[Pacemaker] PE-Engine-Processing

[Andrew Beekhof]

On Jan 18, 2008, at 10:47 AM, Andreas Mock wrote:

>> -----Ursprüngliche Nachricht-----
>> Von: The Pacemaker cluster resource manager <pacemaker at clusterlabs.org 
>> >
>> Gesendet: 18.01.08 08:26:13
>> An: The Pacemaker cluster resource manager  
>> <pacemaker at clusterlabs.org>
>> Betreff: Re: [Pacemaker] PE-Engine-Processing
>>
>> No need to apologize,  lmb and I just need to get our stories
>> straight :-)
>>
>> I'll continue to also answer pacemaker/crm questions on linux-ha but,
>> personally, I'd prefer questions to be sent to this list.
>> Particularly since not everyone will be using the heartbeat stack
>> anymore.
>
> I agree with Lars to not seperate the HA community when not necessary.

I agree in theory, _but_
- it doesn't seem appropriate to hijack the heartbeat mailing list for  
use by people not using heartbeat (ie. the openais guys)
- threats of censorship have been made in the past
- the heartbeat mailing list has already been shut down once because  
someone didn't like what they saw

> I just want to know where to post to. You know some guys don't like
> crossposting.

nod, personally though it doesn't bother me so much.

Just take a guess.  If you send it here and it's more of a heartbeat  
question, we can just bounce it to the heartbeat list.
Generally speaking though, if it involves XML, then this is the right  
place :-)

Oh, and if you have a question about a specific component (eg. the  
lrmd is using a lot of CPU), try to use the list for the relevant  
project (http://clusterlabs.org/#Architecture)

>> In answer to your original question, yeah, there will have been some
>> errors earlier on that might make this particular input "interesting"
>> to an admin/debugger at a later point.
>> So it saves it as pe-warn-* (instead of pe-input-*) and logs that
>> message.
>
> My initial question is not completely answered. Probably I simply  
> don't understand. ;-)
> The log entry is saying: Hey guy, there's something which caused a  
> warning
> in the PE. I save this in pe-warn-* for you.
> O.K. so far so good. But when I look at this file and put it through  
> crm_verify I
> can't see any messages which would explain why the initial warning  
> was issued.

Ah.

crm_verify doesn't do a full simulation (because it doesn't always  
have the status section, I need to make it smarter at figuring out  
when it does), so it doesn't always catch everything.

try ptest instead... that will do exactly what the pengine process does

> So, where do I have to look at to see what caused the warning  
> message in the log.

further back in the log... try grepping for "pengine.*WARN:"