[Linux-cluster] [UPDATE] IP monitor failing periodically

[Aravind Parchuri]

I'm not sure about the segfaults, but we are facing the same issues on 
RHEL5 and FC6, i368 - random failovers due to ip-check failures. This 
workaround seems to help, for now at least:

http://www.redhat.com/archives/linux-cluster/2006-March/msg00329.html

I'll check if it is indeed the ping segfaulting and report back when I 
get some time.

Aravind.

chris at cmiware.com wrote:
> We reinstalled our machines with RHEL 5 x86_64 (we were running i386) a 
> few weeks ago and the mysterious IP monitoring failures have 
> disappeared. I believe it was postulated that a compiler bug regarding 
> -fpie might be causing segfaults in i386 binaries, so this would support 
> that theory to some degree, although I did not really attempt to confirm 
> it further.  I thought the architecture change fixing the random 
> failovers was noteworthy.
> 
> ### previous thread below
> 
> Hi Chris,
> 
> I am experiencing the same problem on RHEL 5 and I have a support 
> request in with RedHat.
> 
> I was asked to increase the debug level by changing the <rm> line in the 
> cluster configuration to:
> 
> <rm log_facility="local4" log_level="7">
> 
> I then needed to add "local4.* /var/log/cluster" to /etc/syslog.conf and 
> run "service syslog restart".
> 
> To update the cluster configuration I needed to propagate the cluster 
> configuration to both nodes:
> 
> # ccs_tool update /etc/cluster/cluster.conf
> 
> After a week I have not had the problem with the increased logging 
> despite the problem occurring regularly prior to that - 2 to 3 times a 
> day. One day last week out of curiosity I reverted to the default 
> settings and within a few hours I had the failure to ping error on one 
> of the clustered IP addresses and the service was restarted.
> 
> I now have the logging back at 7 and the support request is pending.
> 
> Regards