35.070000000 PM 09-SEP-16 08.00.45.075000000 PM 782 09-SEP-16 08.00.45.075000000 PM 09-SEP-16 09.00.54.573000000 PM 774 09-SEP-16 09.00.54.573000000 PM 09-SEP-16 10.00.04.401000000 PM 768 ... HOW DID I DO IT? STEP 3: VIEWER FRIENDLY DATES
© 2016 Pythian 30 HOW DID I DO IT? STEP 4: TIMELINE FILTERING SELECT * FROM (SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id) WHERE from_time BETWEEN To_date('11092016', 'DDMMYYYY') AND To_date('16092016', 'DDMMYYYY') ORDER BY from_time; FROM_TIME TO_TIME DELTA_VALUE ------------------------------- ------------------------------- ----------- 11-SEP-16 12.00.05.638000000 AM 11-SEP-16 01.00.23.855000000 AM 785 11-SEP-16 01.00.23.855000000 AM 11-SEP-16 02.00.40.073000000 AM 1057 11-SEP-16 02.00.40.073000000 AM 11-SEP-16 03.00.46.109000000 AM 1096 ... 15-SEP-16 09.00.04.192000000 PM 15-SEP-16 10.00.16.087000000 PM 786 15-SEP-16 10.00.16.087000000 PM 15-SEP-16 11.00.35.078000000 PM 779 15-SEP-16 11.00.35.078000000 PM 16-SEP-16 12.00.51.092000000 AM 831
© 2016 Pythian 31 HOW DID I DO IT? STEP 5: AGGREGATION SELECT Trunc(from_time, 'DD') t_from_time, SUM(delta_value) sum_delta_value FROM (SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id) WHERE from_time BETWEEN To_date('11092016', 'DDMMYYYY') AND To_date('16092016', 'DDMMYYYY') GROUP BY Trunc(from_time, 'DD') ORDER BY t_from_time; T_FROM_TIME SUM_DELTA_VALUE ------------------------------- --------------- 11.09.2016 00:00:00 21602 12.09.2016 00:00:00 29173 13.09.2016 00:00:00 32772 14.09.2016 00:00:00 30477 15.09.2016 00:00:00 26799
© 2016 Pythian 32 HOW DO I DO IT? STEP 6: VISUALIZATION
© 2016 Pythian 33 EXAMPLE 1 SQL Performance
© 2016 Pythian 34 EXAMPLE 1 THE PROBLEM • Amazon RDS for Oracle – db.m3.xlarge (4 vCPU, 15GiB) – 11.2.0.4 • Alert from the client – Could you please take a look at the DB utilization in prod? We're seeing 100% utilization and 795 connections. Could you please let us know if any specific connection is causing high CPU utilization.
© 2016 Pythian 35 EXAMPLE 1 5MIN ASH 10:21:04 SQL> @ashtop session_id,sql_id "event is null" sysdate-5/25/60 sysdate Total Seconds AAS %This SESSION_ID SQL_ID FIRST_SEEN LAST_SEEN ... --------- ------- ------- ---------- ------------- ------------------- ------------------- ... 285 1.0 12% | 738 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1336 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1358 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 745 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 882 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 2046 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 99 .3 4% | 1472 grbz54xqkr425 2016-06-06 10:17:04 2016-06-06 10:18:43 ... 64 .2 3% | 1472 fy4c407vamaks 2016-06-06 10:18:52 2016-06-06 10:21:26 ... 37 .1 2% | 2176 2016-06-06 10:18:29 2016-06-06 10:19:50 ... 22 .1 1% | 977 7tdudtm4x03np 2016-06-06 10:16:40 2016-06-06 10:17:01 ... 16 .1 1% | 795 3b8s69ct88pur 2016-06-06 10:16:40 2016-06-06 10:18:13 ... 15 .1 1% | 989 2016-06-06 10:17:04 2016-06-06 10:17:19 ... 13 .0 1% | 632 gdhy97b59bk81 2016-06-06 10:16:46 2016-06-06 10:17:07 ... 11 .0 0% | 1792 2016-06-06 10:16:59 2016-06-06 10:21:09 ... 11 .0 0% | 2133 atmhpvm759a97 2016-06-06 10:16:50 2016-06-06 10:21:02 ... Thank you, Tanel! http://blog.tanelpoder.com/files/scripts/ash/ashtop.sql
© 2016 Pythian 36 EXAMPLE 1 5MIN ASH 10:21:04 SQL> @ashtop session_id,sql_id "event is null" sysdate-5/25/60 sysdate Total Seconds AAS %This SESSION_ID SQL_ID FIRST_SEEN LAST_SEEN ... --------- ------- ------- ---------- ------------- ------------------- ------------------- ... 285 1.0 12% | 738 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1336 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1358 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 745 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 882 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 2046 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 99 .3 4% | 1472 grbz54xqkr425 2016-06-06 10:17:04 2016-06-06 10:18:43 ... 64 .2 3% | 1472 fy4c407vamaks 2016-06-06 10:18:52 2016-06-06 10:21:26 ... 37 .1 2% | 2176 2016-06-06 10:18:29 2016-06-06 10:19:50 ... 22 .1 1% | 977 7tdudtm4x03np 2016-06-06 10:16:40 2016-06-06 10:17:01 ... 16 .1 1% | 795 3b8s69ct88pur 2016-06-06 10:16:40 2016-06-06 10:18:13 ... 15 .1 1% | 989 2016-06-06 10:17:04 2016-06-06 10:17:19 ... 13 .0 1% | 632 gdhy97b59bk81 2016-06-06 10:16:46 2016-06-06 10:17:07 ... 11 .0 0% | 1792 2016-06-06 10:16:59 2016-06-06 10:21:09 ... 11 .0 0% | 2133 atmhpvm759a97 2016-06-06 10:16:50 2016-06-06 10:21:02 ... Thank you, Tanel! http://blog.tanelpoder.com/files/scripts/ash/ashtop.sql
© 2016 Pythian 37 Just one query to address! How hard can it be? EXAMPLE 1 SQL PERFORMANCE
© 2016 Pythian 38 EXAMPLE 1 THE EXECUTION PLAN 10:26:35 SQL> select * from table(dbms_xplan.display_awr('4db73upm43ck1',82218884)); Plan hash value: 82218884 -------------------------------------------------------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time | -------------------------------------------------------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | | | | 54163 (100)| | | 1 | TEMP TABLE TRANSFORMATION | | | | | | | | 2 | LOAD AS SELECT | | | | | | | | 3 | VIEW | | 179K| 13M| | 10337 (1)| 00:02:05 | | 4 | HASH JOIN | | 179K| 9830K| 5800K| 10337 (1)| 00:02:05 | | 5 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 3686K| | 794 (1)| 00:00:10 | | 6 | TABLE ACCESS FULL | ACTIVE_USER_VIEWS_BY_TOPIC | 1543K| 51M| | 5823 (1)| 00:01:10 | | 7 | VIEW | | 1 | 13 | | 2 (0)| 00:00:01 | | 8 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 9 | LOAD AS SELECT | | | | | | | | 10 | VIEW | | 110K| 8415K| | 10228 (1)| 00:02:03 | | 11 | HASH JOIN | | 110K| 6042K| 3568K| 10228 (1)| 00:02:03 | | 12 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 2265K| | 794 (1)| 00:00:10 | | 13 | TABLE ACCESS FULL | ACTIVE_USER_VIEWS_BY_TOPIC | 1543K| 51M| | 5823 (1)| 00:01:10 | | 14 | VIEW | | 1 | 13 | | 2 (0)| 00:00:01 | | 15 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | ... 1
© 2016 Pythian 39 EXAMPLE 1 THE EXECUTION PLAN ... | 16 | LOAD AS SELECT | | | | | | | | 17 | INTERSECTION | | | | | | | | 18 | SORT UNIQUE | | 179K| 7021K| 8472K| 2359 (1)| 00:00:29 | | 19 | VIEW | | 179K| 7021K| | 518 (1)| 00:00:07 | | 20 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 21 | SORT UNIQUE | | 110K| 4315K| 5208K| 1452 (1)| 00:00:18 | | 22 | VIEW | | 110K| 4315K| | 319 (1)| 00:00:04 | | 23 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 24 | LOAD AS SELECT | | | | | | | | 25 | WINDOW BUFFER | | 1 | 27 | | 860 (1)| 00:00:11 | | 26 | HASH GROUP BY | | 1 | 27 | | 860 (1)| 00:00:11 | | 27 | VIEW | | 110K| 2913K| | 856 (1)| 00:00:11 | | 28 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 110K| 4315K| | 856 (1)| 00:00:11 | | 29 | LOAD AS SELECT | | | | | | | | 30 | WINDOW BUFFER | | 42 | 756 | | 523 (2)| 00:00:07 | | 31 | HASH GROUP BY | | 42 | 756 | | 523 (2)| 00:00:07 | | 32 | VIEW | | 179K| 3159K| | 518 (1)| 00:00:07 | | 33 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 34 | LOAD AS SELECT | | | | | | | | 35 | WINDOW BUFFER | | 42 | 756 | | 322 (2)| 00:00:04 | | 36 | HASH GROUP BY | | 42 | 756 | | 322 (2)| 00:00:04 | | 37 | VIEW | | 110K| 1942K| | 319 (1)| 00:00:04 | ... 2
© 2016 Pythian 40 EXAMPLE 1 THE EXECUTION PLAN ... | 38 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 39 | FILTER | | | | | | | | 40 | SORT GROUP BY ROLLUP | | 1 | 579 | | 28077 (1)| 00:05:37 | | 41 | MERGE JOIN OUTER | | 10 | 5790 | | 28076 (1)| 00:05:37 | | 42 | MERGE JOIN OUTER | | 10 | 5790 | | 20379 (1)| 00:04:05 | | 43 | HASH JOIN OUTER | | 10 | 5660 | | 11818 (1)| 00:02:22 | | 44 | HASH JOIN OUTER | | 1 | 526 | | 11705 (1)| 00:02:21 | | 45 | HASH JOIN OUTER | | 1 | 495 | | 11383 (1)| 00:02:17 | | 46 | HASH JOIN OUTER | | 1 | 464 | | 10860 (1)| 00:02:11 | | 47 | HASH JOIN OUTER | | 1 | 407 | | 6041 (1)| 00:01:13 | | 48 | HASH JOIN OUTER | | 1 | 389 | | 6039 (1)| 00:01:13 | | 49 | HASH JOIN OUTER | | 1 | 371 | | 6037 (1)| 00:01:13 | | 50 | HASH JOIN OUTER | | 1 | 327 | | 5266 (1)| 00:01:04 | | 51 | HASH JOIN OUTER | | 1 | 296 | | 1603 (1)| 00:00:20 | | 52 | VIEW | | 1 | 148 | | 802 (1)| 00:00:10 | | 53 | HASH GROUP BY | | 1 | 75 | | 802 (1)| 00:00:10 | | 54 | HASH JOIN OUTER | | 9 | 675 | | 801 (1)| 00:00:10 | | 55 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 7 | 217 | | 8 (0)| 00:00:01 | | 56 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 5 (0)| 00:00:01 | | 57 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 58 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 7723K| | 792 (1)| 00:00:10 | | 59 | VIEW | | 1 | 148 | | 801 (1)| 00:00:10 | ... 3
© 2016 Pythian 41 EXAMPLE 1 THE EXECUTION PLAN ... | 60 | HASH GROUP BY | | 1 | 75 | | 801 (1)| 00:00:10 | | 61 | HASH JOIN OUTER | | 8 | 600 | | 800 (1)| 00:00:10 | | 62 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 7 | 217 | | 8 (0)| 00:00:01 | | 63 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 5 (0)| 00:00:01 | | 64 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 65 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 4747K| | 792 (1)| 00:00:10 | | 66 | VIEW | | 1 | 31 | | 3663 (1)| 00:00:44 | | 67 | HASH JOIN | | 1 | 62 | | 3663 (1)| 00:00:44 | | 68 | VIEW | | 1 | 31 | | 1658 (1)| 00:00:20 | | 69 | HASH GROUP BY | | 1 | 96 | | 1658 (1)| 00:00:20 | | 70 | HASH JOIN | | 1 | 96 | | 1657 (1)| 00:00:20 | | 71 | NESTED LOOPS | | 1 | 75 | | 863 (1)| 00:00:11 | | 72 | NESTED LOOPS | | 7 | 75 | | 863 (1)| 00:00:11 | | 73 | VIEW | | 1 | 40 | | 856 (1)| 00:00:11 | | 74 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 1 | 40 | | 856 (1)| 00:00:11 | | 75 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 4 (0)| 00:00:01 | | 76 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 77 | TABLE ACCESS BY INDEX ROWID| ACTIVE_USER_VIEWS_BY_TOPIC | 1 | 35 | | 7 (0)| 00:00:01 | | 78 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 3686K| | 794 (1)| 00:00:10 | | 79 | VIEW | | 1 | 31 | | 2004 (1)| 00:00:25 | | 80 | HASH GROUP BY | | 1 | 96 | | 2004 (1)| 00:00:25 | | 81 | HASH JOIN | | 1 | 96 | | 2003 (1)| 00:00:25 | ... 4
© 2016 Pythian 42 EXAMPLE 1 THE EXECUTION PLAN ... | 82 | NESTED LOOPS | | 1 | 75 | | 1209 (1)| 00:00:15 | | 83 | NESTED LOOPS | | 987 | 75 | | 1209 (1)| 00:00:15 | | 84 | VIEW | | 1 | 40 | | 856 (1)| 00:00:11 | | 85 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 1 | 40 | | 856 (1)| 00:00:11 | | 86 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 4 (0)| 00:00:01 | | 87 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 88 | TABLE ACCESS BY INDEX ROWID| ACTIVE_USER_VIEWS_BY_TOPIC | 1 | 35 | | 353 (0)| 00:00:05 | | 89 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 2265K| | 794 (1)| 00:00:10 | | 90 | VIEW | | 41 | 1804 | | 771 (1)| 00:00:10 | | 91 | HASH GROUP BY | | 41 | 1886 | | 771 (1)| 00:00:10 | | 92 | HASH JOIN | | 157 | 7222 | | 770 (1)| 00:00:10 | | 93 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 157 | 5338 | | 354 (0)| 00:00:05 | | 94 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 5 (0)| 00:00:01 | | 95 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 96 | TABLE ACCESS FULL | ASSESSMENTS | 270K| 3166K| | 415 (1)| 00:00:05 | | 97 | VIEW | | 42 | 756 | | 2 (0)| 00:00:01 | | 98 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 99 | VIEW | | 42 | 756 | | 2 (0)| 00:00:01 | | 100 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 101 | VIEW | | 42 | 2394 | | 4819 (1)| 00:00:58 | | 102 | HASH JOIN OUTER | | 42 | 6678 | | 4819 (1)| 00:00:58 | | 103 | HASH JOIN OUTER | | 42 | 5376 | | 4816 (1)| 00:00:58 | ... 5
© 2016 Pythian 43 EXAMPLE 1 THE EXECUTION PLAN ... | 104 | HASH JOIN OUTER | | 42 | 4074 | | 4813 (1)| 00:00:58 | | 105 | MERGE JOIN OUTER | | 42 | 2394 | | 4698 (1)| 00:00:57 | | 106 | VIEW | VW_FOJ_0 | 42 | 1848 | | 4 (0)| 00:00:01 | | 107 | HASH JOIN FULL OUTER | | 42 | 2604 | | 4 (0)| 00:00:01 | | 108 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 109 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 110 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 111 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE02_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 112 | BUFFER SORT | | 1 | 13 | | 4698 (1)| 00:00:57 | | 113 | VIEW | | 1 | 13 | | 112 (1)| 00:00:02 | | 114 | VIEW | | 1 | 13 | | 112 (1)| 00:00:02 | | 115 | SORT AGGREGATE | | 1 | 13 | | | | | 116 | VIEW | | 110K| 1402K| | 112 (1)| 00:00:02 | | 117 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 118 | VIEW | | 1 | 40 | | 115 (4)| 00:00:02 | | 119 | WINDOW BUFFER | | 1 | 40 | | 115 (4)| 00:00:02 | | 120 | HASH GROUP BY | | 1 | 40 | | 115 (4)| 00:00:02 | | 121 | VIEW | | 110K| 4315K| | 112 (1)| 00:00:02 | | 122 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 123 | VIEW | | 30 | 930 | | 3 (34)| 00:00:01 | | 124 | WINDOW BUFFER | | 30 | 930 | | 3 (34)| 00:00:01 | | 125 | HASH GROUP BY | | 30 | 930 | | 3 (34)| 00:00:01 | ... 6
© 2016 Pythian 44 EXAMPLE 1 THE EXECUTION PLAN ... | 126 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 127 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 128 | VIEW | | 30 | 930 | | 3 (34)| 00:00:01 | | 129 | WINDOW BUFFER | | 30 | 930 | | 3 (34)| 00:00:01 | | 130 | HASH GROUP BY | | 30 | 930 | | 3 (34)| 00:00:01 | | 131 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 132 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE02_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 133 | VIEW | | 42 | 1302 | | 523 (2)| 00:00:07 | | 134 | HASH GROUP BY | | 42 | 756 | | 523 (2)| 00:00:07 | | 135 | VIEW | | 179K| 3159K| | 518 (1)| 00:00:07 | | 136 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 137 | VIEW | | 42 | 1302 | | 322 (2)| 00:00:04 | | 138 | HASH GROUP BY | | 42 | 756 | | 322 (2)| 00:00:04 | | 139 | VIEW | | 110K| 1942K| | 319 (1)| 00:00:04 | | 140 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 141 | VIEW | | 110K| 4315K| | 112 (1)| 00:00:02 | | 142 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 143 | BUFFER SORT | | 1 | 13 | | 20379 (1)| 00:04:05 | | 144 | VIEW | | 1 | 13 | | 856 (1)| 00:00:11 | | 145 | VIEW | | 1 | 13 | | 856 (1)| 00:00:11 | | 146 | SORT AGGREGATE | | 1 | | | | | | 147 | VIEW | | 110K| | | 856 (1)| 00:00:11 | ... 7
© 2016 Pythian 45 ... | 148 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 110K| 4315K| | 856 (1)| 00:00:11 | | 149 | BUFFER SORT | | 1 | | | 27221 (1)| 00:05:27 | | 150 | VIEW | | 1 | | | 770 (1)| 00:00:10 | | 151 | VIEW | | 1 | | | 770 (1)| 00:00:10 | | 152 | SORT AGGREGATE | | 1 | 25 | | | | | 153 | HASH JOIN | | 157 | 3925 | | 770 (1)| 00:00:10 | | 154 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 157 | 2512 | | 354 (0)| 00:00:05 | | 155 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 5 (0)| 00:00:01 | | 156 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 157 | TABLE ACCESS FULL | ASSESSMENTS | 270K| 2374K| | 415 (1)| 00:00:05 | -------------------------------------------------------------------------------------------------------------------------------------- EXAMPLE 1 THE EXECUTION PLAN 8
© 2016 Pythian 46 EXAMPLE 1 THE EXECUTION PLAN - #SHTF
© 2016 Pythian 47 EXAMPLE 1 THE OPTIONS • What can one do in this case? – Tune the query? – Check / collect fresh statistics? – Flush the shared pool? (hoping it's bind variable peeking) – Run SQL Tuning Advisor? – or... • How did the SQL statement execute in the past?
© 2016 Pythian 48 EXAMPLE 1 AWR_SQLID_PERF_TREND_BY_PLAN.SQL select hss.instance_number inst, to_char(trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24,'dd.mm.yyyy hh24:mi:ss') time, plan_hash_value, sum(hss.executions_delta) executions, round(sum(hss.elapsed_time_delta)/1000000,3) elapsed_time_s, round(sum(hss.cpu_time_delta)/1000000,3) cpu_time_s, round(sum(hss.iowait_delta)/1000000,3) iowait_s, round(sum(hss.clwait_delta)/1000000,3) clwait_s, round(sum(hss.apwait_delta)/1000000,3) apwait_s, round(sum(hss.ccwait_delta)/1000000,3) ccwait_s, round(sum(hss.rows_processed_delta),3) rows_processed, round(sum(hss.buffer_gets_delta),3) buffer_gets, round(sum(hss.disk_reads_delta),3) disk_reads, round(sum(hss.direct_writes_delta),3) direct_writes from dba_hist_sqlstat hss, dba_hist_snapshot hs where hss.sql_id='&sql_id' and hss.snap_id=hs.snap_id and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 group by hss.instance_number, trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24, plan_hash_value having sum(hss.executions_delta)>0
© 2016 Pythian 49 10:21:28 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 5 24 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- 02.06.2016 00:00:00 1597670781 291 278.018 254.272 22.717 19787274.000 50740.000 03.06.2016 00:00:00 1597670781 205 204.755 189.479 14.583 14986545.000 71451.000 04.06.2016 00:00:00 1597670781 3 8.275 4.483 3.936 314594.000 15790.000 04.06.2016 00:00:00 4142332636 2 6.149 3.933 2.307 119766.000 47343.000 05.06.2016 00:00:00 82218884 47 11805.808 5766.836 22.799 1247372099.000 173254.000 06.06.2016 00:00:00 82218884 127 7216.393 3751.640 9.431 522088617.000 24677.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE LOOKING AT 5 DAYS OF PERFORMANCE DATA AT ONCE!
© 2016 Pythian 50 10:21:28 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 5 24 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- 02.06.2016 00:00:00 1597670781 291 278.018 254.272 22.717 19787274.000 50740.000 03.06.2016 00:00:00 1597670781 205 204.755 189.479 14.583 14986545.000 71451.000 04.06.2016 00:00:00 1597670781 3 8.275 4.483 3.936 314594.000 15790.000 04.06.2016 00:00:00 4142332636 2 6.149 3.933 2.307 119766.000 47343.000 05.06.2016 00:00:00 82218884 47 11805.808 5766.836 22.799 1247372099.000 173254.000 06.06.2016 00:00:00 82218884 127 7216.393 3751.640 9.431 522088617.000 24677.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE LOOKING AT 5 DAYS OF PERFORMANCE DATA AT ONCE!
© 2016 Pythian 51 EXAMPLE 1 SQL PLAN BASELINE FROM AWR exec DBMS_SQLTUNE.CREATE_SQLSET(sqlset_name => 'CR1064802', description => 'Plan for sql_id 4db73upm43ck1'); DECLARE cur sys_refcursor; BEGIN OPEN cur FOR SELECT VALUE(P) FROM table(dbms_sqltune.select_workload_repository(11434,12880,'sql_id=''4db73upm43ck1'' and plan_hash_value=1597670781' , NULL, NULL, NULL, NULL, NULL, NULL, 'ALL')) P; DBMS_SQLTUNE.LOAD_SQLSET(load_option=>'MERGE',sqlset_name => 'CR1064802', populate_cursor => cur); CLOSE cur; END; VARIABLE cnt NUMBER EXECUTE :cnt := DBMS_SPM.LOAD_PLANS_FROM_SQLSET( sqlset_name => 'CR1064802', basic_filter => 'sql_id=''4db73upm43ck1'''); comment on table cl_report_owner.ACTIVE_USER_VIEWS_BY_TOPIC is '';
