[Pacemaker] Cannot start VirtualDomain resource after restart

[Phil Frost]

On 06/20/2012 01:09 PM, Kadlecsik József wrote:
> On Wed, 20 Jun 2012, Phil Frost wrote:
>
>> Firstly, I'd try running "crm_simulate -LS -D pacemaker.dot", then viewing
>> the generated pacemaker.dot with graphviz [1] (specifically "dot". It might
>> also be helpful to pass pacemaker.dot through "tred" first, to make it more
>> readable). This asks crm_simulate to simulate what pacemaker would like to do
>> (-S), given the current live state (-L). Probably it will tell you it would
>> do nothing, because it's already running in the desired (by pacemaker,
>> anyway) state. However, I have seen instances in testing where Pacemaker will
>> be stuck in some start ->  monitor ->  timeout loop that's not immediately
>> obvious in crm_mon. This will reveal that.
> I have tried it (attached the resulted .dot file). Looking at the graph, I
> couldn't really spot why lx0 isn't started. :-(

Well, if the dot file you attached is the output of "crm_simulate -LS -D 
pacemaker.dot", then this at least tells you that the policy engine, 
given the current state of things, would like to do something else. 
Normally when you run this you get an empty graph, because normally the 
policy engine should be able to reach the target state that it 
calculates. It looks like in your case it is not.

I really have no idea why lx0 is starting, but you can see from the 
graph on what actions it depends. It's likely one of them is failing, 
forcing the policy engine to recalculate its actions. We can see that 
the policy engine thinks that in order to start lx0, it first has to 
migrate mail0, then migrate caladan. It's also interesting that there 
are a whole bunch of other migrations it thinks are necessary. Without 
being intimately familiar with your environment it's hard to say if this 
is expected or not. If you aren't expecting that, then you must have 
some constraints configure that aren't what you intended.