[Openais] Corosync can't start pacemaker due to syslog and creates a lots of corosync child processes

[Steven Dake]

On Tue, 2010-03-23 at 13:09 +1000, Thomas Guthmann wrote:
> Hey guys,
> 
> Just posting that again [1] for everybody because I've lost half a day 
> to find why pacemaker wasn't running properly and sending a bunch of 
> lines like :
> 
> Mar 23 12:18:34 db2 crmd: [5543]: info: do_cib_control: Could not 
> connect to the CIB service: connection failed
> Mar 23 12:18:34 db2 crmd: [5543]: WARN: do_cib_control: Couldn't 
> complete CIB registration 24 times... pause and retry
> 
> So that's because corosync is launched _before_ rsyslog. To fix that you 
> just have to start corosync _after_ rsyslog. Quick fix is :
> 
>    vi +9 /etc/init.d/corosync
>      Change all 20 to 99 to have : chkconfig: - 99 99
>    chkconfig corosync off
>    chkconfig corosync on
>    killall -9 corosync && reboot
> 
> I hope it has been fixed in 1.2.1 coming tomorrow if I've understood 
> correctly :) Weirdly enough, I have 32 bits clusters which don't have 
> this problem though corosync is in S20. It happened on my new 64 bits 
> cluster, I don't know if it's related to the architecture though.
> 

This is a distro specific bug.  Please file a bugzilla with the
appropriate distro to work out the runlevels on their system.  For
fedora which I test on mostly, rsyslog is runlevel 12.  Other distros
may be different.  The distributed init script is only a guide - it
isn't perfect for all distros by default.

Regards
-steve


> Cheers,
> Thomas
> 
> [1]: hj lee found the issue the 2nd Feb 2010. See thread "Multiple 
> corosync processes are started" but I could remember when :( Too bad 
> cost me a lot of time.
> 
> 
> 
> 
> _______________________________________________
> Openais mailing list
> Openais at lists.linux-foundation.org
> https://lists.linux-foundation.org/mailman/listinfo/openais