[Pacemaker] CMAN integration questions

[Andrew Beekhof]

On Thu, Mar 24, 2011 at 9:27 AM, Vladislav Bogdanov
<bubble at hoster-ok.com> wrote:
> 23.03.2011 21:38, Pavel Levshin wrote:
>> 23.03.2011 15:56, Vladislav Bogdanov:
>>
>>
>>> After 1 minute vd01-d takes over DC role.
>>> ============
>>> Mar 23 10:10:03 vd01-d crmd: [1875]: info: update_dc: Set DC to vd01-d
>>> (3.0.5)
>>
>> Excuse me, I have not much knowledge of cman integration, but don't you
>> think that DC election should be much faster?
>
> I do. But this could depend on many factors, number of nodes in a
> cluster (16 in my case), totem transport used on underlying layer
> (UDPU), etc. Probably Andrew can clarify this, I cannot.

Not really, I know very little of how cman works or is configured.

Potentially its related to the messaging timeouts used by corosync
when its reading the configuration from cluster.conf
Pacemaker can only react to the information its been given - and the
timeouts may affect how long it takes for that information to reach
pacemaker.

But its impossible to say much about what the cluster is doing based
on the provided log fragments.

>> Pacemaker hardly can work without DC. And STONITH of unexpectedly down
>> node should be much faster, too. It's not clear from your log excerpts
>> why fencing the node take so long.
>
> I understand. The main point was that fenced does the same much faster
> if it has fence devices configured.

Yes, but then then you create an internal split brain condition.

>
> I checked this, and fencing by fenced takes only 20 seconds on my setup.
> Then DLM unlocks and cluster continues to work. The only drawback is
> that node is killed twice, once by fenced, and once by pacemaker. But
> this is very minor issue comparing to 3-4 minutes DLM lock.
>
> Maybe it could be possible to make pacemaker fencing faster in this
> particular case, but this may require some efforts if it is even possible.
>
> Best,
> Vladislav
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>