Archive for the ‘Plan Stability’ Category
Sometimes you do what you have to do. So here I confess doing something I usually avoid: forcing an Execution Plan (which is not the same as using a more conventional method for Plan stability).
This is a case on 126.96.36.199.0 base release where the application vendor sets the optimizer to 9i, and tweaks other CBO parameters in questionable ways, then some queries produce suboptimal plans (as expected); and you are called to help without changing the obvious.
There is a family of queries from an ad-hoc query generator that permits users to issue queries without a set of selective predicates. These queries join several large tables and their performance is poor (as expected as well!). On top of the previous, all these queries include the /*+ FIRST_ROWS */ CBO Hint and the questionable DISTINCT keyword. Note: it is quite common for developers to throw a DISTINCT keyword “to avoid duplicates” where the mere existence of duplicates would be an indication of an application bug; so “why fix it if I can hide it, right?”.
There is one caveat although: these queries include a generic predicate “rownum <= :b1”, and value passed defaults to 5000, so users rationale is “if I only want the first X rows my query should return fast”. This highlights still another questionable practice since it is hard to imagine a user scrolling 5000 rows and making any sense of such large set, especially when the full “filtered” set would be several million rows long. So the original problem is questionable in several ways. Nevertheless, sometimes we are called to help besides providing advice. And no, we are not allowed to slap hands 😉
The good news is that we can use this extra predicate on rownum and make these queries to return the first X rows really fast; and I mean less than 5 seconds instead of over one hour or more! And if users want not 5000 but 500 or even 50 rows, then we can be in the sub-second range!
You may be thinking FIRST_ROWS optimization, and that was my first try. Unfortunately, on 188.8.131.52.0, even reversing all the suboptimal CBO parameters at the session level, I would consistently get an Execution Plan with a few Hash Joins and a large Cost; and if I were to force a Nested Loop Plan, the cost would be several orders of magnitude larger so the CBO would not pick it! Nevertheless, such a “Nest Loop only” Execution Plan would fulfill the user’s expectations, regardless the validity of the initial request. And yes, CBO statistics are OK, not perfect but simply OK. One more piece of info: this is not Exadata! (if it were Exadata most probably these same Execution Plans with full table scans and Hash Joins would simply fly!).
So, my issue became: How do I force an Execution Plan that only contains Nested Loops? If I could do that, then the COUNT STOP operation could help me to halt my SQL execution once I fetched the first X rows (Hash Join does not allow me do that). Remember: these tables have literally millions of rows. I could pepper these queries with a ton of CBO Hints and I would get my desired “Nested Loop only” Execution Plan… But that would be a lot of work and tricky at best.
SQL Patch to the rescue
I could had used a SQL Profile, but I think this dirty trick of suppressing Hash Joins and Sort Merge Joins, would be better served with a SQL Patch. I also thought Siebel: They do tweak CBO parameters as well, and they suppress Hash Joins, but they change System and Session level parameters… Since I wanted my change to be very localized, SQL Patch could provide me just what I needed.
Under the Downloads section on the margin of this page, there is a “cscripts” link that includes the sqlpch.sql script. I used this script and passed as the second parameter the following string (1st parameter is SQL_ID). With a SQL Patch generated this way, I could systematically produce a “Nested-Loops only” Execution Plan for these few queries. I did not have to change the original SQL, nor change the CBO environment at the System or Session level, neither restrict the query generator, and I did not had to “educate” the users to avoid such unbounded queries.
OPT_PARAM("_optimizer_sortmerge_join_enabled" "FALSE") OPT_PARAM("_hash_join_enabled" "FALSE")
I have to concede doing something questionable, in this case using a SQL Patch to force a desired Execution Plan instead of fixing the obvious, simply because that was the shortest path to alleviate the user’s pain.
I consider this technique above a temporary work-around and not a solution to the actual issue. In this case the right way to handle this issue would be:
- Have the application vendor certify their application to the latest release of the database and reset all CBO related parameters, plus
- Have the application vendor remove CBO Hints and DISTINCT keyword from queries, plus
- Configure the ad-hoc query generator to restrict users from executing queries without selective predicates, then
- Tune those outlier queries that may still need some work to perform as per business requirements, and possibly
- Educate the users to provide as many selective predicates as possible
Anyways, the potential of using a SQL Patch to tweak an Execution Plan in mysterious ways is quite powerful, and something we may want to keep in the back of our minds for a rainy day…
Quite often I learn of a system where “nobody changed anything” and suddenly the system is experiencing some strange behavior. Then after diligent investigation it turns out someone changed a little parameter at the System level, but somehow disregarded mentioning it since he/she thought it had no connection to the unexpected behavior. As we all know, System parameters are big knobs that we don’t change lightly, still we often see “unknown” changes like the one described.
