Oracle DBA – A lifelong learning experience

Automatically running sql_advisor tasks from ADDM reports

Posted by John Hallas on November 12, 2009

STOP PRESS – 17 Nov 2009  – updated with latest code which works against both 10g and 11g databases

 I am attaching scripts which I wrote a while ago to automatically pick any sql_ids reported in the latest ADDM and then run sql_advisor to report on any tuning advice. I am not suggesting that the information they provide is not available from EM or indeed every task reported needs resolving but it can be a good heads-up on a system you don’t know very well.

These are enabled every hour (can be less depending upon your snapshot interval) and they create a daily file which can be easily reviewed.

I find the real benefit is not on production databases but on dev and test databases that are being used for development prior to production implementation. This is for two reasons, firstly I hope that the team has a good handle on what is happening in production and are aware of issues and secondly we are most likely  to be able to add most value and benefit in development environments before the code is made live.

A couple of ‘issuettes’. The output from the ADDM report is different between 10g and 11g so I have amended the awk file to cater for both versions. I have been having an ongoing problem with tghe sql_advisor tuning task timing out on some systems and consequently leaving the task created for the next run. I have therefore amended the loop to drop the task at several points which looks untidy in the output file bit does seem to resolve the problem. 

I hoped to attach a zip file containing 4 scripts but cannot see how to do it without a plug-in which is a problem on my works PC. so in the meantime I have pasted the code of each of 4 files.

tuning_recommendations.ksh which is the controlling script

#! /bin/ksh
# loop though the file produced from get_addm_report.sql and put the gathered sql-ids into a flat file
# awk the file to get just the SQL_ID
# for each sql_id create a task, execute that task, run the report and then delete the task
#
# The delete tuning task job is run an additional twice because if the tuning task times out then it does not clean up properly
# Better to see a few failures in this job that not run the sql_tuning_advisor at all.
#if [ -d /home/oracle/logs ]
then
   rm  /home/oracle/logs/tmp*.log
   else
   mkdir /home/oracle/logs
   exit
fi
if [ $# -ne 1 ]
then
    echo "No ORACLE SID  - exiting"
    exit
fi
# execute ORACLE's .profile
#
#. ~/.profile
unset ORAENV_ASK
#
# set up environment variables.
#

ORACLE_SID=$1
. /usr/local/bin/oraenv ${ORACLE_SID}
export ORACLE_HOME=`cat /etc/oratab | grep $ORACLE_SID | awk -F: '{print $2 }`
export PATH=$ORACLE_HOME/bin:$PATH
export ORAENV_ASK=NO
today=`date +%d-%b-%Y`; export today
LOGDIR=$HOME/logs
LOGFILE=$LOGDIR/get_addm_${today}.log
REPORTFILE=$LOGDIR/sql_advisor_report_${ORACLE_SID}_${today}.log

#
#
sqlplus -s /nolog  <<SQLEND
connect / as sysdba
     spool $LOGDIR/tmp_${ORACLE_SID}_1.log
     @/shared/oracle/performance/get_addm_report.sql
     spool off
     exit
SQLEND

<p>&nbsp;</p>

cat $LOGDIR/tmp_${ORACLE_SID}_1.log|awk -f /shared/oracle/performance/tuning_recommendations.awk |awk '!a[$0]++' > $LOGDIR/tmp_${ORACLE_SID}_2.log
cat  $LOGDIR/tmp_${ORACLE_SID}_2.log | awk '$0!~/^$/ {print $0}' > $LOGDIR/tmp_${ORACLE_SID}_3.log

for PLAN in  `cat $LOGDIR/tmp_${ORACLE_SID}_3.log`
do
sqlplus -s /nolog  <<SQLEND >> $REPORTFILE
connect / as sysdba
        begin
        DBMS_SQLTUNE.drop_tuning_task('test_task1');
        end;
        /
     SELECT status FROM USER_ADVISOR_TASKS WHERE task_name = 'test_task1';
     @/shared/oracle/performance/sql_advisor.sql $PLAN
     exit
SQLEND
done

