[Pacemaker] Differences in scoring between similar resources

[Andrew Beekhof]

On Tue, Aug 23, 2011 at 2:44 AM, Bobbie Lind <blind at sms-fed.com> wrote:
> I tested this out on our duplicate dev system to see if it was old scores or
> a bad configuration.  I got the exact same ptest scores on a new cluster set
> up exactly the same way.  This lead me to believe that it IS my
> configuration that needed to be changed.  So I started looking at what was
> different.  I changed the following lines of my configuration:
>
> from:
>
> colocation colocOSS1OSS2 -inf: anchorOSS2 anchorOSS1
> colocation colocOSS1OSS4 -inf: anchorOSS4 anchorOSS1
> colocation colocOSS1group 300: ( resOST0000 resOST0004 resOST0008 )
> anchorOSS1
> colocation colocOSS2OSS3 -inf: anchorOSS3 anchorOSS2
> colocation colocOSS2group 300: ( resOST0001 resOST0005 resOST0009 )
> anchorOSS2
> colocation colocOSS3OSS4 -inf: anchorOSS4 anchorOSS3
> colocation colocOSS3group 300: ( resOST0002 resOST0006 resOST000a )
> anchorOSS3
> colocation colocOSS4group 300: ( resOST0003 resOST0007 resOST000b )
> anchorOSS4
>
>
> to:
>
> colocation colocOSS1OSS2 -inf: anchorOSS2 anchorOSS1
> colocation colocOSS1group 300: ( resOST0000 resOST0004 resOST0008 )
> anchorOSS1
> colocation colocOSS2OSS3 -inf: anchorOSS3 anchorOSS2
> colocation colocOSS2group 300: ( resOST0001 resOST0005 resOST0009 )
> anchorOSS2
> colocation colocOSS3OSS4 -inf: anchorOSS4 anchorOSS3
> colocation colocOSS3group 300: ( resOST0002 resOST0006 resOST000a )
> anchorOSS3
> colocation colocOSS4OSS1 -inf: anchorOSS1 anchorOSS4
> colocation colocOSS4group 300: ( resOST0003 resOST0007 resOST000b )
> anchorOSS4
>
> This fixed the scoring error!  Now my question is why?

The order resources are listed in a colocation constraint is significant.

   colocation colocOSS4OSS1 -inf: anchorOSS1 anchorOSS4

means: figure out where to put anchorOSS4, and then put anchorOSS1 there too.
anchorOSS4 will inherits anchorOSS1's node preferences to make it more
likely that anchorOSS4 picks a node they'll both be happy with.

resource stickiness also plays a part.


> What was so screwy about the original set up that changing it made the
> difference?
>
>
> Bobbie Lind
> Systems Engineer
> Solutions Made Simple, Inc (SMSi)
>
>
>
> _______________________________________________
> 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
>
>