Slow Query Issues Oracle 11GR1

Question

Hi All,
Oracle 11.1.0.6 on AIX 6.1. CPU-8 CORES, Memory - 32G db size - 500G
Recently we did a OS Hardening(mainly on OS security) followed by upgrade of RAM (from 16G to 32G) in a production environment.
We restarted the DB but did not restart the server.
1> Just after the activity our Stored Programs in PL/SQL all running very slow.
2> Other issues present from before the activity;
a. High log file switch completion (avg > 140ms)
b. High waits in db sequential reads, db file scattered read, direct path write temp
c. Redo log file siizes are 500M, switching every 2 mins without much activity.
3> data, redo, control, arch all share 2 mount points.
4> We never run any dbms_stats as we recently went to production.
5) After running EXEC DBMS_STATS.GATHER_FIXED_OBJECTS_STATS(NULL); the Stored Programs were more slow.
6) ping and tnsping was Ok.
What should be the action plan to resolve the immediate and long term issues..?

Answer 1

user13355115 wrote:
Hi All,
Oracle 11.1.0.6 on AIX 6.1. CPU-8 CORES, Memory - 32G db size - 500G
Recently we did a OS Hardening(mainly on OS security) followed by upgrade of RAM (from 16G to 32G) in a production environment.
We restarted the DB but did not restart the server.
1> Just after the activity our Stored Programs in PL/SQL all running very slow.
2> Other issues present from before the activity;It is expected that the upgrades/patches you did (OS and RAM) would not affect the IO times. The bigger RAM should have reduced the need to do IO though. You need address your IO system if you need faster responses there.
>
a. High log file switch completion (avg > 140ms)
b. High waits in db sequential reads, db file scattered read, direct path write temp
c. Redo log file siizes are 500M, switching every 2 mins without much activity.There must be some activity that generates that redo. It is either directly or indirectly (delayed block cleanout) generated by a user. The overall amount of redo (15G per hour) is kind of high, but not extreme for most modern hardware.
>
3> data, redo, control, arch all share 2 mount points.
4> We never run any dbms_stats as we recently went to production.Please gather application tables/indexes statistics – they are essential for achieving optimal performance.
>
5) After running EXEC DBMS_STATS.GATHER_FIXED_OBJECTS_STATS(NULL); the Stored Programs were more slow.
6) ping and tnsping was Ok.
What should be the action plan to resolve the immediate and long term issues..?Iordan Iotzov
http://iiotzov.wordpress.com/