© 2016 Pythian 52 11:03:11 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 2 1 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- ... 06.06.2016 00:00:00 82218884 3 4.600 4.019 .608 1425936.000 965.000 06.06.2016 01:00:00 82218884 4 14.374 11.740 2.756 4726784.000 12836.000 06.06.2016 02:00:00 82218884 20 34.437 33.125 .619 7475353.000 264.000 06.06.2016 03:00:00 82218884 3 4.527 4.485 .032 1205412.000 14.000 06.06.2016 04:00:00 82218884 18 13.543 13.366 .132 1346631.000 57.000 06.06.2016 05:00:00 82218884 22 43.248 40.954 2.217 14290668.000 4439.000 06.06.2016 06:00:00 82218884 11 78.463 77.897 .311 45601072.000 79.000 06.06.2016 07:00:00 82218884 2 3.338 2.155 1.237 165446.000 3369.000 06.06.2016 08:00:00 82218884 10 9.241 9.098 .098 900049.000 61.000 06.06.2016 09:00:00 82218884 34 7010.622 3554.801 1.422 444951266.000 2593.000 06.06.2016 10:00:00 82218884 15 10942.980 4653.979 4.637 586951399.000 2078.000 06.06.2016 10:00:00 1597670781 1 7.910 2.047 .323 117577.000 28.000 06.06.2016 11:00:00 1597670781 1 1.028 1.008 .020 117468.000 28.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE Yea, but ... ... what if the good plan is not captured in the AWR?
© 2016 Pythian 53 -- Create the procedure that will be used to collect the execution plans create or replace procedure PERF.XXSTABILITY_CAPTURE_PLANS is cur sys_refcursor; cursor clist is with raw_sqlids as (select sql_id from v$sqlarea where sql_id in (select sql_id from v$active_session_history where sample_time>=sysdate-1/24 and sql_plan_hash_value>0 and machine like '%-hesvc-app0%') and plan_hash_value>0 and executions>2 union select sql_id from v$sqlarea where sql_id in (select sql_id from v$open_cursor where sid in (select sid from v$session where machine like '%-hesvc-app0%')) and plan_hash_value>0 and executions>2 and parsing_schema_name not in ('ANONYMOUS','APEX_030200','APEX_040000','APEX_SSO','APPQOSSYS', 'CTXSYS','DBSNMP','DIP','EXFSYS','FLOWS_FILES','MDSYS','OLAPSYS','ORACLE_OCM','ORDDATA','ORDPLUGINS','ORDSYS','OUTLN','OWBSYS ','SI_INFORMTN_SCHEMA','SQLTXADMIN','SQLTXPLAIN','SYS','SYSMAN','SYSTEM','TRCANLZR','WMSYS','XDB','XS$NULL')), sqlids as (select distinct ''''||sql_id||'''' sql_id, trunc((rownum-1)/200) rn from raw_sqlids) select listagg(s.sql_id,',') within group (order by s.sql_id) as sql_id_filter from sqlids s group by s.rn; c clist%rowtype; BEGIN for c in clist loop OPEN cur FOR SELECT VALUE(P) FROM table(dbms_sqltune.select_cursor_cache(basic_filter=>'sql_id in('||c.sql_id_filter||')', attribute_list=>'ALL')) P; DBMS_SQLTUNE.LOAD_SQLSET(load_option=>'MERGE',sqlset_name => 'XXSTABILITY_EXECUTION_PLANS', populate_cursor => cur); CLOSE cur; end loop; Commit; END; / EXAMPLE 1 PRESERVING GOOD EXECUTION PLANS
© 2016 Pythian 54 -- Create the Scheduler Job which will run on every hour's 47th minute Begin Dbms_scheduler.create_job (job_name => 'COLLECT_XXSTABILITY_PLANS', job_type => 'STORED_PROCEDURE', job_action => 'PERF.XXSTABILITY_CAPTURE_PLANS', start_date => SYSDATE, repeat_interval => 'FREQ=HOURLY; INTERVAL=1; BYMINUTE=47;', enabled => TRUE); End; / -- Loading the baseline from SELECT_SQLSET for a specific sql_id and plan VARIABLE cnt NUMBER EXECUTE :cnt := DBMS_SPM.LOAD_PLANS_FROM_SQLSET( - sqlset_name => 'XXSTABILITY_EXECUTION_PLANS', - basic_filter => 'sql_id=''dtzsc12fbbjk2'' and plan_hash_value=2896891279'); EXAMPLE 1 PRESERVING EXECUTION PLANS
© 2016 Pythian 55 EXAMPLE 2 IO performance
© 2016 Pythian 56 • Oracle e-Business Suite – R12.2 – 12.1.0.2 • Alert from the client – Starting from September 5th, everything has become a lot more slower. Entering a sales order can take up to 15 minutes instead of usual 15 seconds. • Not enough detail to look at any specific process or query EXAMPLE 2 THE PROBLEM
© 2016 Pythian 57 select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, stat_name, sum(delta_value) value_delta from (select hss.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, stat_name, value, value-(lag(value,1) over(partition by hs.startup_time, stat_name order by hss.snap_id)) delta_value from dba_hist_sys_time_model hss, dba_hist_snapshot hs where hss.snap_id=hs.snap_id and hss.instance_number=hs.instance_number and hs.end_interval_time>=trunc(sysdate)-&days_history+1 and hss.stat_name like '&stat_name') group by time, stat_name order by to_date(time,'DD.MM.YYYY HH24:MI:SS'), 1; EXAMPLE 2 AWR_SYS_TIME_MODEL_TREND.SQL
© 2016 Pythian 58 EXAMPLE 2 CPU OR WAITING? SQL> @awr_sys_time_model_trend.sql "DB CPU" 14 24 TIME STAT_NAME VALUE_DELTA ------------------- ---------------------------------------------------------------- ----------------- 28.08.2016 00:00:00 DB CPU 204502770965 29.08.2016 00:00:00 DB CPU 240124304792 30.08.2016 00:00:00 DB CPU 243171465732 31.08.2016 00:00:00 DB CPU 271839933025 01.09.2016 00:00:00 DB CPU 398156022996 02.09.2016 00:00:00 DB CPU 343376689897 03.09.2016 00:00:00 DB CPU 262729582269 04.09.2016 00:00:00 DB CPU 79584869551 05.09.2016 00:00:00 DB CPU 87671787128 06.09.2016 00:00:00 DB CPU 103383834430 07.09.2016 00:00:00 DB CPU 16642806685 08.09.2016 00:00:00 DB CPU 215983418676 09.09.2016 00:00:00 DB CPU 100324311282 10.09.2016 00:00:00 DB CPU 11631538710 We're spending less time on CPU, so it's likely we're spending more time somewhere else!
© 2016 Pythian 59 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES) select * from ( select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, wait_class, sum(delta_time_waited)/1000000 wait_time_s from (select hse.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, EVENT_NAME, WAIT_CLASS, TIME_WAITED_MICRO-(lag(TIME_WAITED_MICRO,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_time_waited from DBA_HIST_SYSTEM_EVENT hse, DBA_HIST_SNAPSHOT hs where hse.snap_id=hs.snap_id and hs.end_interval_time>=trunc(sysdate)-&days_history+1) group by time, event_name, wait_class) pivot ( sum(wait_time_s) for wait_class in ('Administrative','Application','Cluster' ,'Commit','Concurrency','Configuration' ,'Network','Other','Queueing' ,'Scheduler','System I/O','User I/O','Idle') ) order by to_date(time,'DD.MM.YYYY HH24:MI:SS');
© 2016 Pythian 60 SQL> @awr_wait_class_hist.sql 14 1 TIME 'Administrative' 'Application' 'Commit' 'Concurrency' 'Other' 'Scheduler' 'System I/O' 'User I/O' 'Idle' ------------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- 28.08.2016 01:00:00 0 29 27 25 35 0 411 9494 1167334 28.08.2016 02:00:00 0 28 12 16 84 0 435 5944 1220988 28.08.2016 03:00:00 0 28 15 1 17 0 72 5955 1102848 28.08.2016 04:00:00 0 27 13 16 17 0 68 4715 1160580 28.08.2016 05:00:00 0 28 23 0 81 0 361 3585 1100722 28.08.2016 06:00:00 0 28 18 0 58 0 264 3281 1162927 28.08.2016 07:00:00 0 29 13 0 19 0 61 1908 1219200 28.08.2016 08:00:00 0 29 14 0 23 0 67 3752 1121374 28.08.2016 09:00:00 0 29 16 0 22 0 68 2042 1102494 28.08.2016 10:00:00 0 28 14 0 20 0 67 2191 1317052 ... 03.09.2016 19:00:00 0 11 16 1 79 0 238 2943 1191940 03.09.2016 20:00:00 0 15 18 2 96 0 205 3640 1284953 03.09.2016 21:00:00 0 13 25 1 102 0 439 3985 1155491 03.09.2016 22:00:00 0 47 7068 965 1147 0 10445 11709 1201118 03.09.2016 23:00:00 0 43 19698 512 2354 0 56763 16627 1119525 04.09.2016 01:00:00 4 500 50406 6114 14543 0 131447 79139 2224995 04.09.2016 04:00:00 55 1067 65711 4350 16350 0 165858 199986 3820512 04.09.2016 05:00:00 0 29 12118 766 2378 0 13428 55055 1160076 04.09.2016 06:00:00 1 32 2510 54 1242 0 7708 47570 1295052 04.09.2016 07:00:00 0 21 546 1 305 0 3139 42468 1299331 04.09.2016 08:00:00 0 18 265 16 200 0 1788 28516 1242703 04.09.2016 09:00:00 0 13 377 1 174 0 1777 13472 1194265 04.09.2016 10:00:00 0 15 447 1 215 0 2371 15196 1242622 04.09.2016 11:00:00 0 13 404 2 205 0 2312 14984 1207877 04.09.2016 12:00:00 0 13 405 1 238 0 2664 16214 1217226 04.09.2016 13:00:00 0 15 597 1 354 0 3093 14902 1232148 04.09.2016 14:00:00 0 17 486 1 224 0 2252 12952 1313141 04.09.2016 15:00:00 0 5 486 1 254 0 2404 13285 1176055 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
© 2016 Pythian 61 SQL> @awr_wait_class_hist.sql 14 1 TIME 'Administrative' 'Application' 'Commit' 'Concurrency' 'Other' 'Scheduler' 'System I/O' 'User I/O' 'Idle' ------------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- 28.08.2016 01:00:00 0 29 27 25 35 0 411 9494 1167334 28.08.2016 02:00:00 0 28 12 16 84 0 435 5944 1220988 28.08.2016 03:00:00 0 28 15 1 17 0 72 5955 1102848 28.08.2016 04:00:00 0 27 13 16 17 0 68 4715 1160580 28.08.2016 05:00:00 0 28 23 0 81 0 361 3585 1100722 28.08.2016 06:00:00 0 28 18 0 58 0 264 3281 1162927 28.08.2016 07:00:00 0 29 13 0 19 0 61 1908 1219200 28.08.2016 08:00:00 0 29 14 0 23 0 67 3752 1121374 28.08.2016 09:00:00 0 29 16 0 22 0 68 2042 1102494 28.08.2016 10:00:00 0 28 14 0 20 0 67 2191 1317052 ... 03.09.2016 19:00:00 0 11 16 1 79 0 238 2943 1191940 03.09.2016 20:00:00 0 15 18 2 96 0 205 3640 1284953 03.09.2016 21:00:00 0 13 25 1 102 0 439 3985 1155491 03.09.2016 22:00:00 0 47 7068 965 1147 0 10445 11709 1201118 03.09.2016 23:00:00 0 43 19698 512 2354 0 56763 16627 1119525 04.09.2016 01:00:00 4 500 50406 6114 14543 0 131447 79139 2224995 04.09.2016 04:00:00 55 1067 65711 4350 16350 0 165858 199986 3820512 04.09.2016 05:00:00 0 29 12118 766 2378 0 13428 55055 1160076 04.09.2016 06:00:00 1 32 2510 54 1242 0 7708 47570 1295052 04.09.2016 07:00:00 0 21 546 1 305 0 3139 42468 1299331 04.09.2016 08:00:00 0 18 265 16 200 0 1788 28516 1242703 04.09.2016 09:00:00 0 13 377 1 174 0 1777 13472 1194265 04.09.2016 10:00:00 0 15 447 1 215 0 2371 15196 1242622 04.09.2016 11:00:00 0 13 404 2 205 0 2312 14984 1207877 04.09.2016 12:00:00 0 13 405 1 238 0 2664 16214 1217226 04.09.2016 13:00:00 0 15 597 1 354 0 3093 14902 1232148 04.09.2016 14:00:00 0 17 486 1 224 0 2252 12952 1313141 04.09.2016 15:00:00 0 5 486 1 254 0 2404 13285 1176055 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
© 2016 Pythian 62 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
© 2016 Pythian 63 • What could it be? – We're doing way more IOs – The IOs are much slower – … or both EXAMPLE 2 “USER I/O” IS MISBEHAVING
© 2016 Pythian 64 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O? select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, event_name, sum(delta_total_waits) total_waits, round(sum(delta_time_waited/1000000),3) total_time_s, round(sum(delta_time_waited)/decode(sum(delta_total_waits),0,null,sum(delta_total_waits))/1000,3) avg_time_ms from (select hse.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, EVENT_NAME, WAIT_CLASS, TOTAL_WAITS-(lag(TOTAL_WAITS,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_total_waits, TIME_WAITED_MICRO-(lag(TIME_WAITED_MICRO,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_time_waited from DBA_HIST_SYSTEM_EVENT hse, DBA_HIST_SNAPSHOT hs where hse.snap_id=hs.snap_id and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 and hse.WAIT_CLASS like '&wait_class') group by time, event_name order by 2, to_date(time,'DD.MM.YYYY HH24:MI:SS');
© 2016 Pythian 65 SQL> @awr_wait_class_events_trend.sql "User I/O" 14 24 TIME EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- ... 28.08.2016 00:00:00 db file scattered read 110851 2332.624 21.043 29.08.2016 00:00:00 db file scattered read 553580 7823.429 14.132 30.08.2016 00:00:00 db file scattered read 710412 8074.224 11.366 31.08.2016 00:00:00 db file scattered read 686109 8779.227 12.796 01.09.2016 00:00:00 db file scattered read 3697362 79473.993 21.495 02.09.2016 00:00:00 db file scattered read 905800 12899.301 14.241 03.09.2016 00:00:00 db file scattered read 181632 5821.909 32.053 04.09.2016 00:00:00 db file scattered read 142537 44193.256 310.048 05.09.2016 00:00:00 db file scattered read 326866 32658.860 99.915 06.09.2016 00:00:00 db file scattered read 258082 29108.440 112.788 07.09.2016 00:00:00 db file scattered read 534720 44248.693 82.751 08.09.2016 00:00:00 db file scattered read 158424 22083.148 139.393 09.09.2016 00:00:00 db file scattered read 313153 47113.132 150.448 10.09.2016 00:00:00 db file scattered read 145106 14431.633 99.456 ... EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
© 2016 Pythian 66 SQL> @awr_wait_class_events_trend.sql "User I/O" 14 24 TIME EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- ... 28.08.2016 00:00:00 db file sequential read 15851686 24652.524 1.555 29.08.2016 00:00:00 db file sequential read 12892354 41750.047 3.238 30.08.2016 00:00:00 db file sequential read 37228593 89856.611 2.414 31.08.2016 00:00:00 db file sequential read 23431904 71298.237 3.043 01.09.2016 00:00:00 db file sequential read 26866886 82063.899 3.054 02.09.2016 00:00:00 db file sequential read 20273240 60593.329 2.989 03.09.2016 00:00:00 db file sequential read 5114798 28039.911 5.482 04.09.2016 00:00:00 db file sequential read 7142647 401434.020 56.202 05.09.2016 00:00:00 db file sequential read 8949913 292546.728 32.687 06.09.2016 00:00:00 db file sequential read 19490699 665357.848 34.137 07.09.2016 00:00:00 db file sequential read 16387403 541792.005 33.061 08.09.2016 00:00:00 db file sequential read 6989352 241881.608 34.607 09.09.2016 00:00:00 db file sequential read 12549290 453219.850 36.115 10.09.2016 00:00:00 db file sequential read 4823556 119185.469 24.709 ... EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
© 2016 Pythian 67 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
© 2016 Pythian 68 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
© 2016 Pythian 69 The Issue was passed on to the storage team. They found 2 faulty drives. EXAMPLE 2 THE ROOT CAUSE
© 2016 Pythian 70 EXAMPLE 3 Redo size
© 2016 Pythian 71 • Oracle e-Business Suite – R12.1 – 12.1.0.2 • Alert from the monitoring – We're running out of space on the mount point that's used as a long term storage for archived logs (10 days) EXAMPLE 3 THE PROBLEM
© 2016 Pythian 72 select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, stat_name, sum(delta_value) value_delta from (select hss.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, stat_name, value, value-(lag(value,1) over(partition by hs.startup_time, stat_name order by hss.snap_id)) delta_value from dba_hist_sysstat hss, dba_hist_snapshot hs where hss.snap_id=hs.snap_id and hss.instance_number=hs.instance_number and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 and hss.stat_name like ' ') group by time, stat_name order by 2, to_date(time,'DD.MM.YYYY HH24:MI:SS'); EXAMPLE 3 AWR_STAT_TREND.SQL
© 2016 Pythian 73 SQL> @awr_stat_trend "redo size" 30 4 TIME STAT_NAME VALUE_DELTA ------------------- ---------------- ----------------- ... 22.10.2016 20:00:00 redo size 29672405900 23.10.2016 00:00:00 redo size 45868325988 23.10.2016 04:00:00 redo size 28625802412 23.10.2016 08:00:00 redo size 46741701432 23.10.2016 12:00:00 redo size 48761724412 23.10.2016 16:00:00 redo size 46302390088 23.10.2016 20:00:00 redo size 29891530168 24.10.2016 00:00:00 redo size 39012726684 24.10.2016 04:00:00 redo size 33850511384 24.10.2016 08:00:00 redo size 27530642800 24.10.2016 12:00:00 redo size 103480430448 24.10.2016 16:00:00 redo size 124196709984 24.10.2016 20:00:00 redo size 30110908608 25.10.2016 00:00:00 redo size 38282203780 25.10.2016 04:00:00 redo size 34092919668 EXAMPLE 3 WHEN DO DID WE GENERATE MORE REDO?
© 2016 Pythian 74 EXAMPLE 3 WHEN DO DID WE GENERATE MORE REDO?
© 2016 Pythian 75 EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385764.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385765.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385766.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385767.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385768.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385769.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385770.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385771.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385772.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385773.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385774.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385775.arc'); EXEC DBMS_LOGMNR.START_LOGMNR(OPTIONS => DBMS_LOGMNR.DICT_FROM_ONLINE_CATALOG); EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
© 2016 Pythian 76 set pages 50000 lines 500 tab off time on serverout on col owner for a30 col object_name for a30 col subobject_name for a30 col object_type for a30 select data_obj#, owner, object_name, subobject_name, count(*) REDO_CNT, sum(R_SIZE_B)/1024/1024 REDO_MB from ( select data_obj#, THREAD#, SEQUENCE#, RBABLK, RBABYTE, RBABLK*512+RBABYTE b_offset, (lead(RBABLK*512+RBABYTE) over (partition by RBASQN order by RBABLK*512+RBABYTE))-(RBABLK*512+RBABYTE) R_SIZE_B from V$LOGMNR_CONTENTS) lc, dba_objects o where o.data_object_id(+)=lc.data_obj# group by data_obj#, owner, object_name, subobject_name order by 6; EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
© 2016 Pythian 77 DATA_OBJ# OWNER OBJECT_NAME SUBOBJECT_NAME REDO_CNT REDO_MB ---------- ----------- ------------------------------ ----------------- ---------- ---------- ... 4214022 AR SYS_LOB0004214021C00002$$ 16997 55.844368 4214007 AR SYS_LOB0004214006C00002$$ 11991 60.0471535 160758 APPLSYS MO_GLOB_ORG_ACCESS_TMP_U1 431892 80.7151947 112443 784710 142.132919 0 CTXSYS DR$DBO 859097 253.653015 137676 3774665 565.485004 181691 APPS XX_OM_PRICING_DETAILS_GTT_N100 3840811 709.441868 1026278 APPS XX_OM_PRICING_DETAILS_GTT_NS 3870531 772.384396 181690 APPS XX_OM_PRICING_DETAILS_GTT_N400 3886342 817.994175 181692 APPS XX_OM_PRICING_DETAILS_GTT_N200 3943281 882.903255 880271 APPS XX_OM_PRICING_DETAIL_GTT_N20 4118952 1245.52386 181689 APPS XX_OM_PRICING_DETAILS_GTT_N300 4238632 1464.44829 307 rows selected. EXEC DBMS_LOGMNR.END_LOGMNR; EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
© 2016 Pythian 78 One of the concurrent programs inserting into the GTT was executed more often during the time of the issue A) Added +append and set alter session set temp_undo_enabled=true B) The frequency of the concurrent program was reduced C) 4 of 6 indexes on the GTT were dropped EXAMPLE 3 THE ROOT CAUSE Thanks, @davidmkurtz
© 2016 Pythian 79 Redo Size (M) EXAMPLE 3 THE ROOT CAUSE Conventional load Direct path load (+append) Temporary Undo = False 674 200 Temporary Undo = True 13 8
© 2016 Pythian 80 SUMMARY
© 2016 Pythian 81 • AWR reports are not good in displaying how performance changes over time. • Many performance metrics stored in AWR can be: – Put on a timeline to see how they change over time – Useful for Troubleshooting – Useful for Prediction • Remember, you're still working with aggregated data SUMMARY
© 2016 Pythian 82 SUMMARY Photo by Kate Ter Haar / CC BY 2.0
© 2016 Pythian 83 AWR is a complex beast, but it’s not necessary to know much to start mining Be Brave! But buy Diagnostics pack licenses first.
© 2016 Pythian 84 THANK YOU @MarisDBA Elsins@pythian.com

