Don't fail if the application_name is not set properly

[frost242]

If you set application_name to a wrong value, pgsqlms returns a "OCF_ERR_CONFIGURED" that will trigger fencing of the nodes where the service are started. It should return "OCF_ERR_ARGS" instead.

pgsqlms should return a lower level error.

[ioguix]

Hi @frost242,

In fact, it returns OCF_ERR_ARGS, but I suspect Pacemaker is trying to stop your resource in reaction and the stop action to fail for some other reasons with the OCF_ERR_CONFIGURED.

I'll try to reproduce and check what we can do.

[blogh]

Hi @ioguix @frost242 ,

That's it:

Monitor fails

Feb 18 01:10:48 srv2 pgsqlms(pgsqld)[11678]: ERROR: Recovery template file must contain in primary_conninfo parameter "application_name=srv2"
Feb 18 01:10:48 srv2 lrmd[6192]:   notice: pgsqld_monitor_16000:11678:stderr [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2" ]
Feb 18 01:10:48 srv2 crmd[6195]:   notice: srv2-pgsqld_monitor_16000:25 [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2"\n ]

And again (the default value for on-fail is restart except for the stop action):

Feb 18 01:10:48 srv2 pgsqlms(pgsqld)[11689]: ERROR: Recovery template file must contain in primary_conninfo parameter "application_name=srv2"
Feb 18 01:10:48 srv2 lrmd[6192]:   notice: pgsqld_notify_0:11689:stderr [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2" ]
Feb 18 01:10:48 srv2 crmd[6195]:   notice: Result of notify operation for pgsqld on srv2: 0 (ok)
Feb 18 01:10:48 srv2 crmd[6195]:   notice: srv2-pgsqld_monitor_16000:25 [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2"\n ]

Then Stop fails

Feb 18 01:10:48 srv2 pgsqlms(pgsqld)[11700]: ERROR: Recovery template file must contain in primary_conninfo parameter "application_name=srv2"
Feb 18 01:10:48 srv2 lrmd[6192]:   notice: pgsqld_stop_0:11700:stderr [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2" ]
Feb 18 01:10:48 srv2 crmd[6195]:   notice: Result of stop operation for pgsqld on srv2: 2 (invalid parameter)
Feb 18 01:10:48 srv2 crmd[6195]:   notice: srv2-pgsqld_stop_0:28 [ ocf-exit-reason:Recovery template file must contain in primary_conninfo parameter "application_name=srv2"\n ]

As @ioguix said, Failed stop = Fence:

Feb 18 01:10:48 srv2 stonith-ng[6191]:   notice: Operation reboot of srv2 by srv1 for [email protected]: OK
Feb 18 01:10:48 srv2 crmd[6195]:     crit: We were allegedly just fenced by srv1 for srv1!

[ioguix]

Thank you for this study @blogh!

Right now, I'm not sure what is the best option to this issue. The obvious one would be to relax the pg_stop function to raise OCF_SUCCESS if we "feel" the local instance is really stopped. But I'm still uncomfortable with this presently :/