don't remove LOCK file when PGSQL stopped with quorum lost.

[greenx]

Hi!
I again found trouble. )
I doing experiment with quorum lost.
After quorum lost LOCK not deleted.
I see code.

    if  [ "$1" = "master" -a "$OCF_RESKEY_CRM_meta_notify_slave_uname" = " " ]; then
        ocf_log info "Removing $PGSQL_LOCK."
        rm -f $PGSQL_LOCK
    fi

Where defined $OCF_RESKEY_CRM_meta_notify_slave_uname ? what he mean?

[greenx]

One answer found http://clusterlabs.org/doc/en-US/Pacemaker/1.1/html-single/Pacemaker_Explained/#_multi_state_notifications

But the meaning is not quite clear.
Lost quorum - the resources will stopped.
I made ​​a stand of four nodes.
If two node down - quorum lost and pacemaker apply policy 'stop'.
But there is one slave.

Or should we switched on and off in the correct order.

[t-matsuo]

If two node down - quorum lost and pacemaker apply policy 'stop'.
But there is one slave.

It may be a bug of pacemaker.

[greenx]

You mean to say, that after the loss of a quorum pacemaker should send a message to stop all the resources and reset this variable($OCF_RESKEY_CRM_meta_notify_slave_uname), because resource master-slave too must be stopped?

[t-matsuo]

You mean to say, that after the loss of a quorum pacemaker should send a message to stop all the resources
and reset this variable($OCF_RESKEY_CRM_meta_notify_slave_uname), because resource master-slave too
must be stopped?

Yes.

[Wintermute3]

In the comparison above:

"$OCF_RESKEY_CRM_meta_notify_slave_uname" = " "

Why the space between the " " for the comparison target? Seems like an error to me. The OCF_RESKEY_CRM_meta_notify_slave_uname got trim'ed upon import, so blank strings would have become empty strings, no?

[playmobil77d]

Hi,
Is there a solution about this problem ?

[greenx]

I dont know. My project was closed. I almost was not interested in the topic.

[furynick]

I have the same problem with 2 nodes cluster.

The LOCK is never deleted as OCF_RESKEY_CRM_meta_notify_slave_uname always contains so the slave can't be restarted.

Perhaps this file can be deleted on slave successful stop after wal sync check

[t-matsuo]

Hi furynick

Sorry, I switched to another task 5 years ago.

This agent was merged into ClusterLabs repository and maintained at its community.
Could you open new topic at ClusterLabs resource-agent repository ?
Someone may respond.

[furynick]

Thanks for reply, I already found related issue at ClusterLabs#699