Comments on: The simplest query for checking what’s happening in a database http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/ Oracle troubleshooting, internals and performance tuning Thu, 11 Mar 2010 20:28:36 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: tanelp http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-497 tanelp Tue, 12 Aug 2008 01:43:07 +0000 http://tanelpoder.wordpress.com/?p=109#comment-497 Yep, you can use also SQL_ID (or whatever other column of interest). SQL_ID doesn't exist in 9i so I'm still used to use SQL hash value. Btw, SQL_ID is just a fancy presentation of SQL hash value, the library cache is still organized by hash value internally. Yep, you can use also SQL_ID (or whatever other column of interest). SQL_ID doesn’t exist in 9i so I’m still used to use SQL hash value.

Btw, SQL_ID is just a fancy presentation of SQL hash value, the library cache is still organized by hash value internally.

]]> By: mdinh http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-496 mdinh Mon, 11 Aug 2008 17:57:31 +0000 http://tanelpoder.wordpress.com/?p=109#comment-496 Why not use SQL_ID versus SQL_HASH_VALUE which in turn can be used with dbms_xplan.display_cursor() Why not use SQL_ID versus SQL_HASH_VALUE which in turn can be used with dbms_xplan.display_cursor()

]]> By: tanelp http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-498 tanelp Mon, 11 Aug 2008 04:14:03 +0000 http://tanelpoder.wordpress.com/?p=109#comment-498 Hi Jerry, I usually run the query few times to get a better statistic sample of what's happening in the instance. The SUM(SECONDS_IN_WAIT) could be deceiving as if there's a single session which has been waiting for some lock for long time (but there are no other waiters) then the significance of this wait might be pushed too high in the output report. When diagnosing *instance* performance state here, it makes sense to concentrate on the event on which the highest number of sessions happen to wait.. and that's the philosophy behind ASH enterprise manager screens as well. Hi Jerry,

I usually run the query few times to get a better statistic sample of what’s happening in the instance.

The SUM(SECONDS_IN_WAIT) could be deceiving as if there’s a single session which has been waiting for some lock for long time (but there are no other waiters) then the significance of this wait might be pushed too high in the output report. When diagnosing *instance* performance state here, it makes sense to concentrate on the event on which the highest number of sessions happen to wait.. and that’s the philosophy behind ASH enterprise manager screens as well.

]]> By: tanelp http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-499 tanelp Mon, 11 Aug 2008 04:09:20 +0000 http://tanelpoder.wordpress.com/?p=109#comment-499 Hi Tom, Oracle asks the process CPU usage numbers from OS, using times() syscall for example. Thus the CPU usage numbers are relatively correct, no such measurement like 2x difference should happen at OS level. The issue you are experiencing may be due Oracle updating the session the CPU usage only at the end of the db call. Thus if you have a long-running SQL statement, the "CPU used by this session" is not updated before the statement execution/fetch finishes. If you happen to take a sample just when the execution (which has lasted for X hours) finishes, you'd see all this CPU time reported at once, thus it can be larger than wallclock time during your sampling period. Starting from 10g you can sample V$SESS_TIME_MODEL instead (as my Snapper and Sesspack tools do), this way the CPU time is updated every 5 seconds. Hi Tom,

Oracle asks the process CPU usage numbers from OS, using times() syscall for example. Thus the CPU usage numbers are relatively correct, no such measurement like 2x difference should happen at OS level.

The issue you are experiencing may be due Oracle updating the session the CPU usage only at the end of the db call. Thus if you have a long-running SQL statement, the “CPU used by this session” is not updated before the statement execution/fetch finishes. If you happen to take a sample just when the execution (which has lasted for X hours) finishes, you’d see all this CPU time reported at once, thus it can be larger than wallclock time during your sampling period.

Starting from 10g you can sample V$SESS_TIME_MODEL instead (as my Snapper and Sesspack tools do), this way the CPU time is updated every 5 seconds.

]]> By: kenny r. http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-502 kenny r. Fri, 08 Aug 2008 04:11:36 +0000 http://tanelpoder.wordpress.com/?p=109#comment-502 hi tanel, thanks for the tips. your blog is always a great read. hi tanel,

thanks for the tips.

your blog is always a great read.

]]> By: Jerry http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-501 Jerry Thu, 07 Aug 2008 16:05:46 +0000 http://tanelpoder.wordpress.com/?p=109#comment-501 Hi Tanel, Do you think it would be useful to grab SUM(SECONDS_IN_WAIT) as well? Jerry p.s. I really enjoy reading your blog, thanks for doing this. Hi Tanel,

Do you think it would be useful to grab SUM(SECONDS_IN_WAIT) as well?

Jerry

p.s. I really enjoy reading your blog, thanks for doing this.

]]> By: Tom http://blog.tanelpoder.com/2008/08/07/the-simplest-query-for-checking-whats-happening-in-a-database/comment-page-1/#comment-500 Tom Wed, 06 Aug 2008 20:38:17 +0000 http://tanelpoder.wordpress.com/?p=109#comment-500 Tanel, on a related note concerning "On CPU/runqueue", is it accurate to say that (for simplicity) on a 1 cpu server Oracle could record more than one hour of CPU usage? Specifically, I'm seeing on occasion, 2 hours of "CPU used by this session" in a 1 hour period. Could I be correct in attributing this extra hour to waiting on cpu, ie...run queue? Thanks much for any guidance you can provide! Tanel, on a related note concerning “On CPU/runqueue”, is it accurate to say that (for simplicity) on a 1 cpu server Oracle could record more than one hour of CPU usage? Specifically, I’m seeing on occasion, 2 hours of “CPU used by this session” in a 1 hour period. Could I be correct in attributing this extra hour to waiting on cpu, ie…run queue? Thanks much for any guidance you can provide!

]]>