[Pacemaker] Several problems with simple virtual-IP config

[Klaus Darilion]

Am 09.02.2011 09:48, schrieb Florian Haas:
> On 2011-02-09 09:25, Klaus Darilion wrote:
>> Am 08.02.2011 19:17, schrieb Michael Schwartzkopff:
>>>> Then I changed node server2 to standby: # crm node standby server2.
>>>>>
>>>>>
>>>>>   Node server2: standby
>>>>>   Online: [ server1 ]
>>>>>
>>>>>   failover-ip     (ocf::heartbeat:IPaddr):        Started server1
>>>>>    Clone Set: clonePing
>>>>>        Started: [ server1 ]
>>>>>        Stopped: [ pingtest:0 ]
>>>>>
>>>>> Why is the output "Stopped: [ pingtest:0 ]" and not as I would expect
>>>>> "Stopped: [ server2 ]"?
>>> No. The clone instances are pingtest:0 and pingtest:1.
>>
>> So, when a clone instance is started, the server's name is displayed,
>> and when a clone instance is stooped the instance's name is displayed,
>> somehow inconsistent?
> 
> I believe you're operating on the assumption that clone instances are
> somehow tied to particular nodes, which they are not.
> 
> Pacemaker has no crystal ball. It has no way of knowing what node the
> instance is going to be started on next.
> 
> You could have an 8-node cluster and a clone limited to two instances
> (with meta clone-max=2). Then you'd have instance IDs 0 and 1, and only
> one instance running (for whatever reason), and the currently stopped
> instance would next run on one of 7 different nodes. Why should
> Pacemaker display all 7 node names?

This is a valid point!

> Hope this clears things up.

Yes it does. Thanks
Klaus