Fwd: Multi-node patching woes

From: Patrick Jolliffe <jolliffe_at_gmail.com>
Date: Thu, 25 May 2017 12:05:32 +0800
Message-ID: <CABx0cSXJ6T6HV8=r9eT1vO_gypoG=FBJYdqjZhMMOtub3uMOPg_at_mail.gmail.com>



Still waiting on Oracle Support for this, however I have just attempted to reproduce using latest version of opatch, 12.2.0.1.9 and cannot. This is because with 12.2.0.1.9 opatch appears to have 'lost' the connection between the two homes.

When I execute the following (note OPatch.backup directory is the 12.2.0.1.8 opatch version):
/u01/app/oracle/product/12.1.0.2/dbhome_2/OPatch.backup/opatch lsinventory
I get the following at the end of the output: Rac system comprising of multiple nodes
  Local node = hkexdb01
  Remote node = hkexdb02
When I execute the following (12.2.0.1.9 opatch version):
/u01/app/oracle/product/12.1.0.2/dbhome_2/OPatch/opatch lsinventory
That information is missing.

Is anyone else able to check with latest and previous versions OPatch and see whether they also see this change in behavior? TIA
Patrick

We have two Exadata compute nodes, node1 and node2, the Oracle home on both nodes is RAC configured, however we are not using RAC, and actually expect the homes on node1 and node2 to be running different patch levels (and yes we now realize that this is a problem, and we are going to resolve this using runInstaller -updateNodeList)

I have the following reproducible test case: 1. Initially node1 has 22652097 <2265%202097> (the (in)famous adaptive patch) on top of October DBBP.
2. Initially node2 has July DBBP only.
3. I apply October DBBP to node2, and choose Y when prompted, to also patch node1.
4. opatch lsinventory reveals the adaptive patch has disappeared from node1. 5. I apply the adaptive patch on node2, but chose N when prompted, to not patch node1.
6. opatch lsinventory reveals the adaptive patch still not on node1. 7. I apply October OJVM patch to node2, and choose Y when prompted, to also patch node1.
8. opatch lsinventory reveals the adaptive patch has re-appeared on node1 9. Testing reveals that the patch isn't really installed on node1 (no ADAPTIVE_PLANS/ADAPTIVE_STATS parameters)

I kind of understand what is happening (opatch is making assumptions that the homes are in-sync, whereas in reality they are not). My question is, whether this is expected or acceptable behavior given our configuration?
Ie is this worth raising as a bug, or is it our mistake in having the wrong setup.
Regards
Patrick

--
http://www.freelists.org/webpage/oracle-l
Received on Thu May 25 2017 - 06:05:32 CEST

Original text of this message