[Linux-HA] Command line option to fail back a master/slave resource

[Andrew Beekhof]

On Thu, Feb 25, 2010 at 4:40 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
> Hi,
>
> On Wed, Feb 24, 2010 at 10:45:16AM -0800, Bob Schatz wrote:

>> QUESTION #2:  Why would these operations abort?  They are serial commands?
>
> We have to ask Andrew about that.

What kind of abort do you mean?
Do you mean the crmd logs?  Thats pacemaker saying "Hold on, something
else changed.  We need to recompute before doing anything else"


>> SS0:1 promotion score on fc12-64-2: 1000000
>> SS0:0 promotion score on fc12-64-1: 1000000
>
>> QUESTION #4:  The promotion score for SS0:0 and SS0:1 is
>> 1000000. How does the CRM use the "location" attribute when
>> making a decision and the scores are the same?
>
> I think it's pseudo-random. Not intentionally though. CRM also
> tries to balance running resources between nodes.

http://www.clusterlabs.org/doc/en-US/Pacemaker/1.0/html/Pacemaker_Explained/node-score-equal.html