Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> 8.1.7 cache buffer chains contention

8.1.7 cache buffer chains contention

From: Scott Gamble <zifnab_at_NOSPAM.reddragon.org>
Date: Thu, 23 Aug 2001 14:26:59 GMT
Message-ID: <To8h7.6141$e8.2596175@e3500-chi1.usenetserver.com>


OS = TRU64 4.0F
CPUS = 12
DB = 8.1.7.1 + one off patch to address 4031 issues

Has anyone run into very high 'cache buffers chains' contention after upgrading to 8.1.7.1 with no application changes at all. Sounds _very_ similiar to a bug that was reported in addition to the above mentioned one off patch we applied. However it was included in the 'fix list' for the 8.1.7.1 path so supposedly is fixed, yet we are showing extraordinary contention.

Our application did not change, and the execution plans did not change either. Yet with the upgrade from 8.0.6.1 to 8.1.7.1 our latch contention has gone through the roof, to the point where the application is almost unusable.

Having been on the phone with Oracle support much of the last 3 days, we have tried to change a few hidden parameters with no success.

   _db_block_hash_latches, and _db_block_hash_buckets

the default calculations for both the above drastically changed with 8.1.7 (confirmed by Oracle) (dont think he latches parameter was even settable in 8.0.6)

We changed _spin_count last night with no real noticeable results to the application though as expected 'sleeps' on the latch did go down.

Just for reference the application does have a 'hot' index root block that is under the 'hot' latch, but this has not changed for years. Since we bought the application that block has been 'hot'.

I can post all the details if anyone is interested (statspacks, etc)

Oracle has not been able to provide a solution to the problem and are positive that we are not running into any of the 'known' bugs.

Just curious if anyone else has run into this and actually managed to solve the problem. Metalink is full of descriptions of the same issues, just no resolutions. Realizing that the application is at fault for the one hot block still doesnt really address why its so much worse in 8.1.7.1

Scott Gamble
Allegiance Healthcare
gambles_at_allegiance.net Received on Thu Aug 23 2001 - 09:26:59 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US