# tidy up the report file
# tidy up reports > 14 days old
find  $LOGDIR -name "sql_advisor_repor*.log" -mtime +14 -print -exec rm -f {} \;

get_addm_report.sql which gets each task from the last snapshot from dba_advisor_tasks

set long  10000000
set pagesize 50000
column get_clob format a80

select dbms_advisor.get_task_report (task_name) as ADDM_report
from dba_advisor_tasks
where task_id = (
        select max(t. task_id)
        from dba_advisor_tasks t, dba_advisor_log l
        where t.task_id = l.task_id
        and t.advisor_name = 'ADDM'
        and l.status = 'COMPLETED');

tuning_recommendations.awk is a short awk script used to process the output from get_addm_report.sql

BEGIN{
#start at the first line
#OUTFILE="$HOME/logs/outfile.log"
}
{
        {
                if (($1=="RATIONALE:") && ($2=="SQL")) #10G ADDM format
                {
                        F1=$6
                }
                if (($1=="Run") && ($2=="SQL") && ($3=="Tuning") && ($4=="Advisor") && ($7=="SQL") && ($10=="SQL_ID")) #11G ADDM format
                {
                        F1=$11
                }

        }
VAR1=substr(F1,2,13)
print VAR1
}
END{
}

sql_advisor.sql runs the sql_advisor package against each task found.

DECLARE
my_task_name   VARCHAR2 (30);
my_sqltext     CLOB;
my_sqlid        varchar2(30);

BEGIN
my_sqlid := '&1';
my_task_name := dbms_sqltune.create_tuning_task (sql_id=> my_sqlid,
       scope         => 'COMPREHENSIVE',
       time_limit    => 300,
       task_name     => 'test_task1',
       description   => 'test_task1'
    );
END;
/

BEGIN
dbms_sqltune.execute_tuning_task (task_name => 'test_task1');
END;
/

SELECT status FROM USER_ADVISOR_TASKS WHERE task_name = 'test_task1';
SET LONG 10000
SET LONGCHUNKSIZE 10000
SET LINESIZE 100
set pages 60
SELECT DBMS_SQLTUNE.REPORT_TUNING_TASK( 'test_task1')
FROM DUAL;
begin
DBMS_SQLTUNE.drop_tuning_task('test_task1');
end;
/

We have a read only NFS mounted disk available on all database servers and the files are placed in there and initiated by a cron entry for each SID on an hourly basis
40 * * * * /shared/oracle/performance/tuning_recommendations.ksh SID >/dev/null 2>&1

Output is created in a folder $HOME/logs and 14 days worth of reports are kept.

A sample output report (only one task shown but certainly on this Peoplesoft database it would show many tasks)

DBMS_SQLTUNE.REPORT_TUNING_TASK('TEST_TASK1')
----------------------------------------------------------------------------------------------------
GENERAL INFORMATION SECTION
-------------------------------------------------------------------------------
Tuning Task Name                  : test_task1
Tuning Task Owner                 : SYS
Scope                             : COMPREHENSIVE
Time Limit(seconds)               : 60
Completion Status                 : COMPLETED
Started at                        : 11/12/2009 07:43:33
Completed at                      : 11/12/2009 07:44:16
Number of Statistic Findings      : 1
Number of Index Findings          : 1

-------------------------------------------------------------------------------
Schema Name: SYSADM
SQL ID     : cy3fmjha2sjnr
SQL Text   : SELECT M.EMPLID, M.EMPL_RCD, M.SCH_PRIM_ALT_IND,
             TO_CHAR(M.DUR,'YYYY-MM-DD'), M.SEQ_NO, M.CHNG_PRIMARY,
             M.SCHEDULE_GRP, M.SETID, M.WRKDAY_ID, M.SHIFT_ID, M.SCHED_HRS,
             M.SCH_CONFIG1, M.SCH_CONFIG2, M.SCH_CONFIG3, M.SCH_CONFIG4,
             TO_CHAR(M.START_DTTM,'YYYY-MM-DD-HH24.MI.SS.&amp;quot;000000&amp;quot;'),
             TO_CHAR(M.END_DTTM,'YYYY-MM-DD-HH24.MI.SS.&amp;quot;000000&amp;quot;'),
             M.SCHED_SOURCE, M.OFFDAY_IND, A.TIMEZONE, A.SCH_CATEGORY From
             PS_SCH_MNG_SCH_TBL M, PS_SCH_ADHOC_DTL A Where M.EMPLID = :1 and
             M.EMPL_RCD = :2 and M.SCH_PRIM_ALT_IND = :3 and M.DUR between
             TO_DATE(:4,'YYYY-MM-DD') and TO_DATE(:5,'YYYY-MM-DD') and
             A.EMPLID = M.EMPLID and A.EMPL_RCD = M.EMPL_RCD and
             A.SCH_PRIM_ALT_IND = M.SCH_PRIM_ALT_IND and A.DUR = M.DUR and
             A.SEQ_NO = M.SEQ_NO and A.SEQNUM = 1 Order By M.DUR Asc,
             M.SCHED_SOURCE Desc, M.SEQ_NO Desc