Script below produces a list of changes to System parameter values, indicating when a parameter was changed and from which value into which value. It does not filter out cache re-sizing operations, or resource manager plan changes. Both would be easy to exclude, but I’d rather see those global changes listed as well.
Note: This script below should only be executed if your site has a license for the Oracle Diagnostics pack (or Tuning pack), since it reads from AWR.
WITH all_parameters AS ( SELECT snap_id, dbid, instance_number, parameter_name, value, isdefault, ismodified, lag(value) OVER (PARTITION BY dbid, instance_number, parameter_hash ORDER BY snap_id) prior_value FROM dba_hist_parameter ) SELECT TO_CHAR(s.begin_interval_time, 'YYYY-MM-DD HH24:MI') begin_time, TO_CHAR(s.end_interval_time, 'YYYY-MM-DD HH24:MI') end_time, p.snap_id, p.dbid, p.instance_number, p.parameter_name, p.value, p.isdefault, p.ismodified, p.prior_value FROM all_parameters p, dba_hist_snapshot s WHERE p.value != p.prior_value AND s.snap_id = p.snap_id AND s.dbid = p.dbid AND s.instance_number = p.instance_number ORDER BY s.begin_interval_time DESC, p.dbid, p.instance_number, p.parameter_name /
While delivering a session at KSCOPE 14, I was asked a very good question: Can I create a SQL Plan Baseline for an Original SQL out of a Modified SQL? In other words, query Q1, which performs poorly, has SQL_ID1 and SQL_HANDLE1 with a poorly performing Execution Plan with Hash Value PHV1. With a small modification to this query, like adding a CBO Hint or removing one, we obtain query Q2, which performs well, and has SQL_ID2, SQL_HANDLE2 and PHV2. So what we want it to associate PHV2 to SQL_ID1. The questions is: how do we do that with an API?
Maria Colgan explains on this blog how to manually do exactly what is requested above. So the question becomes: can we do this manual procedure with an easy to use API?
Script coe_load_sql_baseline.sql below inputs SQL_ID1, SQL_ID2 and PHV2, which correspond to the Original SQL_ID, the Modified SQL_ID and the Plan Hash Value from the Modified SQL (it could have more than one plan); then it produces a SQL Plan Baseline for the Original SQL out of the Execution Plan of the Modified SQL. By doing this we can “fix” a Query performing poorly by attaching to it a good performing Execution Plan that was produced by making a small temporary change to the Original SQL, like adding or removing a small set of CBO Hints. So we can generate good Execution Plans without actually changing the application.
Script coe_load_sql_baseline.sql is provided inside SQLT (MOS 215187.1) under sqlt/utl directory. Use always the latest version. As of June 2014 the latest version is the one below. This script does not require to install SQLT.
SPO coe_load_sql_baseline.log; SET DEF ON TERM OFF ECHO ON FEED OFF VER OFF HEA ON LIN 2000 PAGES 100 LONG 8000000 LONGC 800000 TRIMS ON TI OFF TIMI OFF SERVEROUT ON SIZE 1000000 NUM 20 SQLP SQL>; SET SERVEROUT ON SIZE UNL; REM REM $Header: 215187.1 coe_load_sql_baseline.sql 184.108.40.206 2013/05/10 carlos.sierra $ REM REM Copyright (c) 2000-2013, Oracle Corporation. All rights reserved. REM REM AUTHOR REM email@example.com REM REM SCRIPT REM coe_load_sql_baseline.sql REM REM DESCRIPTION REM This script loads a plan from a modified SQL into the SQL REM Plan Baseline of the original SQL. REM If a good performing plan only reproduces with CBO Hints REM then you can load the plan of the modified version of the REM SQL into the SQL Plan Baseline of the orignal SQL. REM In other words, the original SQL can use the plan that was REM generated out of the SQL with hints. REM REM PRE-REQUISITES REM 1. Have in cache or AWR the text for the original SQL. REM 2. Have in cache the plan for the modified SQL REM (usually with hints). REM REM PARAMETERS REM 1. ORIGINAL_SQL_ID (required) REM 2. MODIFIED_SQL_ID (required) REM 3. PLAN_HASH_VALUE (required) REM REM EXECUTION REM 1. Connect into SQL*Plus as user with access to data dictionary REM and privileges to create SQL Plan Baselines. Do not use SYS. REM 2. Execute script coe_load_sql_baseline.sql passing first two REM parameters inline or until requested by script. REM 3. Provide plan hash value of the modified SQL when asked. REM REM EXAMPLE REM # sqlplus system REM SQL> START coe_load_sql_baseline.sql gnjy0mn4y9pbm b8f3mbkd8bkgh REM SQL> START coe_load_sql_baseline.sql; REM REM NOTES REM 1. This script works on 11g or higher. REM 2. For a similar script for 10g use coe_load_sql_profile.sql, REM which uses custom SQL Profiles instead of SQL Baselines. REM 3. For possible errors see coe_load_sql_baseline.log REM 4. Use a DBA user but not SYS. Do not connect as SYS as the staging REM table cannot be created in SYS schema and you will receive an error: REM ORA-19381: cannot create staging table in SYS schema REM SET TERM ON ECHO OFF; PRO PRO Parameter 1: PRO ORIGINAL_SQL_ID (required) PRO DEF original_sql_id = '&1'; PRO PRO Parameter 2: PRO MODIFIED_SQL_ID (required) PRO DEF modified_sql_id = '&2'; PRO WITH p AS ( SELECT DISTINCT plan_hash_value FROM gv$sql_plan WHERE sql_id = TRIM('&&modified_sql_id.') AND other_xml IS NOT NULL ), m AS ( SELECT plan_hash_value, SUM(elapsed_time)/SUM(executions) avg_et_secs FROM gv$sql WHERE sql_id = TRIM('&&modified_sql_id.') AND executions > 0 GROUP BY plan_hash_value ) SELECT p.plan_hash_value, ROUND(m.avg_et_secs/1e6, 3) avg_et_secs FROM p, m WHERE p.plan_hash_value = m.plan_hash_value ORDER BY avg_et_secs NULLS LAST; PRO PRO Parameter 3: PRO PLAN_HASH_VALUE (required) PRO DEF plan_hash_value = '&3'; PRO PRO Values passed to coe_load_sql_baseline: PRO ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ PRO ORIGINAL_SQL_ID: "&&original_sql_id." PRO MODIFIED_SQL_ID: "&&modified_sql_id." PRO PLAN_HASH_VALUE: "&&plan_hash_value." PRO WHENEVER SQLERROR EXIT SQL.SQLCODE; SET TERM OFF ECHO ON; -- trim parameters COL original_sql_id NEW_V original_sql_id FOR A30; COL modified_sql_id NEW_V modified_sql_id FOR A30; COL plan_hash_value NEW_V plan_hash_value FOR A30; SELECT TRIM('&&original_sql_id.') original_sql_id, TRIM('&&modified_sql_id.') modified_sql_id, TRIM('&&plan_hash_value.') plan_hash_value FROM DUAL; -- open log file SPO coe_load_sql_baseline_&&original_sql_id..log; GET coe_load_sql_baseline.log; . -- get user COL connected_user NEW_V connected_user FOR A30; SELECT USER connected_user FROM DUAL; VAR sql_text CLOB; VAR plan_name VARCHAR2(30); EXEC :sql_text := NULL; EXEC :plan_name := NULL; -- get sql_text from memory BEGIN SELECT REPLACE(sql_fulltext, CHR(00), ' ') INTO :sql_text FROM gv$sqlarea WHERE sql_id = TRIM('&&original_sql_id.') AND ROWNUM = 1; EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE('getting original sql_text from memory: '||SQLERRM); :sql_text := NULL; END; / -- get sql_text from awr BEGIN IF :sql_text IS NULL OR NVL(DBMS_LOB.GETLENGTH(:sql_text), 0) = 0 THEN SELECT REPLACE(sql_text, CHR(00), ' ') INTO :sql_text FROM dba_hist_sqltext WHERE sql_id = TRIM('&&original_sql_id.') AND sql_text IS NOT NULL AND ROWNUM = 1; END IF; EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE('getting original sql_text from awr: '||SQLERRM); :sql_text := NULL; END; / -- sql_text as found SELECT :sql_text FROM DUAL; -- check is sql_text for original sql is available SET TERM ON; BEGIN IF :sql_text IS NULL THEN RAISE_APPLICATION_ERROR(-20100, 'SQL_TEXT for original SQL_ID &&original_sql_id. was not found in memory (gv$sqlarea) or AWR (dba_hist_sqltext).'); END