More Related Content

PDF
Migrating and Running DBs on Amazon RDS for Oracle
Maris Elsins
 
PDF
Oracle Databases on AWS - Getting the Best Out of RDS and EC2
Maris Elsins
 
PDF
An AWS DMS Replication Journey from Oracle to Aurora MySQL
Maris Elsins
 
PDF
Managing multi tenant resource toward Hive 2.0
Kai Sasaki
 
PDF
RDS for MySQL, No BS Operations and Patterns
Laine Campbell
 
PPTX
Best Practices for running the Oracle Database on EC2 webinar
Tom Laszewski
 
PPTX
Advanced data migration techniques for Amazon RDS
Tom Laszewski
 
PDF
AWS Database Services-Philadelphia AWS User Group-4-17-2018
Bert Zahniser
 
Migrating and Running DBs on Amazon RDS for Oracle
Maris Elsins
 
Oracle Databases on AWS - Getting the Best Out of RDS and EC2
Maris Elsins
 
An AWS DMS Replication Journey from Oracle to Aurora MySQL
Maris Elsins
 
Managing multi tenant resource toward Hive 2.0
Kai Sasaki
 
RDS for MySQL, No BS Operations and Patterns
Laine Campbell
 
Best Practices for running the Oracle Database on EC2 webinar
Tom Laszewski
 
Advanced data migration techniques for Amazon RDS
Tom Laszewski
 
AWS Database Services-Philadelphia AWS User Group-4-17-2018
Bert Zahniser
 

What's hot (11)

