{ "date": "2017-08-23T19:41:01", "slug": "improving-statspack-experience", "link": "https://www.dbi-services.com/blog/improving-statspack-experience/", "title": { "rendered": "Improving Statspack Experience" }, "content": { "rendered": "

By Franck Pachot

\n

.
\nI’ve published a few month ago an article in the UKOUG OracleScene magazine on Improving Statspack Experience: quick setup script, changing settings, filling Idle Events,etc. In the article, I used dbms_job to schedule the snapshots, because I have this script for years and never took the time to do it with dbms_scheduler. Thanks to Nicolas Jardot here is the equivalent script using dbms_scheduler.
\n
\nThe idea is to have a script to run on each instance (when in RAC) in order to have a job calling statspack.snap and statspack.purge on each instance.

\n
\nDECLARE\n    instno NUMBER;\n    snapjob VARCHAR2(30);\n    purgejob VARCHAR2(30);\nBEGIN\n    select instance_number into instno from v$instance;\n    snapjob  := 'PERFSTAT.STATSPACK_SNAP_' || instno;\n    purgejob := 'PERFSTAT.STATSPACK_PURGE_' || instno;\n \n    DBMS_SCHEDULER.CREATE_JOB (\n       job_name => snapjob, \n       job_type => 'PLSQL_BLOCK', \n       job_action => 'statspack.snap;',\n       number_of_arguments => 0,\n       start_date => systimestamp,\n       repeat_interval => 'FREQ=HOURLY;BYTIME=0000;BYDAY=MON,TUE,WED,THU,FRI,SAT,SUN',\n       end_date => NULL,\n       enabled => TRUE,\n       auto_drop => FALSE,\n       comments => 'Take hourly Statspack snapshot');\n \n    DBMS_SCHEDULER.CREATE_JOB (\n       job_name => purgejob, \n       job_type => 'PLSQL_BLOCK', \n       job_action => 'statspack.purge(i_num_days=>31,i_extended_purge=>true);',\n       number_of_arguments => 0,\n       start_date => systimestamp,\n       repeat_interval => 'FREQ=WEEKLY;BYTIME=120000;BYDAY=SUN',\n       end_date => NULL,\n       enabled => TRUE,\n       auto_drop => FALSE,\n       comments => 'Weekly purge Statspack snapshot');\n \n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => snapjob, attribute => 'logging_level', value => DBMS_SCHEDULER.LOGGING_OFF );\n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => snapjob, attribute => 'INSTANCE_ID', value=>instno);\n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => snapjob, attribute => 'INSTANCE_STICKINESS', value=>TRUE);\n \n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => purgejob, attribute => 'logging_level', value => DBMS_SCHEDULER.LOGGING_OFF );\n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => purgejob, attribute => 'INSTANCE_ID', value=>instno);\n    DBMS_SCHEDULER.SET_ATTRIBUTE( name => purgejob, attribute => 'INSTANCE_STICKINESS', value=>TRUE);  \nEND;\n/\n
\n

I used the Oracle Cloud Service to provision quickly a two nodes RAC database to validate, and I’ll check the scheduling:

\n
\n[oracle@rac-dg01-1 admin]$ alias sql='JAVA_HOME=$ORACLE_HOME/jdk bash $ORACLE_HOME/sqldeveloper/sqlcl/bin/sql'\n[oracle@rac-dg01-1 admin]$ TWO_TASK=//10.31.143.86/pdb1.a511644.oraclecloud.internal sql sys/\"Ach1z0#d\" as sysdba\n \nSQLcl: Release 12.2.0.1.0 RC on Wed Aug 23 18:57:12 2017\n \nCopyright (c) 1982, 2017, Oracle.  All rights reserved.\n \nConnected to:\nOracle Database 12c EE Extreme Perf Release 12.2.0.1.0 - 64bit Production\n \nSQL> set sqlformat ansiconsole\nSQL> select job_name, state, enabled, next_run_date, instance_stickiness, instance_id from dba_scheduler_jobs where owner='PERFSTAT';\nJOB_NAME           STATE      ENABLED  NEXT_RUN_DATE                           INSTANCE_STICKINESS  INSTANCE_ID\n--------           -----      -------  -------------                           -------------------  -----------\nSTATSPACK_SNAP_1   SCHEDULED  TRUE     23-AUG-17 07.00.00.981193000 PM +00:00  TRUE                 1\nSTATSPACK_PURGE_1  SCHEDULED  TRUE     27-AUG-17 12.00.00.074939000 PM +00:00  TRUE                 1\nSTATSPACK_SNAP_2   SCHEDULED  TRUE     23-AUG-17 07.00.00.644681000 PM +00:00  TRUE                 2\nSTATSPACK_PURGE_2  SCHEDULED  TRUE     27-AUG-17 12.00.00.755685000 PM +00:00  TRUE                 2\n
\n

One hour later, the job has run on each instance:

\n
\nJOB_NAME           STATE      ENABLED  NEXT_RUN_DATE                           INSTANCE_STICKINESS  INSTANCE_ID\n--------           -----      -------  -------------                           -------------------  -----------\nSTATSPACK_SNAP_1   SCHEDULED  TRUE     23-AUG-17 08.00.00.325755000 PM +00:00  TRUE                 1\nSTATSPACK_PURGE_1  SCHEDULED  TRUE     27-AUG-17 12.00.00.074939000 PM +00:00  TRUE                 1\nSTATSPACK_SNAP_2   SCHEDULED  TRUE     23-AUG-17 08.00.00.644681000 PM +00:00  TRUE                 2\nSTATSPACK_PURGE_2  SCHEDULED  TRUE     27-AUG-17 12.00.00.755685000 PM +00:00  TRUE                 2\n
\n

Now running a spreport to see the instances having snapshots:

\n
\n[oracle@rac-dg01-1 admin]$ TWO_TASK=//10.31.143.86/pdb1.a511644.oraclecloud.internal/cdb12 sqlplus sys/\"Ach1z0#d\" as sysdba @ spreport\n \nInstances in this Statspack schema\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n \n   DB Id    Inst Num DB Name      Instance     Host\n----------- -------- ------------ ------------ ------------\n  924704606        2 CDB1         cdb12        rac-dg01-2\n  924704606        1 CDB1         cdb11        rac-dg01-1\n \nUsing  924704606 for database Id\nUsing          2 for instance number\n
\n

Here it is. dbms_job is deprecated. Let’s use dbms_scheduler.

\n", "protected": false } }