How to patch a database (home) in Restart environment

[ThiloSolbrig]

Hi,

Could you please kindly shed some light on how to patch a database home with active databases when Oracle Restart is used !?

I found we can set stop_processes=true in db_patches_config. But this means also to stop the listener. With Oracle Restart, this doesn't make sense because the listener doesn't use the home affected (and even fails when role_separation=true). At the other hand, if stop_processes is left "false", the databases are left running while the software is patched. This at least leads to warnings continuously appearing in alert.log and likely violates the requirement to shutdown any processes using the home to be patched.

Thanks and regards,
Thilo.

[Rendanic]

My first instinct was that it’s an error, but actually only the listeners belonging to the same ORACLE_HOME that is being patched via db_home_config should be stopped.

If all listeners are stopped — including those that do not belong to the corresponding ORACLE_HOME — then that’s a bug.

When installing a Release Update, the listener must be stopped because OPatch complains about running processes, which is correct since the binaries need to be relinked afterward.

For a OneOff patch, one could consider separating this behavior, but that would require an additional attribute or a module adjustment to check whether running processes need to be stopped before executing opatch apply.

Conclusion:
As long as the listener is a binary that belongs to the database home, it should currently be stopped.
If a listener in a different, non-patched ORACLE_HOME is stopped, then that’s a bug.

[ThiloSolbrig]

Sorry for the delay, Thorsten.
I guess there actually is a bug in the "ps" command checking if the listener is running from the home to be patched. Doesn't it rather has to be like:

index f2aec54..745dd91 100644
--- a/plugins/modules/oracle_opatch.py
+++ b/plugins/modules/oracle_opatch.py
@@ -401,7 +401,7 @@ def stop_process(module, oracle_home):
                 elif len(line.split(':')) >= 2 and line.split(':')[1] == oracle_home:
                     # Find listener for ORACLE_HOME
                     p = subprocess.Popen(
-                        'ps -o cmd -C tnslsnr | grep "^%s/bin/tnslsnr "' % (line),
+                        'ps -o cmd -C tnslsnr | grep "^%s/bin/tnslsnr "' % (oracle_home),
                         shell=True,
                         stdout=subprocess.PIPE,
                         stderr=subprocess.PIPE,

[ThiloSolbrig]

...and just one more question: Did I overlook something in oraswdb_manage_patches or are databases intentionally left down after patching?

[Rendanic]

I never paid attention to that, since I always start the playbook opatch.yml for patching. After patching the binaries, it triggers the “datapatch” via the role oradb_datapatch, which in turn starts the database beforehand.

As a first idea, I’d suggest keeping this behavior as it is, because that way it’s clear that after patching the binaries, the “datapatch” step is still missing.

[ThiloSolbrig]

Thanks for the quick response. So that was the point I actually overlooked. I'm going to use opatch.yml now.