PPTX
October 2016 HUG: Architecture of an Open Source RDBMS powered by HBase and ...
Yahoo Developer Network
 
PPTX
SQLIO - measuring storage performance
valerian_ceaus
 
PDF
Scaling Pinterest
C4Media
 
PPTX
Building Effective Near-Real-Time Analytics with Spark Streaming and Kudu
Jeremy Beard
 
PPTX
Achieving 100k Queries per Hour on Hive on Tez
DataWorks Summit/Hadoop Summit
 
PPTX
Migrate Oracle database to Amazon RDS
Jesus Guzman
 
PPTX
Optimizing Apache HBase for Cloud Storage in Microsoft Azure HDInsight
HBaseCon
 
PPTX
Using Kafka and Kudu for fast, low-latency SQL analytics on streaming data
Mike Percy
 
PPTX
Azure + DataStax Enterprise (DSE) Powers Office365 Per User Store
DataStax Academy
 
PPTX
SQL Server to Redshift Data Load Using SSIS
Marc Leinbach
 
PPTX
January 2015 HUG: Using HBase Co-Processors to Build a Distributed, Transacti...
Yahoo Developer Network
 
October 2016 HUG: Architecture of an Open Source RDBMS powered by HBase and ...
Yahoo Developer Network
 
SQLIO - measuring storage performance
valerian_ceaus
 
Scaling Pinterest
C4Media
 
Building Effective Near-Real-Time Analytics with Spark Streaming and Kudu
Jeremy Beard
 
Achieving 100k Queries per Hour on Hive on Tez
DataWorks Summit/Hadoop Summit
 
Migrate Oracle database to Amazon RDS
Jesus Guzman
 
Optimizing Apache HBase for Cloud Storage in Microsoft Azure HDInsight
HBaseCon
 
Using Kafka and Kudu for fast, low-latency SQL analytics on streaming data
Mike Percy
 
Azure + DataStax Enterprise (DSE) Powers Office365 Per User Store
DataStax Academy
 
SQL Server to Redshift Data Load Using SSIS
Marc Leinbach
 
January 2015 HUG: Using HBase Co-Processors to Build a Distributed, Transacti...
Yahoo Developer Network
 
Ad

Viewers also liked (20)

PDF
Node.js - async for the rest of us.
Mike Brevoort
 
PPTX
NoSQL and SOA
Guido Schmutz
 
PPT
Oracle Event Delivery Network (EDN) of SOA Suite 11g
Guido Schmutz
 
PDF
How_to_build_GameServer_2
Peter Rybar
 
PDF
MongoDB performance tuning and load testing, NOSQL Now! 2013 Conference prese...
ronwarshawsky
 
PPTX
The True State of the Oracle Public Cloud - Dutch Oracle Architects Platform ...
Lucas Jellema
 
PDF
AMIS Oracle OpenWorld 2013 Review Part 1 - Intro Overview Innovation, Hardwar...
Getting value from IoT, Integration and Data Analytics
 
PPTX
Review Oracle OpenWorld 2015 - Overview, Main themes, Announcements and Future
Lucas Jellema
 
PPTX
Systems on the edge - your stepping stones into Oracle Public PaaS Cloud - AM...
Lucas Jellema
 
PDF
10 Key MongoDB Performance Indicators
iammutex
 
PPTX
Mobile Database and Service Oriented Architecture
Lucas Jellema
 
PPTX
Introducing Node.js in an Oracle technology environment (including hands-on)
Lucas Jellema
 
PPTX
Introducing Oracle Real-Time Integration Business Insight
Lucas Jellema
 
PDF
Gluent New World #02 - SQL-on-Hadoop : A bit of History, Current State-of-the...
Mark Rittman
 
PDF
HMC_Industry_Report_Drone_Technology_160321[1]
Robert Cheek
 
PPTX
Oracle Database 12c - Introducing SQL Pattern Recognition through MATCH_RECOG...
Lucas Jellema
 
PPTX
What is the Oracle PaaS Cloud for Developers (Oracle Cloud Day, The Netherlan...
Lucas Jellema
 
PPT
Fast querying indexing for performance (4)
MongoDB
 
PDF
SQL Outer Joins for Fun and Profit
Karwin Software Solutions LLC
 
PPTX
Ranges, ranges everywhere (Oracle SQL)
Stew Ashton
 
Node.js - async for the rest of us.
Mike Brevoort
 
NoSQL and SOA
Guido Schmutz
 
Oracle Event Delivery Network (EDN) of SOA Suite 11g
Guido Schmutz
 
How_to_build_GameServer_2
Peter Rybar
 
MongoDB performance tuning and load testing, NOSQL Now! 2013 Conference prese...
ronwarshawsky
 
The True State of the Oracle Public Cloud - Dutch Oracle Architects Platform ...
Lucas Jellema
 
AMIS Oracle OpenWorld 2013 Review Part 1 - Intro Overview Innovation, Hardwar...
Getting value from IoT, Integration and Data Analytics
 
Review Oracle OpenWorld 2015 - Overview, Main themes, Announcements and Future
Lucas Jellema
 
Systems on the edge - your stepping stones into Oracle Public PaaS Cloud - AM...
Lucas Jellema
 
10 Key MongoDB Performance Indicators
iammutex
 
Mobile Database and Service Oriented Architecture
Lucas Jellema
 
Introducing Node.js in an Oracle technology environment (including hands-on)
Lucas Jellema
 
Introducing Oracle Real-Time Integration Business Insight
Lucas Jellema
 
Gluent New World #02 - SQL-on-Hadoop : A bit of History, Current State-of-the...
Mark Rittman
 
HMC_Industry_Report_Drone_Technology_160321[1]
Robert Cheek
 
Oracle Database 12c - Introducing SQL Pattern Recognition through MATCH_RECOG...
Lucas Jellema
 
What is the Oracle PaaS Cloud for Developers (Oracle Cloud Day, The Netherlan...
Lucas Jellema
 
Fast querying indexing for performance (4)
MongoDB
 
SQL Outer Joins for Fun and Profit
Karwin Software Solutions LLC
 
Ranges, ranges everywhere (Oracle SQL)
Stew Ashton
 
Ad

Similar to Mining AWR V2 - Trend Analysis (20)

PDF
Maris+Elsins+-+Mining+AWR+V2(1)11111.pdg
cookie1969
 
PPTX
AWR DB performance Data Mining - Collaborate 2015
Yury Velikanov
 
PDF
OTN tour 2015 AWR data mining
Andrejs Vorobjovs
 
PPTX
Awr doag
Marcin Przepiórowski
 
PDF
Whitepaper: Mining the AWR repository for Capacity Planning and Visualization
Kristofferson A
 
PDF
Donatone_Grid Performance(2)11111111.pdg
cookie1969
 
PDF
AWR and ASH in an EM12c World
Kellyn Pot'Vin-Gorman
 
DOC
Analyzing awr report
satish Gaddipati
 
PPTX
AWR and ASH Deep Dive
Kellyn Pot'Vin-Gorman
 
PPTX
Kellyn Pot'Vin-Gorman - Awr and Ash
gaougorg
 
DOCX
Step by stepDoc for Oracle TuningsandAWR.docx
Ganesh Bhosale
 
PDF
Ebs dba con4696_pdf_4696_0001
jucaab
 
DOCX
Predictive performance analysis using sql pattern matching
Horia Berca
 
PPTX
ASH and AWR on DB12c
Kellyn Pot'Vin-Gorman
 
PPTX
Oracle AWR Data mining
Yury Velikanov
 
PDF
Mining the AWR: Alternative Methods for Identification of the Top SQLs (inclu...
Maris Elsins
 
PDF
2013 Collaborate - OAUG - Presentation
Biju Thomas
 
PPT
Using AWR for SQL Analysis
Texas Memory Systems, and IBM Company
 
PDF
VirtaThon 2011 - Mining the AWR
Kristofferson A
 
PPT
Teradata 13.10
Teradata
 
Maris+Elsins+-+Mining+AWR+V2(1)11111.pdg
cookie1969
 
AWR DB performance Data Mining - Collaborate 2015
Yury Velikanov
 
OTN tour 2015 AWR data mining
Andrejs Vorobjovs
 
Awr doag
Marcin Przepiórowski
 
Whitepaper: Mining the AWR repository for Capacity Planning and Visualization
Kristofferson A
 
Donatone_Grid Performance(2)11111111.pdg
cookie1969
 
AWR and ASH in an EM12c World
Kellyn Pot'Vin-Gorman
 
Analyzing awr report
satish Gaddipati
 
AWR and ASH Deep Dive
Kellyn Pot'Vin-Gorman
 
Kellyn Pot'Vin-Gorman - Awr and Ash
gaougorg
 
Step by stepDoc for Oracle TuningsandAWR.docx
Ganesh Bhosale
 
Ebs dba con4696_pdf_4696_0001
jucaab
 
Predictive performance analysis using sql pattern matching
Horia Berca
 
ASH and AWR on DB12c
Kellyn Pot'Vin-Gorman
 
Oracle AWR Data mining
Yury Velikanov
 
Mining the AWR: Alternative Methods for Identification of the Top SQLs (inclu...
Maris Elsins
 
2013 Collaborate - OAUG - Presentation
Biju Thomas
 
Using AWR for SQL Analysis
Texas Memory Systems, and IBM Company
 
VirtaThon 2011 - Mining the AWR
Kristofferson A
 
Teradata 13.10
Teradata
 

More from Maris Elsins (15)

PDF
DB12c: All You Need to Know About the Resource Manager
Maris Elsins
 
PPTX
C15LV: Ins and Outs of Concurrent Processing Configuration in Oracle e-Busine...
Maris Elsins
 
PDF
Database as a Service on the Oracle Database Appliance Platform
Maris Elsins
 
PPTX
LVOUG meetup #2 - Forcing SQL Execution Plan Instability
Maris Elsins
 
PDF
OUG Harmony 2012 - Using SQL Plan Baselines for Performance Testing
Maris Elsins
 
PPTX
LVOUG meetup #4 - Case Study 10g to 11g
Maris Elsins
 
PPTX
Surviving the Crisis With the Help of Oracle Database Resource Manager
Maris Elsins
 
PPTX
Concurrent Processing Performance Analysis for Apps DBAs
Maris Elsins
 
PPTX
Simplify Consolidation with Oracle Database 12c
Maris Elsins
 
PPTX
10 ways to improve your rman script
Maris Elsins
 
PPTX
Wildcard13 - warmup slides for the "Roundtable discussion with Oracle Profess...
Maris Elsins
 
PDF
Whitepaper: Running Oracle e-Business Suite Database on Oracle Database Appli...
Maris Elsins
 
PDF
Running E-Business Suite Database on Oracle Database Appliance
Maris Elsins
 
PDF
Internals of concurent managers
Maris Elsins
 
PDF
Using SQL Plan Management for Performance Testing
Maris Elsins
 
DB12c: All You Need to Know About the Resource Manager
Maris Elsins
 
C15LV: Ins and Outs of Concurrent Processing Configuration in Oracle e-Busine...
Maris Elsins
 
Database as a Service on the Oracle Database Appliance Platform
Maris Elsins
 
LVOUG meetup #2 - Forcing SQL Execution Plan Instability
Maris Elsins
 
OUG Harmony 2012 - Using SQL Plan Baselines for Performance Testing
Maris Elsins
 
LVOUG meetup #4 - Case Study 10g to 11g
Maris Elsins
 
Surviving the Crisis With the Help of Oracle Database Resource Manager
Maris Elsins
 
Concurrent Processing Performance Analysis for Apps DBAs
Maris Elsins
 
Simplify Consolidation with Oracle Database 12c
Maris Elsins
 
10 ways to improve your rman script
Maris Elsins
 
Wildcard13 - warmup slides for the "Roundtable discussion with Oracle Profess...
Maris Elsins
 
Whitepaper: Running Oracle e-Business Suite Database on Oracle Database Appli...
Maris Elsins
 
Running E-Business Suite Database on Oracle Database Appliance
Maris Elsins
 
Internals of concurent managers
Maris Elsins
 
Using SQL Plan Management for Performance Testing
Maris Elsins
 

Recently uploaded (20)

PDF
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
SOFTTECHHUB
 
PPTX
Understanding_Digital_Forensics_Presentation.pptx
ImranKhan423233
 
PDF
Chapter 3 Spatial Domain Image Processing.pdf
Getnet Tigabie Askale -(GM)
 
PDF
NewMind AI Weekly Chronicles - August'25 Week I
NewMind AI
 
PPTX
MYSQL Presentation for SQL database connectivity
Swati270511
 
PDF
The Rise and Fall of 3GPP – Time for a Sabbatical?
3G4G
 
PDF
Electronic commerce courselecture one. Pdf
OmerMohamed64
 
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
Dropbox
 
PDF
Machine learning based COVID-19 study performance prediction
IAESIJAI
 
PDF
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
Akar16
 
PDF
Advanced methodologies resolving dimensionality complications for autism neur...
IAESIJAI
 