IF; END; / -- check phv is found DECLARE l_count NUMBER; BEGIN SELECT COUNT(*) INTO l_count FROM gv$sql WHERE sql_id = TRIM('&&modified_sql_id.') AND plan_hash_value = TO_NUMBER(TRIM('&&plan_hash_value.')); IF l_count = 0 THEN RAISE_APPLICATION_ERROR(-20110, 'PHV &&plan_hash_value. for modified SQL_ID &&modified_sql_id. was not be found in memory (gv$sql).'); END IF; END; / SET ECHO OFF; DECLARE plans NUMBER; description VARCHAR2(500); sys_sql_handle VARCHAR2(30); sys_plan_name VARCHAR2(30); BEGIN -- create sql_plan_baseline for original sql using plan from modified sql plans := DBMS_SPM.LOAD_PLANS_FROM_CURSOR_CACHE ( sql_id => TRIM('&&modified_sql_id.'), plan_hash_value => TO_NUMBER(TRIM('&&plan_hash_value.')), sql_text => :sql_text ); DBMS_OUTPUT.PUT_LINE('Plans Loaded: '||plans); -- find handle and plan_name for sql_plan_baseline just created SELECT sql_handle, plan_name INTO sys_sql_handle, sys_plan_name FROM dba_sql_plan_baselines WHERE creator = USER AND origin = 'MANUAL-LOAD' AND created = ( -- past 1 minute only SELECT MAX(created) max_created FROM dba_sql_plan_baselines WHERE creator = USER AND origin = 'MANUAL-LOAD' AND created > SYSDATE - (1/24/60)); DBMS_OUTPUT.PUT_LINE('sys_sql_handle: "'||sys_sql_handle||'"'); DBMS_OUTPUT.PUT_LINE('sys_plan_name: "'||sys_plan_name||'"'); -- update description of new sql_plan_baseline description := UPPER('original:'||TRIM('&&original_sql_id.')||' modified:'||TRIM('&&modified_sql_id.')||' phv:'||TRIM('&&plan_hash_value.')||' created by coe_load_sql_baseline.sql'); plans := DBMS_SPM.ALTER_SQL_PLAN_BASELINE ( sql_handle => sys_sql_handle, plan_name => sys_plan_name, attribute_name => 'description', attribute_value => description ); DBMS_OUTPUT.PUT_LINE(plans||' plan(s) modified description: "'||description||'"'); -- update plan_name of new sql_plan_baseline :plan_name := UPPER(TRIM('&&original_sql_id.')||'_'||TRIM('&&modified_sql_id.')); :plan_name := sys_plan_name; -- avoids ORA-38141: SQL plan baseline SQL_PLAN_64b0jqr2t1h3558b5ab4d does not exist IF :plan_name <> sys_plan_name THEN plans := DBMS_SPM.ALTER_SQL_PLAN_BASELINE ( sql_handle => sys_sql_handle, plan_name => sys_plan_name, attribute_name => 'plan_name', attribute_value => :plan_name ); DBMS_OUTPUT.PUT_LINE(plans||' plan(s) modified plan_name: "'||:plan_name||'"'); END IF; -- drop baseline staging table for original sql (if one exists) BEGIN DBMS_OUTPUT.PUT_LINE('dropping staging table "STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.'))||'"'); EXECUTE IMMEDIATE 'DROP TABLE STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.')); EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE('staging table "STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.'))||'" did not exist'); END; -- create baseline staging table for original sql DBMS_OUTPUT.PUT_LINE('creating staging table "STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.'))||'"'); DBMS_SPM.CREATE_STGTAB_BASELINE ( table_name => 'STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.')), table_owner => '&&connected_user.' ); -- packs new baseline for original sql DBMS_OUTPUT.PUT_LINE('packaging new sql baseline into staging table "STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.'))