-------------------------------------------------------------------------------
FINDINGS SECTION (2 findings)
-------------------------------------------------------------------------------

1- Statistics Finding
---------------------
  Optimizer statistics for index &amp;quot;SYSADM&amp;quot;.&amp;quot;PS_SCH_MNG_SCH_TBL&amp;quot; are stale.

  Recommendation
  --------------
  - Consider collecting optimizer statistics for this index.
    execute dbms_stats.gather_index_stats(ownname =&amp;gt; 'SYSADM', indname =&amp;gt;
            'PS_SCH_MNG_SCH_TBL', estimate_percent =&amp;gt;
            DBMS_STATS.AUTO_SAMPLE_SIZE);

  Rationale
  ---------
    The optimizer requires up-to-date statistics for the index in order to
    select a good execution plan.

2- Index Finding (see explain plans section below)
--------------------------------------------------
  The execution plan of this statement can be improved by creating one or more
  indices.

  Recommendation (estimated benefit: 94.67%)
  ------------------------------------------

DBMS_SQLTUNE.REPORT_TUNING_TASK('TEST_TASK1')
----------------------------------------------------------------------------------------------------
  - Consider running the Access Advisor to improve the physical schema design
    or creating the recommended index.
    create index SYSADM.IDX$$_21C0F0001 on
    SYSADM.PS_SCH_ADHOC_DTL(&amp;quot;EMPLID&amp;quot;,&amp;quot;SEQNUM&amp;quot;,&amp;quot;EMPL_RCD&amp;quot;,&amp;quot;SCH_PRIM_ALT_IND&amp;quot;,&amp;quot;DU
    R&amp;quot;);

  - Consider running the Access Advisor to improve the physical schema design
    or creating the recommended index.
    create index SYSADM.IDX$$_21C0F0002 on
    SYSADM.PS_SCH_MNG_SCH_TBL(&amp;quot;EMPLID&amp;quot;,&amp;quot;DUR&amp;quot;);

  Rationale
  ---------
    Creating the recommended indices significantly improves the execution plan
    of this statement. However, it might be preferable to run &amp;quot;Access Advisor&amp;quot;
    using a representative SQL workload as opposed to a single statement. This
    will allow to get comprehensive index recommendations which takes into
    account index maintenance overhead and additional space consumption.

-------------------------------------------------------------------------------
EXPLAIN PLANS SECTION
-------------------------------------------------------------------------------