PPTX
PA Analog/Digital System: The Backbone of Modern Surveillance and Communication
AVTRON Technologies LLC
 
PDF
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
SynapseIndia
 
PPT
“AI and Expert System Decision Support & Business Intelligence Systems”
ZubinRadhakrishnan
 
PDF
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
Peter Tan
 
PDF
Approach and Philosophy of On baking technology
30anthuong17
 
PPTX
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
muhammadmuneebnaeem7
 
PDF
KodekX | Application Modernization Development
KodekX
 
PDF
Shreyas Phanse Resume: Experienced Backend Engineer | Java • Spring Boot • Ka...
SHREYAS PHANSE
 
PPTX
20250228 LYD VKU AI Blended-Learning.pptx
DatVoNgoc
 
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
SOFTTECHHUB
 
Understanding_Digital_Forensics_Presentation.pptx
ImranKhan423233
 
Chapter 3 Spatial Domain Image Processing.pdf
Getnet Tigabie Askale -(GM)
 
NewMind AI Weekly Chronicles - August'25 Week I
NewMind AI
 
MYSQL Presentation for SQL database connectivity
Swati270511
 
The Rise and Fall of 3GPP – Time for a Sabbatical?
3G4G
 
Electronic commerce courselecture one. Pdf
OmerMohamed64
 
Dropbox Q2 2025 Financial Results & Investor Presentation
Dropbox
 
Machine learning based COVID-19 study performance prediction
IAESIJAI
 
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
Akar16
 
Advanced methodologies resolving dimensionality complications for autism neur...
IAESIJAI
 
PA Analog/Digital System: The Backbone of Modern Surveillance and Communication
AVTRON Technologies LLC
 
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
SynapseIndia
 
“AI and Expert System Decision Support & Business Intelligence Systems”
ZubinRadhakrishnan
 
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
Peter Tan
 
Approach and Philosophy of On baking technology
30anthuong17
 
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
muhammadmuneebnaeem7
 
KodekX | Application Modernization Development
KodekX
 
Shreyas Phanse Resume: Experienced Backend Engineer | Java • Spring Boot • Ka...
SHREYAS PHANSE
 
