[Pacemaker] cluster deadlock/malfunctioning

[Andrew Beekhof]

On Wed, Nov 19, 2008 at 13:54, Raoul Bhatia [IPAX] <r.bhatia at ipax.at> wrote:
> Andrew Beekhof wrote:
>>> i took a look at pe-warn-12143.bz2 but do not know how to interpret the
>>> three different threads i see in the corresponding .dot file.
>>
>> If there are no lines between sets of actions, then Pacemaker thinks
>> they're unrelated (from an ordering perspective).
>> Arrows indicate a (first) --> (then) relationship.
>>
>> Ones with orange text are place-holders
>> Ones with black text are real actions
>> Ones with green borders are actions we'll try to execute
>> Ones with dashed red borders are actions we'd like to execute but
>> can't (for some reason)
>> Anything with an arrow pointing to it from dashed-red action will
>> never be run.
>
> ok, thank you for explaining it again :)
>
> how do i know why we wanted to run the placeholder
> "group_www_data_start_0" but we did not?

invoke ptest with lots of -V arguments and use grep :-)

>
>> Apologies for not replying to your other problem also, I'm simply
>> swamped at the moment.
>
> ok, no problem. i just would like to know if i shall leave the cluster
> in the current state or if i can upgrade to pacemaker 1.0.1, continue
> testing, etc.

upgrading is fine - and may even help

> ps: i am including the .dot file so that you do not have to extract it
> again.

the cib from which the .dot was generated is actually better :-)