Mailing List Archive: drbd + heartbeat / 2 or more drbd devices

>"fsck" or "mounts" executed in datadisk script are sometimes slow.
>I think this could lead to something a bit dangerous: if the former
>primary joins cluster when new primary is executed user heartbeat
>scripts (datadisk drbd0;datadisk drbd1; ....) but does not already reach
>datadisk drbdX , former primary will become primary for drbdX, what we
>don't want!

That may happen - and it will happen, cause nodes that comes up have
to sync each other. So the "primary" node will run trough and start
heartbeat that starts the services/ip's of the other node too, caused
by waiting of the 2nd node for finishing sync.

What you mean is called "nice_failback" in heartbeat - an option.

If nice_failback is enabled a node holds started services even if
the other node is up again ! So disable if dont want this behaviour.

Michael Appeldorn