20250228 LYD VKU AI Blended-Learning.pptx
DatVoNgoc
 

Mining AWR V2 - Trend Analysis

  • 1. © 2016 Pythian 1 Mining AWR V2 Trend Analysis MARIS ELSINS Lead Database Consultant Can you read the line below? Can you read the line below? This is the font size I use for listing code in slides!
  • 2. © 2016 Pythian 2 Maris Elsins Lead Database Consultant At Pythian since 2011 Located in Riga, Latvia Oracle [Apps] DBA since 2005 Speaker at conferences since 2007 @MarisDBA elsins@pythian.com http://bit.ly/getMOSPatchV2
  • 3. © 2016 Pythian 3 ABOUT PYTHIAN Pythian’s 400+ IT professionals help companies adopt and manage disruptive technologies to better compete
  • 4. © 2016 Pythian 4 Systems currently managed by Pythian EXPERIENCED Pythian experts in 35 countries GLOBAL Millennia of experience gathered and shared over 19 years EXPERTS 11,800 2400
  • 5. © 2016 Pythian 5 • “1-Dot” Philosophy • • “2-Dots” Philosophy • • • Data trends in AWR • Examples AGENDA
  • 6. © 2016 Pythian 6 LET'S OVERSIMPLIFY WHAT IS AN AWR REPORT? It's a dot ! •
  • 7. © 2016 Pythian 7 LET'S OVERSIMPLIFY SERIOUSLY? A DOT??? Are you gone totally nuts? Photo by herr_hartmann / CC BY-NC 2.0
  • 8. © 2016 Pythian 8 • AWR Report gives you valuable drill-down capabilities (time interval): – Top Wait Events – Top SQL Statements in different categories – Top Segments by Logical Reads – ... LET'S OVERSIMPLIFY AWR REPORT AS A "DOT"! THINK ABOUT IT!
  • 9. © 2016 Pythian 9 DOTS AND AWR - LET'S INVESTIGATE! HOW DID THE IO PERFORM ON SEPTEMBER 9? - AWR REPORT IS THE ANSWER!
  • 10. © 2016 Pythian 10 LET'S OVERSIMPLIFY AWR REPORT AS A "DOT"! THINK ABOUT IT! • Lot of resource usage / performance information is aggregated down to a single number (a "dot" on a time/value graph) – How much time was spent waiting on User I/O? = 42189 s – How quickly did the db file sequential reads respond? = 0.84 ms – How many transactions per second are completed? = 47.09
  • 11. © 2016 Pythian 11 LET'S OVERSIMPLIFY DB FILE SEQUENTIAL READS PERFORMANCE AS A "DOT" Time Resp. time 09-Sep-16 0.84 ms
  • 12. © 2016 Pythian 12 "The FBI Academy teaches new agents that the best predictor of future behavior is past behavior." Ronald Kessler Who doesn't want to be an FBI agent specializing in database performance analysis? A “dot” doesn't reveal the past behaviour WHY THIS IS IMPORTANT?
  • 13. © 2016 Pythian 13 1-DOT What does it tell us about how things develop over time? Time Value S1 V1
  • 14. © 2016 Pythian 14 DOTS AND AWR COMPARING TWO PERIODS OF TIME (AWR DIFF REPORT)
  • 15. © 2016 Pythian 15 What does it tell us about how things develop over time? 2-DOTS Time Value S1 V1 V2 S2
  • 16. © 2016 Pythian 16 2-DOTS What does it tell us about how things develop over time? Time Value S1 V1 V2 S2
  • 17. © 2016 Pythian 17 DOTS AND AWR • An AWR report describes the time interval between 2 snapshots i.e. – How much time was spent on CPU? – How many buffer gets did a particular query do? – How many Physical IOs did a top query do? – ... etc ... • A single aggregated value, that describes the reporting interval, which is insufficient for trend analysis
  • 18. © 2016 Pythian 18 IS IT WORTH MINING AWR? Trends in AWR
  • 19. © 2016 Pythian 19 MINING DATA TRENDS IN AWR • Each AWR snapshot represents a moment in time. • Multitude of statistics are saved for each snapshot • It's an obvious thing to graph these values! • And the best about it is you pick the metric to graph!
  • 20. © 2016 Pythian 20 SQL> @awr_wait_trend.sql "db file sequential read" 10 1 TIME (SNAP) EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- .... 09.09.2016 00:00:00 db file sequential read 55302 50.670 .916 09.09.2016 01:00:00 db file sequential read 45372 31.303 .690 09.09.2016 02:00:00 db file sequential read 80171 68.995 .861 09.09.2016 03:00:00 db file sequential read 215704 235.555 1.092 09.09.2016 04:00:00 db file sequential read 67104 46.547 .694 09.09.2016 05:00:00 db file sequential read 172241 104.396 .606 09.09.2016 06:00:00 db file sequential read 65497 36.598 .559 09.09.2016 07:00:00 db file sequential read 76487 46.231 .604 09.09.2016 08:00:00 db file sequential read 53193 27.399 .515 09.09.2016 09:00:00 db file sequential read 78568 45.846 .584 09.09.2016 10:00:00 db file sequential read 88612 48.192 .544 09.09.2016 11:00:00 db file sequential read 71483 42.597 .596 09.09.2016 12:00:00 db file sequential read 79741 53.432 .670 09.09.2016 13:00:00 db file sequential read 89707 65.399 .729 09.09.2016 14:00:00 db file sequential read 85208 49.717 .583 09.09.2016 15:00:00 db file sequential read 125878 99.274 .789 09.09.2016 16:00:00 db file sequential read 71224 39.754 .558 09.09.2016 17:00:00 db file sequential read 78324 46.832 .598 09.09.2016 18:00:00 db file sequential read 72536 67.021 .924 09.09.2016 19:00:00 db file sequential read 49199 37.167 .755 09.09.2016 20:00:00 db file sequential read 47840 26.727 .559 ... DOTS AND AWR HOW DID THE IO PERFORM ON SEPTEMBER 9?
  • 21. © 2016 Pythian 21 DOTS AND AWR HOW DID THE IO PERFORM ON SEPTEMBER 9?
  • 22. © 2016 Pythian 22 DOTS AND AWR DO YOU STILL REMEMBER THIS PICTURE?
  • 23. © 2016 Pythian 23 DOTS AND AWR HOW DID THE IO PERFORM BETWEEN SEPTEMBER 2 AND 9?
  • 24. © 2016 Pythian 24 DOTS AND AWR TOTAL WAITS COMPARISON BETWEEN SEPTEMBER 2 AND 9?
  • 25. © 2016 Pythian 25 DOTS AND AWR RESPONSE TIME COMPARISON BETWEEN SEPTEMBER 2 AND 9?
  • 26. © 2016 Pythian 26 HOW DID I DO IT? Extracting the “Past Behaviour” from AWR
  • 27. © 2016 Pythian 27 Information on the DBA_HIST_views can be found in Oracle® Database Reference 12c Release 2 (12.2) http://docs.oracle.com/database/122/REFRN/toc.htm SELECT v.snap_id, v.value FROM dba_hist_sysstat v WHERE v.stat_name = 'user commits' ORDER BY v.snap_id; SNAP_ID VALUE ---------- ---------- 12621 625766 12622 626608 12623 627853 12624 629348 12625 630130 12626 630904 .... HOW DID I DO IT? STEP 1: PICK THE METRIC AND FIND THE CORRECT DBA_HIST_% VIEW
  • 28. © 2016 Pythian 28 SELECT Lag(snap_id) over (ORDER BY v.snap_id) from_snap, v.snap_id to_snap, v.value - Lag(v.value) over (ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v WHERE v.stat_name = 'user commits' ORDER BY v.snap_id; FROM_SNAP TO_SNAP DELTA_VALUE ---------- ---------- ----------- 12621 12621 12622 842 12622 12623 1245 12623 12624 1495 12624 12625 782 12625 12626 774 ... HOW DID I DO IT? STEP 2: CALCULATE DELTAS
  • 29. © 2016 Pythian 29 SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id ORDER BY s.end_interval_time; FROM_TIME TO_TIME DELTA_VALUE ------------------------------- ------------------------------- ----------- 09-SEP-16 04.00.44.539000000 PM 09-SEP-16 04.00.44.539000000 PM 09-SEP-16 05.00.58.620000000 PM 842 09-SEP-16 05.00.58.620000000 PM 09-SEP-16 06.00.22.386000000 PM 1245 09-SEP-16 06.00.22.386000000 PM 09-SEP-16 07.00.35.070000000 PM 1495 09-SEP-16 07.00.35.070000000 PM 09-SEP-16 08.00.45.075000000 PM 782 09-SEP-16 08.00.45.075000000 PM 09-SEP-16 09.00.54.573000000 PM 774 09-SEP-16 09.00.54.573000000 PM 09-SEP-16 10.00.04.401000000 PM 768 ... HOW DID I DO IT? STEP 3: VIEWER FRIENDLY DATES
  • 30. © 2016 Pythian 30 HOW DID I DO IT? STEP 4: TIMELINE FILTERING SELECT * FROM (SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id) WHERE from_time BETWEEN To_date('11092016', 'DDMMYYYY') AND To_date('16092016', 'DDMMYYYY') ORDER BY from_time; FROM_TIME TO_TIME DELTA_VALUE ------------------------------- ------------------------------- ----------- 11-SEP-16 12.00.05.638000000 AM 11-SEP-16 01.00.23.855000000 AM 785 11-SEP-16 01.00.23.855000000 AM 11-SEP-16 02.00.40.073000000 AM 1057 11-SEP-16 02.00.40.073000000 AM 11-SEP-16 03.00.46.109000000 AM 1096 ... 15-SEP-16 09.00.04.192000000 PM 15-SEP-16 10.00.16.087000000 PM 786 15-SEP-16 10.00.16.087000000 PM 15-SEP-16 11.00.35.078000000 PM 779 15-SEP-16 11.00.35.078000000 PM 16-SEP-16 12.00.51.092000000 AM 831
  • 31. © 2016 Pythian 31 HOW DID I DO IT? STEP 5: AGGREGATION SELECT Trunc(from_time, 'DD') t_from_time, SUM(delta_value) sum_delta_value FROM (SELECT Lag(s.end_interval_time) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) from_time, s.end_interval_time to_time, v.value - Lag(v.value) over (PARTITION BY s.dbid, s.startup_time ORDER BY v.snap_id) delta_value FROM dba_hist_sysstat v, dba_hist_snapshot s WHERE v.stat_name = 'user commits' AND s.snap_id = v.snap_id) WHERE from_time BETWEEN To_date('11092016', 'DDMMYYYY') AND To_date('16092016', 'DDMMYYYY') GROUP BY Trunc(from_time, 'DD') ORDER BY t_from_time; T_FROM_TIME SUM_DELTA_VALUE ------------------------------- --------------- 11.09.2016 00:00:00 21602 12.09.2016 00:00:00 29173 13.09.2016 00:00:00 32772 14.09.2016 00:00:00 30477 15.09.2016 00:00:00 26799
  • 32. © 2016 Pythian 32 HOW DO I DO IT? STEP 6: VISUALIZATION
  • 33. © 2016 Pythian 33 EXAMPLE 1 SQL Performance
  • 34. © 2016 Pythian 34 EXAMPLE 1 THE PROBLEM • Amazon RDS for Oracle – db.m3.xlarge (4 vCPU, 15GiB) – 11.2.0.4 • Alert from the client – Could you please take a look at the DB utilization in prod? We're seeing 100% utilization and 795 connections. Could you please let us know if any specific connection is causing high CPU utilization.
  • 35. © 2016 Pythian 35 EXAMPLE 1 5MIN ASH 10:21:04 SQL> @ashtop session_id,sql_id "event is null" sysdate-5/25/60 sysdate Total Seconds AAS %This SESSION_ID SQL_ID FIRST_SEEN LAST_SEEN ... --------- ------- ------- ---------- ------------- ------------------- ------------------- ... 285 1.0 12% | 738 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1336 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1358 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 745 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 882 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 2046 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 99 .3 4% | 1472 grbz54xqkr425 2016-06-06 10:17:04 2016-06-06 10:18:43 ... 64 .2 3% | 1472 fy4c407vamaks 2016-06-06 10:18:52 2016-06-06 10:21:26 ... 37 .1 2% | 2176 2016-06-06 10:18:29 2016-06-06 10:19:50 ... 22 .1 1% | 977 7tdudtm4x03np 2016-06-06 10:16:40 2016-06-06 10:17:01 ... 16 .1 1% | 795 3b8s69ct88pur 2016-06-06 10:16:40 2016-06-06 10:18:13 ... 15 .1 1% | 989 2016-06-06 10:17:04 2016-06-06 10:17:19 ... 13 .0 1% | 632 gdhy97b59bk81 2016-06-06 10:16:46 2016-06-06 10:17:07 ... 11 .0 0% | 1792 2016-06-06 10:16:59 2016-06-06 10:21:09 ... 11 .0 0% | 2133 atmhpvm759a97 2016-06-06 10:16:50 2016-06-06 10:21:02 ... Thank you, Tanel! http://blog.tanelpoder.com/files/scripts/ash/ashtop.sql
  • 36. © 2016 Pythian 36 EXAMPLE 1 5MIN ASH 10:21:04 SQL> @ashtop session_id,sql_id "event is null" sysdate-5/25/60 sysdate Total Seconds AAS %This SESSION_ID SQL_ID FIRST_SEEN LAST_SEEN ... --------- ------- ------- ---------- ------------- ------------------- ------------------- ... 285 1.0 12% | 738 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1336 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 285 1.0 12% | 1358 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 745 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 882 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 284 1.0 12% | 2046 4db73upm43ck1 2016-06-06 10:16:40 2016-06-06 10:21:27 ... 99 .3 4% | 1472 grbz54xqkr425 2016-06-06 10:17:04 2016-06-06 10:18:43 ... 64 .2 3% | 1472 fy4c407vamaks 2016-06-06 10:18:52 2016-06-06 10:21:26 ... 37 .1 2% | 2176 2016-06-06 10:18:29 2016-06-06 10:19:50 ... 22 .1 1% | 977 7tdudtm4x03np 2016-06-06 10:16:40 2016-06-06 10:17:01 ... 16 .1 1% | 795 3b8s69ct88pur 2016-06-06 10:16:40 2016-06-06 10:18:13 ... 15 .1 1% | 989 2016-06-06 10:17:04 2016-06-06 10:17:19 ... 13 .0 1% | 632 gdhy97b59bk81 2016-06-06 10:16:46 2016-06-06 10:17:07 ... 11 .0 0% | 1792 2016-06-06 10:16:59 2016-06-06 10:21:09 ... 11 .0 0% | 2133 atmhpvm759a97 2016-06-06 10:16:50 2016-06-06 10:21:02 ... Thank you, Tanel! http://blog.tanelpoder.com/files/scripts/ash/ashtop.sql
  • 37. © 2016 Pythian 37 Just one query to address! How hard can it be? EXAMPLE 1 SQL PERFORMANCE
  • 38. © 2016 Pythian 38 EXAMPLE 1 THE EXECUTION PLAN 10:26:35 SQL> select * from table(dbms_xplan.display_awr('4db73upm43ck1',82218884)); Plan hash value: 82218884 -------------------------------------------------------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time | -------------------------------------------------------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | | | | 54163 (100)| | | 1 | TEMP TABLE TRANSFORMATION | | | | | | | | 2 | LOAD AS SELECT | | | | | | | | 3 | VIEW | | 179K| 13M| | 10337 (1)| 00:02:05 | | 4 | HASH JOIN | | 179K| 9830K| 5800K| 10337 (1)| 00:02:05 | | 5 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 3686K| | 794 (1)| 00:00:10 | | 6 | TABLE ACCESS FULL | ACTIVE_USER_VIEWS_BY_TOPIC | 1543K| 51M| | 5823 (1)| 00:01:10 | | 7 | VIEW | | 1 | 13 | | 2 (0)| 00:00:01 | | 8 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 9 | LOAD AS SELECT | | | | | | | | 10 | VIEW | | 110K| 8415K| | 10228 (1)| 00:02:03 | | 11 | HASH JOIN | | 110K| 6042K| 3568K| 10228 (1)| 00:02:03 | | 12 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 2265K| | 794 (1)| 00:00:10 | | 13 | TABLE ACCESS FULL | ACTIVE_USER_VIEWS_BY_TOPIC | 1543K| 51M| | 5823 (1)| 00:01:10 | | 14 | VIEW | | 1 | 13 | | 2 (0)| 00:00:01 | | 15 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | ... 1
  • 39. © 2016 Pythian 39 EXAMPLE 1 THE EXECUTION PLAN ... | 16 | LOAD AS SELECT | | | | | | | | 17 | INTERSECTION | | | | | | | | 18 | SORT UNIQUE | | 179K| 7021K| 8472K| 2359 (1)| 00:00:29 | | 19 | VIEW | | 179K| 7021K| | 518 (1)| 00:00:07 | | 20 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 21 | SORT UNIQUE | | 110K| 4315K| 5208K| 1452 (1)| 00:00:18 | | 22 | VIEW | | 110K| 4315K| | 319 (1)| 00:00:04 | | 23 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 24 | LOAD AS SELECT | | | | | | | | 25 | WINDOW BUFFER | | 1 | 27 | | 860 (1)| 00:00:11 | | 26 | HASH GROUP BY | | 1 | 27 | | 860 (1)| 00:00:11 | | 27 | VIEW | | 110K| 2913K| | 856 (1)| 00:00:11 | | 28 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 110K| 4315K| | 856 (1)| 00:00:11 | | 29 | LOAD AS SELECT | | | | | | | | 30 | WINDOW BUFFER | | 42 | 756 | | 523 (2)| 00:00:07 | | 31 | HASH GROUP BY | | 42 | 756 | | 523 (2)| 00:00:07 | | 32 | VIEW | | 179K| 3159K| | 518 (1)| 00:00:07 | | 33 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 34 | LOAD AS SELECT | | | | | | | | 35 | WINDOW BUFFER | | 42 | 756 | | 322 (2)| 00:00:04 | | 36 | HASH GROUP BY | | 42 | 756 | | 322 (2)| 00:00:04 | | 37 | VIEW | | 110K| 1942K| | 319 (1)| 00:00:04 | ... 2
  • 40. © 2016 Pythian 40 EXAMPLE 1 THE EXECUTION PLAN ... | 38 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 39 | FILTER | | | | | | | | 40 | SORT GROUP BY ROLLUP | | 1 | 579 | | 28077 (1)| 00:05:37 | | 41 | MERGE JOIN OUTER | | 10 | 5790 | | 28076 (1)| 00:05:37 | | 42 | MERGE JOIN OUTER | | 10 | 5790 | | 20379 (1)| 00:04:05 | | 43 | HASH JOIN OUTER | | 10 | 5660 | | 11818 (1)| 00:02:22 | | 44 | HASH JOIN OUTER | | 1 | 526 | | 11705 (1)| 00:02:21 | | 45 | HASH JOIN OUTER | | 1 | 495 | | 11383 (1)| 00:02:17 | | 46 | HASH JOIN OUTER | | 1 | 464 | | 10860 (1)| 00:02:11 | | 47 | HASH JOIN OUTER | | 1 | 407 | | 6041 (1)| 00:01:13 | | 48 | HASH JOIN OUTER | | 1 | 389 | | 6039 (1)| 00:01:13 | | 49 | HASH JOIN OUTER | | 1 | 371 | | 6037 (1)| 00:01:13 | | 50 | HASH JOIN OUTER | | 1 | 327 | | 5266 (1)| 00:01:04 | | 51 | HASH JOIN OUTER | | 1 | 296 | | 1603 (1)| 00:00:20 | | 52 | VIEW | | 1 | 148 | | 802 (1)| 00:00:10 | | 53 | HASH GROUP BY | | 1 | 75 | | 802 (1)| 00:00:10 | | 54 | HASH JOIN OUTER | | 9 | 675 | | 801 (1)| 00:00:10 | | 55 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 7 | 217 | | 8 (0)| 00:00:01 | | 56 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 5 (0)| 00:00:01 | | 57 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 58 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 7723K| | 792 (1)| 00:00:10 | | 59 | VIEW | | 1 | 148 | | 801 (1)| 00:00:10 | ... 3
  • 41. © 2016 Pythian 41 EXAMPLE 1 THE EXECUTION PLAN ... | 60 | HASH GROUP BY | | 1 | 75 | | 801 (1)| 00:00:10 | | 61 | HASH JOIN OUTER | | 8 | 600 | | 800 (1)| 00:00:10 | | 62 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 7 | 217 | | 8 (0)| 00:00:01 | | 63 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 5 (0)| 00:00:01 | | 64 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 65 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 4747K| | 792 (1)| 00:00:10 | | 66 | VIEW | | 1 | 31 | | 3663 (1)| 00:00:44 | | 67 | HASH JOIN | | 1 | 62 | | 3663 (1)| 00:00:44 | | 68 | VIEW | | 1 | 31 | | 1658 (1)| 00:00:20 | | 69 | HASH GROUP BY | | 1 | 96 | | 1658 (1)| 00:00:20 | | 70 | HASH JOIN | | 1 | 96 | | 1657 (1)| 00:00:20 | | 71 | NESTED LOOPS | | 1 | 75 | | 863 (1)| 00:00:11 | | 72 | NESTED LOOPS | | 7 | 75 | | 863 (1)| 00:00:11 | | 73 | VIEW | | 1 | 40 | | 856 (1)| 00:00:11 | | 74 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 1 | 40 | | 856 (1)| 00:00:11 | | 75 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 7 | | | 4 (0)| 00:00:01 | | 76 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 77 | TABLE ACCESS BY INDEX ROWID| ACTIVE_USER_VIEWS_BY_TOPIC | 1 | 35 | | 7 (0)| 00:00:01 | | 78 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 179K| 3686K| | 794 (1)| 00:00:10 | | 79 | VIEW | | 1 | 31 | | 2004 (1)| 00:00:25 | | 80 | HASH GROUP BY | | 1 | 96 | | 2004 (1)| 00:00:25 | | 81 | HASH JOIN | | 1 | 96 | | 2003 (1)| 00:00:25 | ... 4
  • 42. © 2016 Pythian 42 EXAMPLE 1 THE EXECUTION PLAN ... | 82 | NESTED LOOPS | | 1 | 75 | | 1209 (1)| 00:00:15 | | 83 | NESTED LOOPS | | 987 | 75 | | 1209 (1)| 00:00:15 | | 84 | VIEW | | 1 | 40 | | 856 (1)| 00:00:11 | | 85 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 1 | 40 | | 856 (1)| 00:00:11 | | 86 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 4 (0)| 00:00:01 | | 87 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 88 | TABLE ACCESS BY INDEX ROWID| ACTIVE_USER_VIEWS_BY_TOPIC | 1 | 35 | | 353 (0)| 00:00:05 | | 89 | TABLE ACCESS FULL | LEARNING_IMPACT_BY_TOPIC | 110K| 2265K| | 794 (1)| 00:00:10 | | 90 | VIEW | | 41 | 1804 | | 771 (1)| 00:00:10 | | 91 | HASH GROUP BY | | 41 | 1886 | | 771 (1)| 00:00:10 | | 92 | HASH JOIN | | 157 | 7222 | | 770 (1)| 00:00:10 | | 93 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 157 | 5338 | | 354 (0)| 00:00:05 | | 94 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 5 (0)| 00:00:01 | | 95 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 96 | TABLE ACCESS FULL | ASSESSMENTS | 270K| 3166K| | 415 (1)| 00:00:05 | | 97 | VIEW | | 42 | 756 | | 2 (0)| 00:00:01 | | 98 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 99 | VIEW | | 42 | 756 | | 2 (0)| 00:00:01 | | 100 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 101 | VIEW | | 42 | 2394 | | 4819 (1)| 00:00:58 | | 102 | HASH JOIN OUTER | | 42 | 6678 | | 4819 (1)| 00:00:58 | | 103 | HASH JOIN OUTER | | 42 | 5376 | | 4816 (1)| 00:00:58 | ... 5
  • 43. © 2016 Pythian 43 EXAMPLE 1 THE EXECUTION PLAN ... | 104 | HASH JOIN OUTER | | 42 | 4074 | | 4813 (1)| 00:00:58 | | 105 | MERGE JOIN OUTER | | 42 | 2394 | | 4698 (1)| 00:00:57 | | 106 | VIEW | VW_FOJ_0 | 42 | 1848 | | 4 (0)| 00:00:01 | | 107 | HASH JOIN FULL OUTER | | 42 | 2604 | | 4 (0)| 00:00:01 | | 108 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 109 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 110 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 111 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE02_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 112 | BUFFER SORT | | 1 | 13 | | 4698 (1)| 00:00:57 | | 113 | VIEW | | 1 | 13 | | 112 (1)| 00:00:02 | | 114 | VIEW | | 1 | 13 | | 112 (1)| 00:00:02 | | 115 | SORT AGGREGATE | | 1 | 13 | | | | | 116 | VIEW | | 110K| 1402K| | 112 (1)| 00:00:02 | | 117 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 118 | VIEW | | 1 | 40 | | 115 (4)| 00:00:02 | | 119 | WINDOW BUFFER | | 1 | 40 | | 115 (4)| 00:00:02 | | 120 | HASH GROUP BY | | 1 | 40 | | 115 (4)| 00:00:02 | | 121 | VIEW | | 110K| 4315K| | 112 (1)| 00:00:02 | | 122 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 123 | VIEW | | 30 | 930 | | 3 (34)| 00:00:01 | | 124 | WINDOW BUFFER | | 30 | 930 | | 3 (34)| 00:00:01 | | 125 | HASH GROUP BY | | 30 | 930 | | 3 (34)| 00:00:01 | ... 6
  • 44. © 2016 Pythian 44 EXAMPLE 1 THE EXECUTION PLAN ... | 126 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 127 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE01_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 128 | VIEW | | 30 | 930 | | 3 (34)| 00:00:01 | | 129 | WINDOW BUFFER | | 30 | 930 | | 3 (34)| 00:00:01 | | 130 | HASH GROUP BY | | 30 | 930 | | 3 (34)| 00:00:01 | | 131 | VIEW | | 42 | 1302 | | 2 (0)| 00:00:01 | | 132 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE02_AAF67829 | 42 | 756 | | 2 (0)| 00:00:01 | | 133 | VIEW | | 42 | 1302 | | 523 (2)| 00:00:07 | | 134 | HASH GROUP BY | | 42 | 756 | | 523 (2)| 00:00:07 | | 135 | VIEW | | 179K| 3159K| | 518 (1)| 00:00:07 | | 136 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFD_AAF67829 | 179K| 4388K| | 518 (1)| 00:00:07 | | 137 | VIEW | | 42 | 1302 | | 322 (2)| 00:00:04 | | 138 | HASH GROUP BY | | 42 | 756 | | 322 (2)| 00:00:04 | | 139 | VIEW | | 110K| 1942K| | 319 (1)| 00:00:04 | | 140 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFE_AAF67829 | 110K| 2697K| | 319 (1)| 00:00:04 | | 141 | VIEW | | 110K| 4315K| | 112 (1)| 00:00:02 | | 142 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBE00_AAF67829 | 110K| 2913K| | 112 (1)| 00:00:02 | | 143 | BUFFER SORT | | 1 | 13 | | 20379 (1)| 00:04:05 | | 144 | VIEW | | 1 | 13 | | 856 (1)| 00:00:11 | | 145 | VIEW | | 1 | 13 | | 856 (1)| 00:00:11 | | 146 | SORT AGGREGATE | | 1 | | | | | | 147 | VIEW | | 110K| | | 856 (1)| 00:00:11 | ... 7
  • 45. © 2016 Pythian 45 ... | 148 | TABLE ACCESS FULL | SYS_TEMP_0FD9DBDFF_AAF67829 | 110K| 4315K| | 856 (1)| 00:00:11 | | 149 | BUFFER SORT | | 1 | | | 27221 (1)| 00:05:27 | | 150 | VIEW | | 1 | | | 770 (1)| 00:00:10 | | 151 | VIEW | | 1 | | | 770 (1)| 00:00:10 | | 152 | SORT AGGREGATE | | 1 | 25 | | | | | 153 | HASH JOIN | | 157 | 3925 | | 770 (1)| 00:00:10 | | 154 | TABLE ACCESS BY INDEX ROWID | ACTIVE_USER_VIEWS_BY_TOPIC | 157 | 2512 | | 354 (0)| 00:00:05 | | 155 | INDEX RANGE SCAN | AUVBT_BY_PPID1_IX1 | 987 | | | 5 (0)| 00:00:01 | | 156 | FAST DUAL | | 1 | | | 2 (0)| 00:00:01 | | 157 | TABLE ACCESS FULL | ASSESSMENTS | 270K| 2374K| | 415 (1)| 00:00:05 | -------------------------------------------------------------------------------------------------------------------------------------- EXAMPLE 1 THE EXECUTION PLAN 8
  • 46. © 2016 Pythian 46 EXAMPLE 1 THE EXECUTION PLAN - #SHTF
  • 47. © 2016 Pythian 47 EXAMPLE 1 THE OPTIONS • What can one do in this case? – Tune the query? – Check / collect fresh statistics? – Flush the shared pool? (hoping it's bind variable peeking) – Run SQL Tuning Advisor? – or... • How did the SQL statement execute in the past?
  • 48. © 2016 Pythian 48 EXAMPLE 1 AWR_SQLID_PERF_TREND_BY_PLAN.SQL select hss.instance_number inst, to_char(trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24,'dd.mm.yyyy hh24:mi:ss') time, plan_hash_value, sum(hss.executions_delta) executions, round(sum(hss.elapsed_time_delta)/1000000,3) elapsed_time_s, round(sum(hss.cpu_time_delta)/1000000,3) cpu_time_s, round(sum(hss.iowait_delta)/1000000,3) iowait_s, round(sum(hss.clwait_delta)/1000000,3) clwait_s, round(sum(hss.apwait_delta)/1000000,3) apwait_s, round(sum(hss.ccwait_delta)/1000000,3) ccwait_s, round(sum(hss.rows_processed_delta),3) rows_processed, round(sum(hss.buffer_gets_delta),3) buffer_gets, round(sum(hss.disk_reads_delta),3) disk_reads, round(sum(hss.direct_writes_delta),3) direct_writes from dba_hist_sqlstat hss, dba_hist_snapshot hs where hss.sql_id='&sql_id' and hss.snap_id=hs.snap_id and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 group by hss.instance_number, trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24, plan_hash_value having sum(hss.executions_delta)>0
  • 49. © 2016 Pythian 49 10:21:28 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 5 24 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- 02.06.2016 00:00:00 1597670781 291 278.018 254.272 22.717 19787274.000 50740.000 03.06.2016 00:00:00 1597670781 205 204.755 189.479 14.583 14986545.000 71451.000 04.06.2016 00:00:00 1597670781 3 8.275 4.483 3.936 314594.000 15790.000 04.06.2016 00:00:00 4142332636 2 6.149 3.933 2.307 119766.000 47343.000 05.06.2016 00:00:00 82218884 47 11805.808 5766.836 22.799 1247372099.000 173254.000 06.06.2016 00:00:00 82218884 127 7216.393 3751.640 9.431 522088617.000 24677.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE LOOKING AT 5 DAYS OF PERFORMANCE DATA AT ONCE!
  • 50. © 2016 Pythian 50 10:21:28 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 5 24 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- 02.06.2016 00:00:00 1597670781 291 278.018 254.272 22.717 19787274.000 50740.000 03.06.2016 00:00:00 1597670781 205 204.755 189.479 14.583 14986545.000 71451.000 04.06.2016 00:00:00 1597670781 3 8.275 4.483 3.936 314594.000 15790.000 04.06.2016 00:00:00 4142332636 2 6.149 3.933 2.307 119766.000 47343.000 05.06.2016 00:00:00 82218884 47 11805.808 5766.836 22.799 1247372099.000 173254.000 06.06.2016 00:00:00 82218884 127 7216.393 3751.640 9.431 522088617.000 24677.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE LOOKING AT 5 DAYS OF PERFORMANCE DATA AT ONCE!
  • 51. © 2016 Pythian 51 EXAMPLE 1 SQL PLAN BASELINE FROM AWR exec DBMS_SQLTUNE.CREATE_SQLSET(sqlset_name => 'CR1064802', description => 'Plan for sql_id 4db73upm43ck1'); DECLARE cur sys_refcursor; BEGIN OPEN cur FOR SELECT VALUE(P) FROM table(dbms_sqltune.select_workload_repository(11434,12880,'sql_id=''4db73upm43ck1'' and plan_hash_value=1597670781' , NULL, NULL, NULL, NULL, NULL, NULL, 'ALL')) P; DBMS_SQLTUNE.LOAD_SQLSET(load_option=>'MERGE',sqlset_name => 'CR1064802', populate_cursor => cur); CLOSE cur; END; VARIABLE cnt NUMBER EXECUTE :cnt := DBMS_SPM.LOAD_PLANS_FROM_SQLSET( sqlset_name => 'CR1064802', basic_filter => 'sql_id=''4db73upm43ck1'''); comment on table cl_report_owner.ACTIVE_USER_VIEWS_BY_TOPIC is '';
  • 52. © 2016 Pythian 52 11:03:11 SQL> @awr_sqlid_perf_trend_by_plan.sql 4db73upm43ck1 2 1 TIME PLAN_HASH_VALUE EXECUTIONS ELAPSED_TIME_S CPU_TIME_S IOWAIT_S BUFFER_GETS DISK_READS ------------------- --------------- ----------- -------------- ------------ ------------ ----------------- ----------------- ... 06.06.2016 00:00:00 82218884 3 4.600 4.019 .608 1425936.000 965.000 06.06.2016 01:00:00 82218884 4 14.374 11.740 2.756 4726784.000 12836.000 06.06.2016 02:00:00 82218884 20 34.437 33.125 .619 7475353.000 264.000 06.06.2016 03:00:00 82218884 3 4.527 4.485 .032 1205412.000 14.000 06.06.2016 04:00:00 82218884 18 13.543 13.366 .132 1346631.000 57.000 06.06.2016 05:00:00 82218884 22 43.248 40.954 2.217 14290668.000 4439.000 06.06.2016 06:00:00 82218884 11 78.463 77.897 .311 45601072.000 79.000 06.06.2016 07:00:00 82218884 2 3.338 2.155 1.237 165446.000 3369.000 06.06.2016 08:00:00 82218884 10 9.241 9.098 .098 900049.000 61.000 06.06.2016 09:00:00 82218884 34 7010.622 3554.801 1.422 444951266.000 2593.000 06.06.2016 10:00:00 82218884 15 10942.980 4653.979 4.637 586951399.000 2078.000 06.06.2016 10:00:00 1597670781 1 7.910 2.047 .323 117577.000 28.000 06.06.2016 11:00:00 1597670781 1 1.028 1.008 .020 117468.000 28.000 A few columns have been removed from the outputs to improve the readability! EXAMPLE 1 SQL PERFORMANCE Yea, but ... ... what if the good plan is not captured in the AWR?
  • 53. © 2016 Pythian 53 -- Create the procedure that will be used to collect the execution plans create or replace procedure PERF.XXSTABILITY_CAPTURE_PLANS is cur sys_refcursor; cursor clist is with raw_sqlids as (select sql_id from v$sqlarea where sql_id in (select sql_id from v$active_session_history where sample_time>=sysdate-1/24 and sql_plan_hash_value>0 and machine like '%-hesvc-app0%') and plan_hash_value>0 and executions>2 union select sql_id from v$sqlarea where sql_id in (select sql_id from v$open_cursor where sid in (select sid from v$session where machine like '%-hesvc-app0%')) and plan_hash_value>0 and executions>2 and parsing_schema_name not in ('ANONYMOUS','APEX_030200','APEX_040000','APEX_SSO','APPQOSSYS', 'CTXSYS','DBSNMP','DIP','EXFSYS','FLOWS_FILES','MDSYS','OLAPSYS','ORACLE_OCM','ORDDATA','ORDPLUGINS','ORDSYS','OUTLN','OWBSYS ','SI_INFORMTN_SCHEMA','SQLTXADMIN','SQLTXPLAIN','SYS','SYSMAN','SYSTEM','TRCANLZR','WMSYS','XDB','XS$NULL')), sqlids as (select distinct ''''||sql_id||'''' sql_id, trunc((rownum-1)/200) rn from raw_sqlids) select listagg(s.sql_id,',') within group (order by s.sql_id) as sql_id_filter from sqlids s group by s.rn; c clist%rowtype; BEGIN for c in clist loop OPEN cur FOR SELECT VALUE(P) FROM table(dbms_sqltune.select_cursor_cache(basic_filter=>'sql_id in('||c.sql_id_filter||')', attribute_list=>'ALL')) P; DBMS_SQLTUNE.LOAD_SQLSET(load_option=>'MERGE',sqlset_name => 'XXSTABILITY_EXECUTION_PLANS', populate_cursor => cur); CLOSE cur; end loop; Commit; END; / EXAMPLE 1 PRESERVING GOOD EXECUTION PLANS
  • 54. © 2016 Pythian 54 -- Create the Scheduler Job which will run on every hour's 47th minute Begin Dbms_scheduler.create_job (job_name => 'COLLECT_XXSTABILITY_PLANS', job_type => 'STORED_PROCEDURE', job_action => 'PERF.XXSTABILITY_CAPTURE_PLANS', start_date => SYSDATE, repeat_interval => 'FREQ=HOURLY; INTERVAL=1; BYMINUTE=47;', enabled => TRUE); End; / -- Loading the baseline from SELECT_SQLSET for a specific sql_id and plan VARIABLE cnt NUMBER EXECUTE :cnt := DBMS_SPM.LOAD_PLANS_FROM_SQLSET( - sqlset_name => 'XXSTABILITY_EXECUTION_PLANS', - basic_filter => 'sql_id=''dtzsc12fbbjk2'' and plan_hash_value=2896891279'); EXAMPLE 1 PRESERVING EXECUTION PLANS
  • 55. © 2016 Pythian 55 EXAMPLE 2 IO performance
  • 56. © 2016 Pythian 56 • Oracle e-Business Suite – R12.2 – 12.1.0.2 • Alert from the client – Starting from September 5th, everything has become a lot more slower. Entering a sales order can take up to 15 minutes instead of usual 15 seconds. • Not enough detail to look at any specific process or query EXAMPLE 2 THE PROBLEM
  • 57. © 2016 Pythian 57 select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, stat_name, sum(delta_value) value_delta from (select hss.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, stat_name, value, value-(lag(value,1) over(partition by hs.startup_time, stat_name order by hss.snap_id)) delta_value from dba_hist_sys_time_model hss, dba_hist_snapshot hs where hss.snap_id=hs.snap_id and hss.instance_number=hs.instance_number and hs.end_interval_time>=trunc(sysdate)-&days_history+1 and hss.stat_name like '&stat_name') group by time, stat_name order by to_date(time,'DD.MM.YYYY HH24:MI:SS'), 1; EXAMPLE 2 AWR_SYS_TIME_MODEL_TREND.SQL
