RE: RAC rolling patch application
Date: Tue, 22 Jul 2008 11:38:56 -0400
There are two plausible reasons for Oracle to declare a patch false on rolling patch.
- It is too much work to prove that the interaction of the new binary starting an instance on a database where the previous binary is running is safe.
- They know for sure it is incompatible and it will either break for sure or you have to get lucky and not experience the incompatible interinstance interaction.
In neither case is it safe to ignore Oracle's warning that it is not a rolling patch.
Since you can apparently handle the load with one instance, you should be able to shut down the non-primary instance, (starting exposure to no active failover target), apply the patch to the non-primary instance's software source, shutdown normal the primary instance, (beginning outage), startup restricted old software primary-switch logfile-archive complete-shutdown normal, start "secondary" instance using new software, (ending outage unless you didn't test properly and it blows up in your face), apply patch to primary, start primary using new software (ending exposure).
Compared to risking it all unsupported, I wouldn't call that much of a choice.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of daniel koehne
Sent: Tuesday, July 22, 2008 10:45 AM
Subject: RAC rolling patch application
We need/want to apply Oracle patch 5907779 to our two node 10.2.0.3 RAC Solaris 9 cluster. We run this cluster with init.ora active_instance_count=1 which makes one instance the primary and the other instance the secondary--in summary all connections are routed to the primary instance.
Querying the patch lists the following as the install actions:
Update /oracle/product/10.2/lib/libserver10.a with /kkxs.o
; make -f ins_rdbms.mk ioracle ORACLE_HOME=/oracle/product/10.2
I also see that this patch is not a rolling upgrade patch. OPatch query -all ->"Patch is a rolling patch: false"
We had initially planned to apply this patch as a rolling upgrade (shutdown our secondary instance and apply patch, then failover [scondary instance becomes primary instance by shutting down primary] and apply the patch to the second node) but I am reluctant to proceed this way. I have been trying to get Oracle support, without much success, to address our specific situation (RAC cluster with active_instance_count=1).
I am wondering if anyone has a feeling for how "out there blowing in the wind" we would be if we went with a rolling patch upgrade for this patch?
I won't tell Oracle support if you won't!
http://www.freelists.org/webpage/oracle-l Received on Tue Jul 22 2008 - 10:38:56 CDT