1- Original
-----------
Plan hash value: 2070933151

----------------------------------------------------------------------------------------------------
------------------------
| Id  | Operation                             | Name               | Rows  | Bytes | Cost (%CPU)| Ti
me     | Pstart| Pstop |
----------------------------------------------------------------------------------------------------
------------------------
|   0 | SELECT STATEMENT                      |                    |     1 |   103 |   387   (1)| 00
:00:01 |       |       |
|*  1 |  FILTER                               |                    |       |       |            |
       |       |       |
|   2 |   SORT ORDER BY                       |                    |     1 |   103 |   387   (1)| 00
:00:01 |       |       |
|   3 |    NESTED LOOPS                       |                    |     1 |   103 |   386   (1)| 00
:00:01 |       |       |
|   4 |     PARTITION RANGE ITERATOR          |                    |    10 |   350 |   371   (1)| 00
:00:01 |   KEY |   KEY |
|   5 |      TABLE ACCESS BY LOCAL INDEX ROWID| PS_SCH_ADHOC_DTL   |    10 |   350 |   371   (1)| 00
:00:01 |   KEY |   KEY |
|*  6 |       INDEX RANGE SCAN                | PS_SCH_ADHOC_DTL   |    10 |       |   369   (1)| 00
:00:01 |   KEY |   KEY |
|   7 |     PARTITION RANGE ITERATOR          |                    |     1 |    68 |     2   (0)| 00
:00:01 |   KEY |   KEY |
|   8 |      TABLE ACCESS BY LOCAL INDEX ROWID| PS_SCH_MNG_SCH_TBL |     1 |    68 |     2   (0)| 00
:00:01 |   KEY |   KEY |
|*  9 |       INDEX UNIQUE SCAN               | PS_SCH_MNG_SCH_TBL |     1 |       |     1   (0)| 00
:00:01 |   KEY |   KEY |
----------------------------------------------------------------------------------------------------
------------------------

Predicate Information (identified by operation id):

DBMS_SQLTUNE.REPORT_TUNING_TASK('TEST_TASK1')
----------------------------------------------------------------------------------------------------
---------------------------------------------------

   1 - filter(TO_DATE(:4,'YYYY-MM-DD')&amp;lt;=TO_DATE(:5,'YYYY-MM-DD'))
   6 - access(&amp;quot;A&amp;quot;.&amp;quot;EMPLID&amp;quot;=:1 AND &amp;quot;A&amp;quot;.&amp;quot;EMPL_RCD&amp;quot;=TO_NUMBER(:2) AND &amp;quot;A&amp;quot;.&amp;quot;SCH_PRIM_ALT_IND&amp;quot;=:3 AND
              &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;&amp;gt;=TO_DATE(:4,'YYYY-MM-DD') AND &amp;quot;A&amp;quot;.&amp;quot;SEQNUM&amp;quot;=1 AND &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;&amp;lt;=TO_DATE(:5,'YYYY
-MM-DD'))
       filter(&amp;quot;A&amp;quot;.&amp;quot;SEQNUM&amp;quot;=1)
   9 - access(&amp;quot;M&amp;quot;.&amp;quot;EMPLID&amp;quot;=:1 AND &amp;quot;M&amp;quot;.&amp;quot;EMPL_RCD&amp;quot;=TO_NUMBER(:2) AND &amp;quot;M&amp;quot;.&amp;quot;SCH_PRIM_ALT_IND&amp;quot;=:3 AND
              &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot; AND &amp;quot;A&amp;quot;.&amp;quot;SEQ_NO&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;SEQ_NO&amp;quot;)
       filter(&amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot;&amp;gt;=TO_DATE(:4,'YYYY-MM-DD') AND &amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot;&amp;lt;=TO_DATE(:5,'YYYY-MM-DD'))

2- Using New Indices
--------------------
Plan hash value: 1209469329

----------------------------------------------------------------------------------------------------
------------------------
| Id  | Operation                             | Name               | Rows  | Bytes | Cost (%CPU)| Ti
me     | Pstart| Pstop |
----------------------------------------------------------------------------------------------------
------------------------
|   0 | SELECT STATEMENT                      |                    |     1 |   103 |    21  (10)| 00
:00:01 |       |       |
|*  1 |  FILTER                               |                    |       |       |            |
       |       |       |
|   2 |   SORT ORDER BY                       |                    |     1 |   103 |    21  (10)| 00
:00:01 |       |       |
|*  3 |    HASH JOIN                          |                    |     1 |   103 |    20   (5)| 00
:00:01 |       |       |
|   4 |     TABLE ACCESS BY GLOBAL INDEX ROWID| PS_SCH_ADHOC_DTL   |    10 |   350 |    10   (0)| 00
:00:01 | ROWID | ROWID |
|*  5 |      INDEX RANGE SCAN                 | IDX$$_21C0F0001    |    10 |       |     4   (0)| 00
:00:01 |       |       |
|*  6 |     TABLE ACCESS BY GLOBAL INDEX ROWID| PS_SCH_MNG_SCH_TBL |    13 |   884 |     9   (0)| 00
:00:01 | ROWID | ROWID |
|*  7 |      INDEX RANGE SCAN                 | IDX$$_21C0F0002    |    13 |       |     3   (0)| 00
:00:01 |       |       |
----------------------------------------------------------------------------------------------------
------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(TO_DATE(:4,'YYYY-MM-DD')&amp;lt;=TO_DATE(:5,'YYYY-MM-DD'))
   3 - access(&amp;quot;A&amp;quot;.&amp;quot;EMPLID&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;EMPLID&amp;quot; AND &amp;quot;A&amp;quot;.&amp;quot;EMPL_RCD&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;EMPL_RCD&amp;quot; AND
              &amp;quot;A&amp;quot;.&amp;quot;SCH_PRIM_ALT_IND&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;SCH_PRIM_ALT_IND&amp;quot; AND &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot; AND
              SYS_OP_DESCEND(&amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;)=SYS_OP_DESCEND(&amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot;) AND &amp;quot;A&amp;quot;.&amp;quot;SEQ_NO&amp;quot;=&amp;quot;M&amp;quot;.&amp;quot;SEQ_NO&amp;quot;)
   5 - access(&amp;quot;A&amp;quot;.&amp;quot;EMPLID&amp;quot;=:1 AND &amp;quot;A&amp;quot;.&amp;quot;SEQNUM&amp;quot;=1 AND &amp;quot;A&amp;quot;.&amp;quot;EMPL_RCD&amp;quot;=TO_NUMBER(:2) AND &amp;quot;A&amp;quot;.&amp;quot;SCH_PRIM_
ALT_IND&amp;quot;=:3 AND
              &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;&amp;gt;=TO_DATE(:4,'YYYY-MM-DD') AND &amp;quot;A&amp;quot;.&amp;quot;DUR&amp;quot;&amp;lt;=TO_DATE(:5,'YYYY-MM-DD'))
   6 - filter(&amp;quot;M&amp;quot;.&amp;quot;SCH_PRIM_ALT_IND&amp;quot;=:3 AND &amp;quot;M&amp;quot;.&amp;quot;EMPL_RCD&amp;quot;=TO_NUMBER(:2))
   7 - access(&amp;quot;M&amp;quot;.&amp;quot;EMPLID&amp;quot;=:1 AND &amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot;&amp;gt;=TO_DATE(:4,'YYYY-MM-DD') AND &amp;quot;M&amp;quot;.&amp;quot;DUR&amp;quot;&amp;lt;=TO_DATE(:5,'YYY
Y-MM-DD'))

-------------------------------------------------------------------------------

PL/SQL procedure successfully completed.

The routine above works well but I am happy to consider any changes or improvements.

PS If anybody knows how to use a code tag and not have those horrible green wraparound marks please let me know

About these ads

2 Responses to “Automatically running sql_advisor tasks from ADDM reports”

  1. Raj jamadagni said

    Nice idea, In 11G, you perhaps can use sql below to directly pick SQLIDs identified by ADDM for tuning, thus reducing your number of scripts. It may work in 10g too but I don’t have a 10g DB.

    with x as (select task_name tn from dba_advisor_tasks where task_id = (select max(t. task_id) from dba_advisor_tasks t, 
    dba_advisor_log l where t.task_id = l.task_id  and t.advisor_name = 'ADDM' and l.status = 'COMPLETED'))
    select task_Name,max(replace(regexp_substr(replace(message,' ',','),'[^,]+',1,5),'"')) as sql_id
    /*,max(regexp_substr(replace(message,' ',','),'[^,]+',1,8)) as execs
    ,max(regexp_substr(replace(message,' ',','),'[^,]+',1,17)) as avg_ela */
    from DBA_ADVISOR_RATIONALE, x where task_name = x.tn and message like 'SQL statement with SQL_ID%'
    group by task_Name, rec_id, replace(message,' ',',')
    

    The commented columns get you number of executions and averaged elapsed time as reported by ADDM.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

Join 196 other followers

%d bloggers like this: