SUMMARY: Sep-24-03 Rec.Patch.Cluster on Ora8i e450-> Performance Hit Issue ?

MANY thanks to all who replied -- in no particular order:  John 
Mitchell, Tom Crummery, Hichael Morton, Grzegorz Bakalarski, Joe Fletcher.


Comments Suggest,

-I'm not the only person in this boat, it seems, and no trivial / 
certain answer has surfaced yet.

-Are you certain you trust your DBA ? :-) Have all recent changes been 
documented? Really!? and ; if on sun support, give 'em a call since they 
should know if "issues" exist with given patch cluster. (Alas the box in 
question is not on sun support). Barring that, read all the Readme docs 
in the patch cluster and look for obvious gotcha's.

-Run a patch tool such as "patch pro" (or similar) which sun provides 
from sunsolve.sun.com - since recommended patch clusters are not going 
to guarantee you get *all* "required patches" for a given system - 
rather, the "highlights and most critical" of patches available.

-Re-run "analyze schema" on the database, on the chance it produces 
usefull info (and even if not, it isn't a bad thing to do from 
time-to-time) ; check driver rev's for HBAs since there may be issues 
with a patch for qlc (qlogic) in a recent patch cluster.



I haven't had any downtime available on this sytem since making my 
posting, so don't have a "final answer" at this time -- but I've now got 
"patch pro" installed and ready to roll with a large bundle of patches, 
including a more recent qlc patch than one in my 
most-recent-patch-cluster-in-question.

BTW, [IMHO] PatchPro is a very-easy-to-use patch manager, and from a 
test run on a development system here, I'm certainly going to be using 
it on all production boxes here in the future, and would recommend it 
strongly to other folks who are currently relying exclusively on 
pre-cooked "patch clusters" from sun for their patching needs. [for more 
info, pay a visit to " http://patchpro.sun.com " ]

When I roll the patches (sometime in the next few weeks likely), I'll 
post a followup summary IFF there is an improvement to the status of the 
performance issue.

Hope this is of slight use to others out there,


--_Tim







ORIGINAL POSTING FOLLOWS
=================================================================================
On Thu, Oct 02, 2003 at 10:22:20AM -0400, Tim Chipman wrote:
Hi All,

I apologize for the lame nature of this query ; unfortunately it is the
kind of thing that only can be gleaned from this sort of forum, potentially.

We applied last FridayPM what was the most-recent "recommended patch
cluster" from sunsolve on all of our servers here (sept-24-03 cluster)
including our primary database machine (Oracle8i on an e450, 4x400mhz
2gig ram, has T3 disk array for primary DB volume, A1000 for some
low-priority DB data).

Since that time, we've observed a not-insignificant hit to performance
when doing "large batch jobs" (for instance, one job that runs nightly
is a large "select", reading approx. 700,000 records ; prior to applying
the patch cluster, this job ran in ~45 minutes ; since applying the
patch cluster, it takes nearly 90 minutes to complete).

Our DBA assures me until blue-in-face that *nothing* else has changed
WRT the oracle config, and we're quite certain that no other changes
were rolled into this system anytime recently (nor since the last reboot
even).

So, this is just a general query - on the "off chance" that anyone else
who applied this patch cluster to an oracleDBserver has observed a
similar hit to performance.

Given the ## of patches applied in a cluster like this, I'm not terribly
optimistic that I can easily diagnose the underlying cause. Options that
are readily open which I can see include, (a) rollback to pre-patch
state using our "disaster recovery dumps" of the system, or (b) apply
the most recent patch cluster just released, dated Oct-1-03. [as an
aside - since when does Sun release new recommended patch clusters this
frequently?!? it used to be far less often a few years ago, IIRC?].

Any thoughts / comments or ideally similar experiences confirming that
I'm not insane -- would be greatly appreciated.

Thanks,

Tim
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers