Comment 12 for bug 1837401

Revision history for this message
Chris MacNaughton (chris.macnaughton) wrote :

It looks like the percona charm is asking the hacluster charm, via relation, to start monitoring the mysql resource *potentially* before the actual percona cluster is a cluster:

2019-07-23 20:00:32 DEBUG juju-log cluster:31: Notifying peers that percona is bootstrapped (uuid=07928334-ad82-11e9-8e6c-7f00ef700f43)
2019-07-23 20:00:32 DEBUG juju-log cluster:31: Waiting on hacluster to complete clustering, not clustered yet.
2019-07-23 20:00:32 INFO juju-log cluster:31: Installing /usr/lib/ocf/resource.d/percona/mysql_monitor
2019-07-23 20:00:33 DEBUG juju-log cluster:31: VIP HA: VIP is set 192.168.33.11
2019-07-23 20:00:33 DEBUG juju-log cluster:31: Checking for minimum of 3 peer units
2019-07-23 20:00:33 INFO juju-log cluster:31: Sufficient number of peer units to form cluster 3
2019-07-23 20:00:33 DEBUG juju-log cluster:31: mysql/0 is not yet clustered
2019-07-23 20:00:33 DEBUG juju-log cluster:31: Waiting on hacluster to complete clustering, not clustered yet.

The referenced "mysql_monitor" file is dropped at https://github.com/openstack/charm-percona-cluster/blob/master/hooks/percona_hooks.py#L583 , and then the hacluster's HA relation is updated.

If the hacluster charm sets the crm resource before percona is clustered, this failure can occur.