||'"'); plans := DBMS_SPM.PACK_STGTAB_BASELINE ( table_name => 'STGTAB_BASELINE_'||UPPER(TRIM('&&original_sql_id.')), table_owner => '&&connected_user.', sql_handle => sys_sql_handle, plan_name => :plan_name ); DBMS_OUTPUT.PUT_LINE(plans||' pla(s) packaged'); END; / -- display details of new sql_plan_baseline SET ECHO ON; REM REM SQL Plan Baseline REM ~~~~~~~~~~~~~~~~~ REM SELECT signature, sql_handle, plan_name, enabled, accepted, fixed--, reproduced (avail on 220.127.116.11) FROM dba_sql_plan_baselines WHERE plan_name = :plan_name; SELECT description FROM dba_sql_plan_baselines WHERE plan_name = :plan_name; SET ECHO OFF; PRO PRO **************************************************************************** PRO * Enter &&connected_user. password to export staging table STGTAB_BASELINE_&&original_sql_id. PRO **************************************************************************** HOS exp &&connected_user. tables=&&connected_user..STGTAB_BASELINE_&&original_sql_id. file=STGTAB_BASELINE_&&original_sql_id..dmp statistics=NONE indexes=N constraints=N grants=N triggers=N PRO PRO If you need to implement this SQL Plan Baseline on a similar system, PRO import and unpack using these commands: PRO PRO imp &&connected_user. file=STGTAB_BASELINE_&&original_sql_id..dmp tables=STGTAB_BASELINE_&&original_sql_id. ignore=Y PRO PRO SET SERVEROUT ON;; PRO DECLARE PRO plans NUMBER;; PRO BEGIN PRO plans := DBMS_SPM.UNPACK_STGTAB_BASELINE('STGTAB_BASELINE_&&original_sql_id.', '&&connected_user.');; PRO DBMS_OUTPUT.PUT_LINE(plans||' plan(s) unpackaged');; PRO END;; PRO / PRO SPO OFF; HOS zip -m coe_load_sql_baseline_&&original_sql_id. coe_load_sql_baseline_&&original_sql_id..log STGTAB_BASELINE_&&original_sql_id..dmp coe_load_sql_baseline.log HOS zip -d coe_load_sql_baseline_&&original_sql_id. coe_load_sql_baseline.log WHENEVER SQLERROR CONTINUE; SET DEF ON TERM ON ECHO OFF FEED 6 VER ON HEA ON LIN 80 PAGES 14 LONG 80 LONGC 80 TRIMS OFF TI OFF TIMI OFF SERVEROUT OFF NUM 10 SQLP SQL>; SET SERVEROUT OFF; UNDEFINE 1 2 3 original_sql_id modified_sql_id plan_hash_value CL COL PRO PRO coe_load_sql_baseline completed.
As I prepare for one of my sessions at ODTUG Kscope14 I came across the typical situation of having a SQL for which I wanted to produce multiple optimal execution Plans on an 11g environment. As you may know, with Adaptive Cursor Sharing (ACS) this is possible and automatic, but the problem is that sometimes the ACS ramp-up process causes some suboptimal Execution Plans. If you want to skip this ACS ramp-up process, lets say for a SQL that is part of a business-critical transaction and which is known to have unstable Plans, then you may want to create a SQL Patch with the BIND_AWARE Hint. Maria Colgan explained this method on this blog post. What I present here is a script I use, so I can easily implement SQL Patches for some SQL where I just need to inject one or two CBO Hints, like this BIND_AWARE. I use SQL Profiles or SQL Plan Management when I need to provide CBO Hints that affect access paths or join order, but if I just need something like skipping ACS ramp-up or a Hint to produce a SQL Monitor report, then I’d rather use SQL Patch.
Script below asks for SQL_ID and for a short list of CBO Hints to include. By default it includes these 3: “GATHER_PLAN_STATISTICS MONITOR BIND_AWARE”. Execute this script connecting as SYS.
---------------------------------------------------------------------------------------- -- -- File name: sqlpch.sql -- -- Purpose: Create Diagnostics SQL Patch for one SQL_ID -- -- Author: Carlos Sierra -- -- Version: 2013/12/28 -- -- Usage: This script inputs two parameters. Parameter 1 the SQL_ID and Parameter 2 -- the set of Hints for the SQL Patch (default to GATHER_PLAN_STATISTICS -- MONITOR BIND_AWARE). -- -- Example: @sqlpch.sql f995z9antmhxn BIND_AWARE -- -- Notes: Developed and tested on 18.104.22.168 and 22.214.171.124 -- --------------------------------------------------------------------------------------- SPO sqlpch.txt; DEF def_hint_text = 'GATHER_PLAN_STATISTICS MONITOR BIND_AWARE'; SET DEF ON TERM OFF ECHO ON FEED OFF VER OFF HEA ON LIN 2000 PAGES 100 LONG 8000000 LONGC 800000 TRIMS ON TI OFF TIMI OFF SERVEROUT ON SIZE 1000000 NUMF "" SQLP SQL>; SET