[Pacemaker] Quorum question

Andrew Beekhof andrew at beekhof.net
Mon Mar 29 12:45:26 EDT 2010


On Mon, Mar 29, 2010 at 6:35 PM, Robert <vanleeuwen at stone-it.com> wrote:
> Hi,
>
> I expected that quorum would be lost if you have half of the nodes
> offline.

Agreed. Can you file a bug for this please?
But sure to include a hb_report with the logs.

I'll be out for most of the next two weeks, but I'll take a look when I return.

(Oh, and please use the correct list address, its @oss.clusterlabs.org
 Thanks :)

>
> However this seems to be the case when I test this:
> I simulating a network failure by blocking all traffic via a crontab
> iptables rule on half of the nodes:
>
> iptables -I INPUT -j DROP
> iptables -I OUTPUT -j DROP
>
> Fencing is configured and the 4 remaining nodes try to fence the 4
> "dead" nodes.
> I would NOT expect this to happen because quorum should be lost...?
>
> Am I missing something?
>
> Below the output of crm status were you can see that 4 nodes are gone
> while quorum is retained.
>
>>>>
> ============
> Last updated: Mon Mar 29 17:52:21 2010
> Stack: openais
> Current DC: blade62 - partition with quorum
> Version: 1.0.7-6e1815972fc236825bf3658d7f8451d33227d420
> 8 Nodes configured, 8 expected votes
> 8 Resources configured.
> ============
>
> Node blade59: UNCLEAN (offline)
> Node blade73: UNCLEAN (offline)
> Node blade72: UNCLEAN (offline)
> Node blade71: UNCLEAN (offline)
> Online: [ blade60 blade61 blade62 blade63 ]
>
>
>
> Thanks,
> Robert van Leeuwen
>
>
>
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>




More information about the Pacemaker mailing list