  • 58. © 2016 Pythian 58 EXAMPLE 2 CPU OR WAITING? SQL> @awr_sys_time_model_trend.sql "DB CPU" 14 24 TIME STAT_NAME VALUE_DELTA ------------------- ---------------------------------------------------------------- ----------------- 28.08.2016 00:00:00 DB CPU 204502770965 29.08.2016 00:00:00 DB CPU 240124304792 30.08.2016 00:00:00 DB CPU 243171465732 31.08.2016 00:00:00 DB CPU 271839933025 01.09.2016 00:00:00 DB CPU 398156022996 02.09.2016 00:00:00 DB CPU 343376689897 03.09.2016 00:00:00 DB CPU 262729582269 04.09.2016 00:00:00 DB CPU 79584869551 05.09.2016 00:00:00 DB CPU 87671787128 06.09.2016 00:00:00 DB CPU 103383834430 07.09.2016 00:00:00 DB CPU 16642806685 08.09.2016 00:00:00 DB CPU 215983418676 09.09.2016 00:00:00 DB CPU 100324311282 10.09.2016 00:00:00 DB CPU 11631538710 We're spending less time on CPU, so it's likely we're spending more time somewhere else!
  • 59. © 2016 Pythian 59 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES) select * from ( select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, wait_class, sum(delta_time_waited)/1000000 wait_time_s from (select hse.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, EVENT_NAME, WAIT_CLASS, TIME_WAITED_MICRO-(lag(TIME_WAITED_MICRO,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_time_waited from DBA_HIST_SYSTEM_EVENT hse, DBA_HIST_SNAPSHOT hs where hse.snap_id=hs.snap_id and hs.end_interval_time>=trunc(sysdate)-&days_history+1) group by time, event_name, wait_class) pivot ( sum(wait_time_s) for wait_class in ('Administrative','Application','Cluster' ,'Commit','Concurrency','Configuration' ,'Network','Other','Queueing' ,'Scheduler','System I/O','User I/O','Idle') ) order by to_date(time,'DD.MM.YYYY HH24:MI:SS');
  • 60. © 2016 Pythian 60 SQL> @awr_wait_class_hist.sql 14 1 TIME 'Administrative' 'Application' 'Commit' 'Concurrency' 'Other' 'Scheduler' 'System I/O' 'User I/O' 'Idle' ------------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- 28.08.2016 01:00:00 0 29 27 25 35 0 411 9494 1167334 28.08.2016 02:00:00 0 28 12 16 84 0 435 5944 1220988 28.08.2016 03:00:00 0 28 15 1 17 0 72 5955 1102848 28.08.2016 04:00:00 0 27 13 16 17 0 68 4715 1160580 28.08.2016 05:00:00 0 28 23 0 81 0 361 3585 1100722 28.08.2016 06:00:00 0 28 18 0 58 0 264 3281 1162927 28.08.2016 07:00:00 0 29 13 0 19 0 61 1908 1219200 28.08.2016 08:00:00 0 29 14 0 23 0 67 3752 1121374 28.08.2016 09:00:00 0 29 16 0 22 0 68 2042 1102494 28.08.2016 10:00:00 0 28 14 0 20 0 67 2191 1317052 ... 03.09.2016 19:00:00 0 11 16 1 79 0 238 2943 1191940 03.09.2016 20:00:00 0 15 18 2 96 0 205 3640 1284953 03.09.2016 21:00:00 0 13 25 1 102 0 439 3985 1155491 03.09.2016 22:00:00 0 47 7068 965 1147 0 10445 11709 1201118 03.09.2016 23:00:00 0 43 19698 512 2354 0 56763 16627 1119525 04.09.2016 01:00:00 4 500 50406 6114 14543 0 131447 79139 2224995 04.09.2016 04:00:00 55 1067 65711 4350 16350 0 165858 199986 3820512 04.09.2016 05:00:00 0 29 12118 766 2378 0 13428 55055 1160076 04.09.2016 06:00:00 1 32 2510 54 1242 0 7708 47570 1295052 04.09.2016 07:00:00 0 21 546 1 305 0 3139 42468 1299331 04.09.2016 08:00:00 0 18 265 16 200 0 1788 28516 1242703 04.09.2016 09:00:00 0 13 377 1 174 0 1777 13472 1194265 04.09.2016 10:00:00 0 15 447 1 215 0 2371 15196 1242622 04.09.2016 11:00:00 0 13 404 2 205 0 2312 14984 1207877 04.09.2016 12:00:00 0 13 405 1 238 0 2664 16214 1217226 04.09.2016 13:00:00 0 15 597 1 354 0 3093 14902 1232148 04.09.2016 14:00:00 0 17 486 1 224 0 2252 12952 1313141 04.09.2016 15:00:00 0 5 486 1 254 0 2404 13285 1176055 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
  • 61. © 2016 Pythian 61 SQL> @awr_wait_class_hist.sql 14 1 TIME 'Administrative' 'Application' 'Commit' 'Concurrency' 'Other' 'Scheduler' 'System I/O' 'User I/O' 'Idle' ------------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- 28.08.2016 01:00:00 0 29 27 25 35 0 411 9494 1167334 28.08.2016 02:00:00 0 28 12 16 84 0 435 5944 1220988 28.08.2016 03:00:00 0 28 15 1 17 0 72 5955 1102848 28.08.2016 04:00:00 0 27 13 16 17 0 68 4715 1160580 28.08.2016 05:00:00 0 28 23 0 81 0 361 3585 1100722 28.08.2016 06:00:00 0 28 18 0 58 0 264 3281 1162927 28.08.2016 07:00:00 0 29 13 0 19 0 61 1908 1219200 28.08.2016 08:00:00 0 29 14 0 23 0 67 3752 1121374 28.08.2016 09:00:00 0 29 16 0 22 0 68 2042 1102494 28.08.2016 10:00:00 0 28 14 0 20 0 67 2191 1317052 ... 03.09.2016 19:00:00 0 11 16 1 79 0 238 2943 1191940 03.09.2016 20:00:00 0 15 18 2 96 0 205 3640 1284953 03.09.2016 21:00:00 0 13 25 1 102 0 439 3985 1155491 03.09.2016 22:00:00 0 47 7068 965 1147 0 10445 11709 1201118 03.09.2016 23:00:00 0 43 19698 512 2354 0 56763 16627 1119525 04.09.2016 01:00:00 4 500 50406 6114 14543 0 131447 79139 2224995 04.09.2016 04:00:00 55 1067 65711 4350 16350 0 165858 199986 3820512 04.09.2016 05:00:00 0 29 12118 766 2378 0 13428 55055 1160076 04.09.2016 06:00:00 1 32 2510 54 1242 0 7708 47570 1295052 04.09.2016 07:00:00 0 21 546 1 305 0 3139 42468 1299331 04.09.2016 08:00:00 0 18 265 16 200 0 1788 28516 1242703 04.09.2016 09:00:00 0 13 377 1 174 0 1777 13472 1194265 04.09.2016 10:00:00 0 15 447 1 215 0 2371 15196 1242622 04.09.2016 11:00:00 0 13 404 2 205 0 2312 14984 1207877 04.09.2016 12:00:00 0 13 405 1 238 0 2664 16214 1217226 04.09.2016 13:00:00 0 15 597 1 354 0 3093 14902 1232148 04.09.2016 14:00:00 0 17 486 1 224 0 2252 12952 1313141 04.09.2016 15:00:00 0 5 486 1 254 0 2404 13285 1176055 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
  • 62. © 2016 Pythian 62 EXAMPLE 2 SO WHAT ARE WE WAITING ON? (WAIT CLASSES)
  • 63. © 2016 Pythian 63 • What could it be? – We're doing way more IOs – The IOs are much slower – … or both EXAMPLE 2 “USER I/O” IS MISBEHAVING
  • 64. © 2016 Pythian 64 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O? select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, event_name, sum(delta_total_waits) total_waits, round(sum(delta_time_waited/1000000),3) total_time_s, round(sum(delta_time_waited)/decode(sum(delta_total_waits),0,null,sum(delta_total_waits))/1000,3) avg_time_ms from (select hse.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.begin_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, EVENT_NAME, WAIT_CLASS, TOTAL_WAITS-(lag(TOTAL_WAITS,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_total_waits, TIME_WAITED_MICRO-(lag(TIME_WAITED_MICRO,1) over(partition by hs.STARTUP_TIME, EVENT_NAME order by hse.snap_id)) delta_time_waited from DBA_HIST_SYSTEM_EVENT hse, DBA_HIST_SNAPSHOT hs where hse.snap_id=hs.snap_id and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 and hse.WAIT_CLASS like '&wait_class') group by time, event_name order by 2, to_date(time,'DD.MM.YYYY HH24:MI:SS');
  • 65. © 2016 Pythian 65 SQL> @awr_wait_class_events_trend.sql "User I/O" 14 24 TIME EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- ... 28.08.2016 00:00:00 db file scattered read 110851 2332.624 21.043 29.08.2016 00:00:00 db file scattered read 553580 7823.429 14.132 30.08.2016 00:00:00 db file scattered read 710412 8074.224 11.366 31.08.2016 00:00:00 db file scattered read 686109 8779.227 12.796 01.09.2016 00:00:00 db file scattered read 3697362 79473.993 21.495 02.09.2016 00:00:00 db file scattered read 905800 12899.301 14.241 03.09.2016 00:00:00 db file scattered read 181632 5821.909 32.053 04.09.2016 00:00:00 db file scattered read 142537 44193.256 310.048 05.09.2016 00:00:00 db file scattered read 326866 32658.860 99.915 06.09.2016 00:00:00 db file scattered read 258082 29108.440 112.788 07.09.2016 00:00:00 db file scattered read 534720 44248.693 82.751 08.09.2016 00:00:00 db file scattered read 158424 22083.148 139.393 09.09.2016 00:00:00 db file scattered read 313153 47113.132 150.448 10.09.2016 00:00:00 db file scattered read 145106 14431.633 99.456 ... EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
  • 66. © 2016 Pythian 66 SQL> @awr_wait_class_events_trend.sql "User I/O" 14 24 TIME EVENT_NAME TOTAL_WAITS TOTAL_TIME_S AVG_TIME_MS ------------------- ---------------------------------------------------------------- --------------- -------------- -------------- ... 28.08.2016 00:00:00 db file sequential read 15851686 24652.524 1.555 29.08.2016 00:00:00 db file sequential read 12892354 41750.047 3.238 30.08.2016 00:00:00 db file sequential read 37228593 89856.611 2.414 31.08.2016 00:00:00 db file sequential read 23431904 71298.237 3.043 01.09.2016 00:00:00 db file sequential read 26866886 82063.899 3.054 02.09.2016 00:00:00 db file sequential read 20273240 60593.329 2.989 03.09.2016 00:00:00 db file sequential read 5114798 28039.911 5.482 04.09.2016 00:00:00 db file sequential read 7142647 401434.020 56.202 05.09.2016 00:00:00 db file sequential read 8949913 292546.728 32.687 06.09.2016 00:00:00 db file sequential read 19490699 665357.848 34.137 07.09.2016 00:00:00 db file sequential read 16387403 541792.005 33.061 08.09.2016 00:00:00 db file sequential read 6989352 241881.608 34.607 09.09.2016 00:00:00 db file sequential read 12549290 453219.850 36.115 10.09.2016 00:00:00 db file sequential read 4823556 119185.469 24.709 ... EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
  • 67. © 2016 Pythian 67 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
  • 68. © 2016 Pythian 68 EXAMPLE 2 AWR_WAIT_CLASS_EVENTS_TREND.SQL - WHAT HAPPENED TO THE USER I/O?
  • 69. © 2016 Pythian 69 The Issue was passed on to the storage team. They found 2 faulty drives. EXAMPLE 2 THE ROOT CAUSE
  • 70. © 2016 Pythian 70 EXAMPLE 3 Redo size
  • 71. © 2016 Pythian 71 • Oracle e-Business Suite – R12.1 – 12.1.0.2 • Alert from the monitoring – We're running out of space on the mount point that's used as a long term storage for archived logs (10 days) EXAMPLE 3 THE PROBLEM
  • 72. © 2016 Pythian 72 select to_char(time,'DD.MM.YYYY HH24:MI:SS') time, stat_name, sum(delta_value) value_delta from (select hss.snap_id, trunc(sysdate-&days_history+1)+trunc((cast(hs.end_interval_time as date)-(trunc(sysdate-&days_history+1)))*24/(&interval_hours))*(&interval_hours)/24 time, stat_name, value, value-(lag(value,1) over(partition by hs.startup_time, stat_name order by hss.snap_id)) delta_value from dba_hist_sysstat hss, dba_hist_snapshot hs where hss.snap_id=hs.snap_id and hss.instance_number=hs.instance_number and hs.begin_interval_time>=trunc(sysdate)-&days_history+1 and hss.stat_name like ' ') group by time, stat_name order by 2, to_date(time,'DD.MM.YYYY HH24:MI:SS'); EXAMPLE 3 AWR_STAT_TREND.SQL
  • 73. © 2016 Pythian 73 SQL> @awr_stat_trend "redo size" 30 4 TIME STAT_NAME VALUE_DELTA ------------------- ---------------- ----------------- ... 22.10.2016 20:00:00 redo size 29672405900 23.10.2016 00:00:00 redo size 45868325988 23.10.2016 04:00:00 redo size 28625802412 23.10.2016 08:00:00 redo size 46741701432 23.10.2016 12:00:00 redo size 48761724412 23.10.2016 16:00:00 redo size 46302390088 23.10.2016 20:00:00 redo size 29891530168 24.10.2016 00:00:00 redo size 39012726684 24.10.2016 04:00:00 redo size 33850511384 24.10.2016 08:00:00 redo size 27530642800 24.10.2016 12:00:00 redo size 103480430448 24.10.2016 16:00:00 redo size 124196709984 24.10.2016 20:00:00 redo size 30110908608 25.10.2016 00:00:00 redo size 38282203780 25.10.2016 04:00:00 redo size 34092919668 EXAMPLE 3 WHEN DO DID WE GENERATE MORE REDO?
  • 74. © 2016 Pythian 74 EXAMPLE 3 WHEN DO DID WE GENERATE MORE REDO?
  • 75. © 2016 Pythian 75 EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385764.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385765.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385766.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385767.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385768.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385769.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385770.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385771.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385772.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385773.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385774.arc'); EXEC DBMS_LOGMNR.ADD_LOGFILE('/arch_ret/prod/prod1_719619092_385775.arc'); EXEC DBMS_LOGMNR.START_LOGMNR(OPTIONS => DBMS_LOGMNR.DICT_FROM_ONLINE_CATALOG); EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
  • 76. © 2016 Pythian 76 set pages 50000 lines 500 tab off time on serverout on col owner for a30 col object_name for a30 col subobject_name for a30 col object_type for a30 select data_obj#, owner, object_name, subobject_name, count(*) REDO_CNT, sum(R_SIZE_B)/1024/1024 REDO_MB from ( select data_obj#, THREAD#, SEQUENCE#, RBABLK, RBABYTE, RBABLK*512+RBABYTE b_offset, (lead(RBABLK*512+RBABYTE) over (partition by RBASQN order by RBABLK*512+RBABYTE))-(RBABLK*512+RBABYTE) R_SIZE_B from V$LOGMNR_CONTENTS) lc, dba_objects o where o.data_object_id(+)=lc.data_obj# group by data_obj#, owner, object_name, subobject_name order by 6; EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
  • 77. © 2016 Pythian 77 DATA_OBJ# OWNER OBJECT_NAME SUBOBJECT_NAME REDO_CNT REDO_MB ---------- ----------- ------------------------------ ----------------- ---------- ---------- ... 4214022 AR SYS_LOB0004214021C00002$$ 16997 55.844368 4214007 AR SYS_LOB0004214006C00002$$ 11991 60.0471535 160758 APPLSYS MO_GLOB_ORG_ACCESS_TMP_U1 431892 80.7151947 112443 784710 142.132919 0 CTXSYS DR$DBO 859097 253.653015 137676 3774665 565.485004 181691 APPS XX_OM_PRICING_DETAILS_GTT_N100 3840811 709.441868 1026278 APPS XX_OM_PRICING_DETAILS_GTT_NS 3870531 772.384396 181690 APPS XX_OM_PRICING_DETAILS_GTT_N400 3886342 817.994175 181692 APPS XX_OM_PRICING_DETAILS_GTT_N200 3943281 882.903255 880271 APPS XX_OM_PRICING_DETAIL_GTT_N20 4118952 1245.52386 181689 APPS XX_OM_PRICING_DETAILS_GTT_N300 4238632 1464.44829 307 rows selected. EXEC DBMS_LOGMNR.END_LOGMNR; EXAMPLE 3 WHAT CAUSED THE EXCESSIVE REDO?
  • 78. © 2016 Pythian 78 One of the concurrent programs inserting into the GTT was executed more often during the time of the issue A) Added +append and set alter session set temp_undo_enabled=true B) The frequency of the concurrent program was reduced C) 4 of 6 indexes on the GTT were dropped EXAMPLE 3 THE ROOT CAUSE Thanks, @davidmkurtz
  • 79. © 2016 Pythian 79 Redo Size (M) EXAMPLE 3 THE ROOT CAUSE Conventional load Direct path load (+append) Temporary Undo = False 674 200 Temporary Undo = True 13 8
  • 80. © 2016 Pythian 80 SUMMARY
  • 81. © 2016 Pythian 81 • AWR reports are not good in displaying how performance changes over time. • Many performance metrics stored in AWR can be: – Put on a timeline to see how they change over time – Useful for Troubleshooting – Useful for Prediction • Remember, you're still working with aggregated data SUMMARY
  • 82. © 2016 Pythian 82 SUMMARY Photo by Kate Ter Haar / CC BY 2.0
  • 83. © 2016 Pythian 83 AWR is a complex beast, but it’s not necessary to know much to start mining Be Brave! But buy Diagnostics pack licenses first.
  • 84. © 2016 Pythian 84 THANK YOU @MarisDBA Elsins@pythian.com