Re: PAF with Pacemaker

Hi,

On Thu, 30 Nov 2023 19:07:34 +0000
Vijaykumar Patil <vijaykumar(dot)patil(at)maersk(dot)com> wrote:

> I have two postgres server one is primary and other one replica, I have setup
> replication and configured pacemaker and corosync.
>
> But still I'm facing issue while creating resource. It is showing invalid
> parameters.
>
> [root(at)scrbtrheldbaas001 heartbeat]# pcs status
> Cluster name: pg_cluster
> Cluster Summary:
> * Stack: corosync (Pacemaker is running)
> * Current DC: scrbtrheldbaas001 (version 2.1.6-8.el8-6fdc9deea29) -
> partition with quorum
> * Last updated: Thu Nov 30 19:04:29 2023 on scrbtrheldbaas001
> * Last change: Thu Nov 30 13:41:53 2023 by root via cibadmin on
> scrbtrheldbaas002
> * 2 nodes configured
> * 2 resource instances configured
>
> Node List:
> * Online: [ scrbtrheldbaas001 scrbtrheldbaas002 ]
>
> Full List of Resources:
> * Clone Set: pgsqld-clone [pgsqld] (promotable):
> * Stopped (invalid parameter): [ scrbtrheldbaas001 scrbtrheldbaas002 ]

Side note: make sure to setup fencing and/or watchdog.

> My postgres version is 15.3 but still is searching recover.conf . please
> find below log.

It does not search for the recovery.conf for v15. In fact, if you setup a
recovery.conf with postgres v15, PAF errors immediately with the appropriate
error message:

https://github.com/ClusterLabs/PAF/blob/master/script/pgsqlms#L1350

> ...
> Nov 30 13:43:32 scrbtrheldbaas002.crb.apmoller.net
> pacemaker-controld[1114133]: notice: Result of probe operation for pgsqld on
> scrbtrheldbaas002: invalid parameter (Recovery template file must contain
> "standby_mode = on")

This is the real error. But this error is only checked for v11 and before. So
now I wonder what version of PAF you are actually using? Is it up to date or a
very old one?

Or maybe the agent failed to parse correctly your actual version from
$PGDATA/PG_VERSION ?

https://github.com/ClusterLabs/PAF/blob/master/script/pgsqlms#L637

++