I have two nodes (both as guest on different Xen Dom0)
- it's nfs on virtual ip on top of DRBD device
(no STONITH configured)
in CRM - order, colocation and group configured
When I stop corosync process on one node - another always promotes DRBD to Primary, next raises virtual IP and then starts NFS - so complete automatic failover.
but
when I kill process or destroy guest or poweroff whole xen machine - sometimes corosync promotes secondary to primary and sometimes not...
What can be the reason?
- is that the case of assuming I/O error by DRBD? (and then according to configured behaviour in drbd.conf)
- bad timeouts values (drbd and corosync *.conf)?
Why there's promotion or no promotion in - I'd say - similar conditions?
