[ClusterLabs] Good idea to enable Pacemaker blackbox by default?

[Chuck Carmack]

----- Forwarded by Chuck Carmack/Rochester/IBM on 05/27/2015 08:35 AM 
-----

From:   Chuck Carmack/Rochester/IBM
To:     users at clusterlabs.org
Date:   05/22/2015 07:31 AM
Subject:        Good idea to enable Pacemaker blackbox by default?


Hi, thanks for the answers.  Yes I'm thinking about auto-dumping.   Would 
this dump every time pacemaker detects any error, such as a timeout, 
resource agent error, resource agent in wrong state, etc?   Or does it 
only dump for pacemaker internal errors?  I don't want frequent 
auto-dumping.

I'm looking for a way to debug problems that can't be debugged with 
/var/log/messages. If most problems can be debugged with /var/log/messages 
then maybe default blackbox isn't worth it.es

Can you say if the performance impact will be noticeable?


> On 14 May 2015, at 4:17 am, Chuck Carmack <carmack at us.ibm.com> wrote:
> 
> Hi - 
> 
> I'm thinking of enabling the blackbox by default in production 
environments (ie. customer env).   This seems like a good way to capture 
detailed debug logs at the time of error, without enabling debug logging 
all the time.  I would have a cron job to prune 
/var/lib/pacemaker/blackbox in case too many files are generated. 

Ah, you’re talking about the auto-dumping when we log an error?

> 
> Is there a reason this is *not* a good idea?   Does blackbox impact 
cluster performance? 

There will be some impact, all the arguments need to be marshalled etc.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20150527/5fa934db/attachment-0003.html>