[Pacemaker] Using crm_resource leaks socket descriptors in cib?

[Andrew Beekhof]

On 09/04/2013, at 3:58 PM, Taneli Leppä <taneli at crasman.fi> wrote:

> Hello,
> 
> I updated last night from RHEL 6.3 to 6.4 (eg. from Pacemaker
> 1.1.7 to 1.1.8). A few hours later I got into a bit trouble
> when the cib process ran out of file handles.
> 
> After restarting stuff, I noticed that /proc/cib-pid/fd/ list
> was steadily growing by a few UNIX socket handles every
> few minutes or so.
> 
> I pinpointed this to my monitoring agent (Zabbix) which uses
> crm_resource to query status of certain services (eg. crm_resource
> -q -r Redis_replication).

Wouldn't -W be a better option than -q?

> After changing crm_resource to 'crm
> resource status' calls, the descriptor list doesn't grow anymore.

That is indeed weird.
I wonder what command it is using instead.

> 
> Could this be related to the way crm_resource is called? I couldn't
> replicate this by calling crm_resource manually from the prompt.
> 
> -- 
>  Taneli Leppä       | CISSP, RHCE, ZCE, CMDEV
>  Crasman Oy         | taneli at crasman.fi
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org