What mail client are you using?
It completely stuffs up threading which is really annoying.

On Tue, Aug 11, 2009 at 8:02 AM, Alain.Moulle<[email protected]> wrote:
> Hi,
> I really don't catch the thing between "no-quorum-policy" value and the
> probability of "race to stonith" ... For me, if there is a problem on
> the whole
> heartbeat network (so for all nodes in the cluster), whatever the number
> of healthy nodes
> (with quorum or not) , there will be a race to stonith .
> Could you  explain to me my misunderstanding ?

Only partitions with quorum attempt to stonith other nodes.
By definition at most one such partition can exist in the cluster.

Only when you tell the cluster to ignore quorum can more than one
partition attempt to shoot the other.


> Moreover, in a two-nodes cluster, I think that "no-quorum-policy=ignore"
> is mandatory,
> otherwise failover will not be possible ... right or not ?

Right. But then I don't normally recommend 2 node clusters either :-)
A cheap PC from 10 years ago is enough to act as a tie-breaker (and
putting it in standby prevents it from running resources).
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to