SERVEROUT ON SIZE UNL; COL hint_text NEW_V hint_text FOR A300; SET TERM ON ECHO OFF; PRO PRO Parameter 1: PRO SQL_ID (required) PRO DEF sql_id_1 = '&1'; PRO PRO Parameter 2: PRO HINT_TEXT (default: &&def_hint_text.) PRO DEF hint_text_2 = '&2'; PRO PRO Values passed: PRO ~~~~~~~~~~~~~ PRO SQL_ID : "&&sql_id_1." PRO HINT_TEXT: "&&hint_text_2." (default: "&&def_hint_text.") PRO SET TERM OFF ECHO ON; SELECT TRIM(NVL(REPLACE('&&hint_text_2.', '"', ''''''), '&&def_hint_text.')) hint_text FROM dual; WHENEVER SQLERROR EXIT SQL.SQLCODE; -- trim sql_id parameter COL sql_id NEW_V sql_id FOR A30; SELECT TRIM('&&sql_id_1.') sql_id FROM DUAL; VAR sql_text CLOB; VAR sql_text2 CLOB; EXEC :sql_text := NULL; EXEC :sql_text2 := NULL; -- get sql_text from memory DECLARE l_sql_text VARCHAR2(32767); BEGIN -- 10g see bug 5017909 FOR i IN (SELECT DISTINCT piece, sql_text FROM gv$sqltext_with_newlines WHERE sql_id = TRIM('&&sql_id.') ORDER BY 1, 2) LOOP IF :sql_text IS NULL THEN DBMS_LOB.CREATETEMPORARY(:sql_text, TRUE); DBMS_LOB.OPEN(:sql_text, DBMS_LOB.LOB_READWRITE); END IF; l_sql_text := REPLACE(i.sql_text, CHR(00), ' '); -- removes NUL characters DBMS_LOB.WRITEAPPEND(:sql_text, LENGTH(l_sql_text), l_sql_text); END LOOP; -- if found in memory then sql_text is not null IF :sql_text IS NOT NULL THEN DBMS_LOB.CLOSE(:sql_text); END IF; EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE('getting sql_text from memory: '||SQLERRM); :sql_text := NULL; END; / SELECT :sql_text FROM DUAL; -- get sql_text from awr DECLARE l_sql_text VARCHAR2(32767); l_clob_size NUMBER; l_offset NUMBER; BEGIN IF :sql_text IS NULL OR NVL(DBMS_LOB.GETLENGTH(:sql_text), 0) = 0 THEN SELECT sql_text INTO :sql_text2 FROM dba_hist_sqltext WHERE sql_id = TRIM('&&sql_id.') AND sql_text IS NOT NULL AND ROWNUM = 1; END IF; -- if found in awr then sql_text2 is not null IF :sql_text2 IS NOT NULL THEN l_clob_size := NVL(DBMS_LOB.GETLENGTH(:sql_text2), 0); l_offset := 1; DBMS_LOB.CREATETEMPORARY(:sql_text, TRUE); DBMS_LOB.OPEN(:sql_text, DBMS_LOB.LOB_READWRITE); -- store in clob as 64 character pieces WHILE l_offset < l_clob_size LOOP IF l_clob_size - l_offset > 64 THEN l_sql_text := REPLACE(DBMS_LOB.SUBSTR(:sql_text2, 64, l_offset), CHR(00), ' '); ELSE -- last piece l_sql_text := REPLACE(DBMS_LOB.SUBSTR(:sql_text2, l_clob_size - l_offset + 1, l_offset), CHR(00), ' '); END IF; DBMS_LOB.WRITEAPPEND(:sql_text, LENGTH(l_sql_text), l_sql_text); l_offset := l_offset + 64; END LOOP; DBMS_LOB.CLOSE(:sql_text); END IF; EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE('getting sql_text from awr: '||SQLERRM); :sql_text := NULL; END; / SELECT :sql_text2 FROM DUAL; SELECT :sql_text FROM DUAL; -- validate sql_text BEGIN IF :sql_text IS NULL THEN RAISE_APPLICATION_ERROR(-20100, 'SQL_TEXT for SQL_ID &&sql_id. was not found in memory (gv$sqltext_with_newlines) or AWR (dba_hist_sqltext).'); END IF; END; / PRO generate SQL Patch for SQL "&&sql_id." with CBO Hints "&&hint_text." SELECT loaded_versions, invalidations, address, hash_value FROM v$sqlarea WHERE sql_id = '&&sql_id.' ORDER BY 1; SELECT child_number, plan_hash_value, executions, is_shareable FROM v$sql WHERE sql_id = '&&sql_id.' ORDER BY 1, 2; -- drop prior SQL Patch WHENEVER SQLERROR CONTINUE; PRO ignore errors EXEC DBMS_SQLDIAG.DROP_SQL_PATCH(name => 'sqlpch_&&sql_id.'); WHENEVER SQLERROR EXIT SQL.SQLCODE; -- create SQL Patch PRO you have to connect as SYS BEGIN SYS.DBMS_SQLDIAG_INTERNAL.I_CREATE_PATCH ( sql_text => :sql_text, hint_text => '&&hint_text.', name => 'sqlpch_&&sql_id.', category => 'DEFAULT', description => '/*+ &&hint_text. */' ); END; / -- flush cursor from shared_pool PRO *** before flush *** SELECT inst_id, loaded_versions, invalidations, address, hash_value FROM gv$sqlarea