[ClusterLabs] MS MySQL resource not beeing promoted?

[Ken Gaillot]

On Fri, 2019-02-08 at 15:00 +0100, Michael Schwartzkopff wrote:
> Hi,
> 
> 
> I set up a mariadb on CentOS7 as master/slave-.
> 
> Manually the sync works.
> 
> 
> when I confiure a ms resource in pacmaker, it comes up correctly on
> the
> first glance.
> 
> 
> "show master status" and "show slave status" in mysql works and shows
> the correct status.
> 
> 
> When I set the master node to standby, pacemaker promotes the
> database
> server correctly, at least in the pacemaker status. BUT: "show master
> status" on the now promoted server shows nothing. The database
> definitly
> in NOT in the master state.
> 
> Setting the first node online again, pacemaker starts the database
> and
> shows it as a slave. "show master status" in the database shows, that
> the first node is master again.
> 
> 
> So, the status in pacemaker does not agree to the real status of the
> database.
> 
> Any ideas?

Make sure the clone is configured with notify=true, and the user
specified in replication_user has the correct database permissions. If
the name to use in CHANGE MASTER is not the host's $(uname -n), then an
extra resource parameter needs to be set. If none of those is it,
probably some log diving and manual attempts are in order.
-- 
Ken Gaillot <kgaillot at redhat.com>