RE: Session With oper EXCL is also waiting - where to now? (systemstate dump)

From: <Christopher.Taylor2_at_parallon.net>
Date: Wed, 19 Sep 2012 15:29:51 -0500
Message-ID: <F05D8DF1FB25F44085DB74CB916678E88515AA3713_at_NADCWPMSGCMS10.hca.corpad.net>



Thanks Tanel - I'm definitely looking into the bug aspect being 10.2.0.4 I've inherited a system that collects statistics every night for all tables and runs for HOURS. I'm betting the contention happened between statistics and the nightly batch processing hitting the same objects - but I'm still guessing right now - trying to confirm by connecting all the dots.

Chris

From: tanel_at_poderc.com [mailto:tanel_at_poderc.com] On Behalf Of Tanel Poder Sent: Wednesday, September 19, 2012 3:26 PM To: Taylor Christopher - Nashville
Cc: oracle-l_at_freelists.org
Subject: Re: Session With oper EXCL is also waiting - where to now? (systemstate dump)

Search for the address a5b6bd2a0 (case insensitive) in the systemstate dump file. If some process still happened to be holding that latch while the dump read its state object, it should dump out which one it was (and which session state object was under it).

But allowing shared pool to automatically shrink is a bad idea in super-critical latency-sensitive (OLTP) systems ... better "waste" some extra memory for shared pool instead of having to live with CPU usage and share pool latch/library cache mutex contention spikes...

Of course in your case the symptom was even more extreme (a hang), which may be due to a bug ...

--
Tanel Poder
Blog - http://blog.tanelpoder.com
App  - http://voic.ee


On Wed, Sep 19, 2012 at 11:01 PM, <Christopher.Taylor2_at_parallon.net<mailto:Christopher.Taylor2_at_parallon.net>> wrote:

(apologies for bad formatting - I'm trying to keep it simple tho)
(pastebin is blocked, but used gist https://gist.github.com/3751878)
3-node RAC 10.2.0.4 We had a "hang" last night from 2:50 to ~3:48 and Oracle was kind enough to provide me with an automated system state dump.
(the bottom of this email is where I'm confused - where do I go from here?)
Looking at the ASH and AWR reports for the 2 hour window: I see: 1.) An SGA Resize occurred: Buffer Cache 30,608M increases to 30,656M Shared Pool Size 5,136M *decreases* to 5,088M -- http://www.freelists.org/webpage/oracle-l
Received on Wed Sep 19 2012 - 15:29:51 CDT

Original text of this message