[ClusterLabs] corosync.log is 5.1GB in a short period

[Klaus Wenninger]

On 08/23/2016 08:31 AM, Kristoffer Grönlund wrote:
> 朱荣 <zhu.rong at eisoo.com> writes:
>
>> Hello:
>> I has a problem about corosync log, my corosync log is increase to 5.1GB in a short time.
>> Then I check the corosync log, it’s show me the same message in short period,like the attachment.
>> What happened about corosync? Thank you!
>> my corosync and pacemaker is:corosync-2.3.4-7.el7.x86_64 pacemaker-1.1.13-10.el7x86_64
>>                                                               by zhu rong
> Something is leaking file descriptors in libqb, but the information
> provided does not yield any further hints.

Remember having had an fd-leak in libqb with pacemaker_remote.
Update of libqb did the trick. Maybe it is something generic.
But I would guess that something else is wrong as well - in my
case it was an error-case leaking fds over time - pacemaker_remote
not connected by a cluster-node.

>
> Check the earlier logs to see if you can find any indication as to what
> may be causing the file descriptor leak.
>
> Use crm_report to collect log data and other information from your
> cluster nodes.
>
> Finally, attach log file excerpts as text attachments, not images.
>
> Cheers,
> Kristoffer
>