[Pacemaker] N+1 and equal priority resource groups

Andrew Beekhof andrew at beekhof.net
Tue Apr 22 21:39:07 EDT 2014

On 23 Apr 2014, at 1:46 am, Igal Baevsky <ibaevsky at marketfactory.com> wrote:

> Andrew Beekhof <andrew at ...> writes:
>> Correct. 
>> Given colocate(A, B, -inf), in order to find out where A can go, we need 
> to know where B is (going to go).
>> Even if we made it so that was no longer the case (which at a stretch 
> might even be possible after all these
>> years), there is still an implicit ordering from their order in the 
> configuration.
>> Unfortunately the human brain is still significantly more intelligent than 
> Pacemaker.
>> What seems obvious to us is not at all obvious to it.
>> Can all three groups run on all nodes?  Or just its own and the shared 
> backup?
> Ideally, I would want all three groups to be able to run on all nodes based 
> on pre-assigned priorities(nodes might have different hardware specs).
> What I want to prevent though, is a situation where one group is being shut 
> down or moved to make space for another group.

So sumarizing your setup as group-{1,2,3}; node-{1,2,3,4} where node-4 is the hot spare; and that group-N prefers node-N...
In what scenario would group-1 legitimately end up on node-2 or node-3?

> This is exactly what's 
> happening now when all nodes are taken.
> _______________________________________________
> 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://bugs.clusterlabs.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20140423/a764b6a8/attachment-0003.sig>

More information about the Pacemaker mailing list