[Pacemaker] Pacemaker multi node configuration

[Luciano Sitzia]

Hello Andreas, I found your advices very helpful, I still have some doubts
about
how to setup stonith, if you can, would you be so kind to send me a link
where this subject
is well explained. On the Pacemaker manuals I found only 3/4 pages with one
example.
Considering that the correct setup of stonith is so important I'd like to
understand
Exactly how to do it.

Thank you again for helping me with "Puzzlemaker" :).

My best regards.

Luciano

-----Original Message-----
From: Andreas Kurz [mailto:andreas.kurz at gmail.com] 
Sent: lunedì 10 ottobre 2011 11.24
To: pacemaker at oss.clusterlabs.org
Subject: Re: [Pacemaker] Pacemaker multi node configuration

Hello,

On 10/10/2011 10:17 AM, Luciano Sitzia wrote:
> Hello Andreas, you are 100% right about leaving an idle node instead of
> balancing all the VM
> Among all the four nodes. But the choice was to have and idle node that
can
> guest all the
> VMs running the core business applications in the three production nodes.
So
> even if all three
> Production nodes will go down all the core business applications will keep
> on running.
> Of course, when I will have clarified exactly how to setup the node in
case
> of fault I
> Can try some different locations in case of fault.
> 
> I hope you don't mind having a look at the attached doc. I followed the
> configuration 
> found in "HA virtualization with KVM -ISCSI - Pacemaker" by Florian Hass,
> adding few 
> lines in order to have the VM running only in the production node and if
it
> will fail 
> will keep on running on the backup node. As I have read, that all the
nodes
> must to be active,
> my doubt is that the same VM will try start on two nodes at the same time.

That is the job of Pacemaker to not let this happen, of course every
person with administrative access to the cluster can start machines
manually ....

> 
> Please let me know if the attached configuration can work or there is
> something missing.

You enable stonith but there is no definition for any stonith resource
in this conf and you really must not start a production setup including
shared storage without working stonith!

There is also no need to define scores for the backup_node as the
default of 0 will do its job.

And don't forget to adjust your action timeouts. The default of 20sec is
typically not enough for the migration and stop action of a VM.

> 
> Thank you so much for helping me.

You are welcome!

Regards,
Andreas

-- 
Need help with Pacemaker?
http://www.hastexo.com/now

> 
> My best regards.
> 
> Luciano
> 
> -----Original Message-----
> From: Andreas Kurz [mailto:andreas at hastexo.com] 
> Sent: sabato 8 ottobre 2011 0.44
> To: pacemaker at oss.clusterlabs.org
> Subject: Re: [Pacemaker] Pacemaker multi node configuration
> 
> Hello,
> 
> On 10/05/2011 03:38 PM, Luciano Sitzia wrote:
>> Hello Nick, I have already read this tutorial and for two node
> configuration
>> is quite clear
>> but I didn't get how to setup four primary nodes switching to the same
>> secondary if one of 
>> them will fail.
> 
> First a  question: Why would you want one node being completely idle
> while it could run VMs and help decreasing per node load?
> 
> One solution I could think of for a such a dedicated cold standby server:
> 
> * for every VM resource set a score e.g. 100 for one preferred node out
> of the four servers
> 
> * set a lower score e.g. 10 for every VM for the cold standby server
> 
> ... on node failure (one of "the four"), the first destination for its
> VMs will be the standby node as it has the next best score. The question
> is what should happen on multi node failures .... or do you plan to not
> let them happen? ;-)
> 
> For such a scenario I'd recommend having a look at the utilization
> feature of pacemaker 1.1 in combination with different priorities for
> your vms.
> 
> Regards,
> Andreas
> 
> 
> 
> 
> _______________________________________________
> 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




_______________________________________________
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