[Pacemaker] Confusing semantics of colocation sets (stopping resource stops others in colocation / order sets)

Michael Chapman mike at very.puzzling.org
Fri Jun 15 12:48:35 UTC 2012


> I...uh...don't really know what to say. Is that a bug, in that it is not 
> reasonable in any way? Or is it a feature, in that it can't be fixed in a 
> backwards compatible way?

Having literally just implemented some logic to handle these quirks today, 
I would certainly hope that any changes to this are introduced through new 
syntax. :-)

What I found was that the logical chaining behaviour can be implemented 
using single-resource sets:

   colocation colo inf: (A) (B) (C D) (E) (F)
   # A -> B -> (C, D) -> E -> F

It's a bit ugly, but if (as in my situation) the config is automatically 
generated, it's workable.

It turns out that:

   colocation colo inf: A B

and:

   colocation colo inf: (A) (B)

effectively do the same thing, but in different ways. That is, *only* the 
"two sequential resources" case differs from all the others. I suspect 
that's due to it generating just the single CIB XML element:

   <rsc_colocation id="colo" rsc="A" with-rsc="B" score="INFINITY" />

rather than a <rsc_colocation> containing a series of <resource_set> 
elements.

- Michael




More information about the Pacemaker mailing list