[Linux-HA] Need help with resource-stickyness and STONITH

[Andreas Mock]

> -----Ursprüngliche Nachricht-----
> Von: "Christian Wörns" <christian.woerns at googlemail.com>
> Gesendet: 04.07.08 17:51:17
> An: linux-ha at lists.linux-ha.org
> Betreff: [Linux-HA] Need help with resource-stickyness and STONITH


> Hallo,

Hi

> 
> I have successfully set up a HB2 / DRBD cluster on two HP ProLiant GL380G5
> (ILO2) with Heartbeat-Documentation provided for SLES10. 

One of the most importatnt questions is:
What version of HA are you using, what version of DRBD?

Failover behaviour concerning groups changed from version to version
as far as I know.

Why do you think STONITH should happen to a node when you shutdown 
mysql by hand? I didn't looked at your config, but I would assume that
the monitor action for that resource will see that there is no mysql anymore
and will restart the group on the same or other node. If this works STONITH is
not needed.

STONITH IS nedded when HA wants to get sure that a node/resource is down
and HA can't be sure.

So, if you want to check stonith, make the stop action of the resource agent return
something bad. Then HA does not know if this resource is stopped and will stop it
definitely by shooting the node, if STONITH is configured properly.

Best regards
Andreas Mock