On 2008-08-15T13:13:46, "Krauth, Alexander" <[EMAIL PROTECTED]> wrote:
> Situation:
> Running a two node cluster. A master/slave resource with clone_max = 2,
> clone_node_max = 1, master_max = 1, master_node_max = 1.
> Startup is working, that means 1 clone is running as a master and the
> other clone is running as a slave on the second node. Now I kill the
> application processes of the master resource. The RA recognices this and
> returns OCF_FAILED_MASTER in the next monitor operation.
>
> My expectation:
> 1. Promote the existing slave on the second node to the new master.
> 2. Restart the old master on the first node as a slave.
>
> What happens is:
> 1. Restart the master as master on the first node.
> 2. Nothing happens to the still running slave.
>
> Is this the intended behavior ? Are there any attributes to change it
> (setting some -INFINITY for fail-counts for only _master_ resources) ?
> I could provide a hb_report on this, if needed.
This is because of the crm_master preference.
First, recovery of the slaves is tried - as your first node is still
eligible to run a slave, a slave is started.
Then, the PE looks as the crm_master values, which indicate which side
is preferable. If they are equal, it amounts to a random pick, and you
end up with the first node.
The alternative - promoting the second slave first, then restarting the
slave, and comparing preferences again - might cause the master to shift
once more than necessary.
So the behaviour you describe is expected as of now (and "correct", even
if not optimal in your use case), but we welcome a thought out proposal
of how to configure & describe a better one ;-)
Regards,
Lars
--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
