Re: v$sql 'db block changes'

Selects can cause write activity, both for sorting/hashing into temp tablespace also for delayed block cleanout (which is written to disk by dbwr in background).

Which version are you on? From 9i you could use v$segment_statistics to get some understanding for which segments most of the writes are done.

Also query v$session_event for all those sessions and make sure that foreground direct writes (or direct lob writes) arent prevalent there.

Tanel.

• Original Message ----- From: "Allen, Brandon" <Brandon.Allen_at_OneNeck.com> To: "Ethan Post" <post.ethan_at_gmail.com> Cc: <oracle-l_at_freelists.org> Sent: Thursday, December 15, 2005 1:45 PM Subject: RE: v$sql 'db block changes'

> If only it were that easy. The problem here is that the application uses
> connection pooling, so there is no single session responsible for the
> heavy updates - it bounces around amongst 40 different sessions. The top
> statements in v$sql in terms of cpu, elapsed_time, buffer_gets and
> disk_reads are all SELECT statements, so I know they're not causing the
> write activity.
>
> Thanks,
> Brandon

--
http://www.freelists.org/webpage/oracle-l