WHERE sql_id = '&&sql_id.' ORDER BY 1; SELECT inst_id, child_number, plan_hash_value, executions, is_shareable FROM gv$sql WHERE sql_id = '&&sql_id.' ORDER BY 1, 2; PRO *** flushing &&sql_id. *** BEGIN FOR i IN (SELECT address, hash_value FROM gv$sqlarea WHERE sql_id = '&&sql_id.') LOOP DBMS_OUTPUT.PUT_LINE(i.address||','||i.hash_value); BEGIN SYS.DBMS_SHARED_POOL.PURGE ( name => i.address||','||i.hash_value, flag => 'C' ); EXCEPTION WHEN OTHERS THEN DBMS_OUTPUT.PUT_LINE(SQLERRM); END; END LOOP; END; / PRO *** after flush *** SELECT inst_id, loaded_versions, invalidations, address, hash_value FROM gv$sqlarea WHERE sql_id = '&&sql_id.' ORDER BY 1; SELECT inst_id, child_number, plan_hash_value, executions, is_shareable FROM gv$sql WHERE sql_id = '&&sql_id.' ORDER BY 1, 2; WHENEVER SQLERROR CONTINUE; SET DEF ON TERM ON ECHO OFF FEED 6 VER ON HEA ON LIN 80 PAGES 14 LONG 80 LONGC 80 TRIMS OFF TI OFF TIMI OFF SERVEROUT OFF NUMF "" SQLP SQL>; SET SERVEROUT OFF; PRO PRO SQL Patch "sqlpch_&&sql_id." will be used on next parse. PRO To drop SQL Patch on this SQL: PRO EXEC DBMS_SQLDIAG.DROP_SQL_PATCH(name => 'sqlpch_&&sql_id.'); PRO UNDEFINE 1 2 sql_id_1 sql_id hint_text_2 hint_text CL COL PRO PRO sqlpch completed. SPO OFF;
I have this query that references a couple of Global Temporary Tables (GTT). These GTT have no CBO Statistics, thus Dynamic Sampling (DS) is used on them. The problem is default value of 2 at the instance level samples only 64 blocks and these GTT are large. Testing with DS on larger samples provides better quality on these dynamic statistics and a better performing Execution Plan for this particular SQL.
This SQL already uses an imported SQL Profile generated by coe_xfr_sql_profile.sql out of SQLT (MOS 215187.1) under sqlt/utl directory. This SQL Profile contains CBO Hints created out of the Outline Data contained on the other_xml column of the Plan. Still a higher level of DS is needed for these two GTT in order to produce a better performing Plan. Setting DS at the instance level or session level is not an option. Modifying the query is not an option. So the question is: “how to embed a dynamic sampling hint on a SQL that already has a SQL Profile based on an Outline“?
On the script generated by coe_xfr_sql_profile.sql add one extra Hint with the DS level that is needed. Look at sample below, where one extra Hint has added right below DB_VERSION. Since the two GTT have no CBO Statistics, they invoke DS, which would use now level 6 instead of 2.
h := SYS.SQLPROF_ATTR( q'[BEGIN_OUTLINE_DATA]', q'[IGNORE_OPTIM_EMBEDDED_HINTS]', q'[OPTIMIZER_FEATURES_ENABLE('126.96.36.199')]', q'[DB_VERSION('188.8.131.52')]', q'[OPT_PARAM('optimizer_dynamic_sampling' 6)]', q'[ALL_ROWS]',
Every so often I see on a distribution list a posting that starts like this: “I upgraded my application from database release X to release Y and now many queries are performing poorly, can you tell why?”
As everyone else on a distribution list, my first impulse is to make an educated guess permeated by a prior set of experiences. The intentions are always good, but the process is painful and time consuming. Many of us have seen this kind of question, and many of us have good hunches. Still I think our eagerness to help blinds us a bit. The right thing to do is to step back and analyze the facts, and I mean all the diagnostics supporting the observation.
What is needed to diagnose a SQL Tuning issue?
The list is large, but I will enumerate some of the most important pieces:
- SQL Text
- Version of the database (before and after upgrade)
- Database parameters (before and after)
- State of the CBO Statistics (before and after)
- Changes on Histograms
- Basics about the architecture (CPUs, memory, etc.)
- Values of binds if SQL has them
- Indexes compare, including state (visible?, usable?)
- Execution Plan (before and after)
- Plan stability? (Stored Outlines, Profiles, SQL Plan Management)
- Performance history as per evidence on AWR or StatsPack
- Trace from Event 10053 to understand the CBO
- Trace from Event 10046 level 8 or 12 to review Waits
- Active Session History (ASH) if 10046 is not available
I could keep adding bullets to the list, but I think you get the point: There are simply too many things to check! And each takes some time to collect. More important, the state of the system changes overtime, so you may need to re-collect the same diagnostics more than once.
SQLTXPLAIN to the rescue
SQLT or SQLTXPLAIN, has been available on MetaLink (now MOS) under note 215187.1 for over a decade. In short, SQLT collects all the diagnostics listed above and a lot more. That is WHY Oracle Support uses it every day. It simply saves a lot of time! So, I always encourage fellow Oracle users to make use of the FREE tool and expedite their own SQL Tuning analysis. When time permits, I do volunteer to help on an analysis. So, if you get to read this, and you want to help yourself while using SQLT but feel intimidated by this little monster, please give it a try and contact me for assistance. If I can help, I will, if I cannot, I will let you know.
It is fun to guess WHY a SQL is not performing as expected, and trying different guesses is educational but very time consuming. If you want to actually find root causes before trying to fix your SQL, you may want to collect relevant diagnostics. SQLT is there to help, and if installing this tool is not something you can do in a short term, consider then SQL Health-Check SQLHC.
Every so often I see systems where there is a good amount of SQL that uses Literals instead of Binds, and executes enough times to create a large number of Cursors. Is this a red flag? As many questions regarding performance, I would say the right answer is: it all depends.
Of course we want to use Binds instead of Literals in order to reduce the frequency of Hard Parses, and in turn reduce CPU consumption and space utilization in the Shared Pool. Does it mean we want to replace all Literals with Binds? Do we declare war on Literals? In my opinion, the answer is simply: NO.
If a SQL has a Predicate on a date column, or a key column, then I would expect the Number of Distinct Values (NDV) for such column to be high, and in some cases as high as the number of rows in the Table (unique values for example). In the other hand, if the Predicate is in one of those columns that denotes a code, like Process Type or Status, and the NDV is small, then I’d rather keep the Literal in place. Specially if the data in such column is skewed and I have (or plan to have) Histograms on it.
What do I propose?
- If the SQL is executed sporadically, then it does not matter (Literals or Binds).
- If the SQL executes frequently, and the Predicate in question is on a Column where the Number of Distinct Values (NDV) is high, then use a Bind instead of a Literal (for this Predicate).
- If the SQL executes frequently, and the Predicate in question is on a Column where the NDV is low, then use a Literal (for this Predicate). This assumes the NDV for these Literals is also small.
- Regardless if using a Literal or a Bind for a particular Predicate: If the data in a Column referenced by a Predicate is heavily skewed, gather Statistics with Histograms on this Column.
Follow-up question: When the NDV is high or low? The answer is also: it all depends. I personally prefer to see Literals if the NDV for this Column (and this Literal) is less than 10 (or so).
Why having Binds and Literals on same SQL is better than having all Binds?
If we have good set of CBO Statistics, and we have Histograms on skewed data, and we are using bind peeking, and we are on 11g, and Adaptive Cursor Sharing (ACS) is enabled, and we plan using SQL Plan Management (SPM), then we are for a treat:
With all the “ands” above, by using Binds on predicates with high NDV and Literals in those with low NDV, then we will end up having a small number of different SQL_IDs for what we consider “the same SQL”. Each incarnation of this SQL could potentially have its own set of optimal Execution Plans created by ACS and the CBO (by making use of Histograms on the data and Selectivity Profiles on ACS). Then, with the aid of SPM we could provide stability to those multiple optimal Execution Plans for each version of the SQL. That means that SQL Q1 with Literal L1 could have a different set of optimal plans than Q1 with Literals L2.
Replacing some Literals with Binds but not all Literals sounds like a lot of work, but actually the extra work may be worth the effort. In my opinion, the end result is better if we replace most, but not all (as per proposal above). WHY? Even when ACS does a good job at finding multiple optimal plans for a SQL by using the selectivity of the predicates; by allowing a small number of cursors for the same SQL given the use of Literals in columns with low NDV, we are basically reducing the times we would have to execute a SQL with a sub-optimal plan due to current ACS ramp-up process. This extra granularity provided by a small number of incarnations of the “otherwise same SQL” could be crucial for tuning complex SQL or corner cases.