[Pacemaker] Log STDERR from OCF scripts

[Andrew Beekhof]

On Fri, Jan 4, 2013 at 8:30 PM, Michal Fiala <fiala at mfiala.net> wrote:
> Hallo,
>
> thanks for information. We are using corosync-1.4.4 and
> pacemaker-1.1.6.1.

Oh, thats the old lrmd.  In which case your original assessment could
well be correct.

> Pacemaker is started as a modul by corosync.
>
> corosync loggin option:
>
> logging {
>         fileline: off
>         to_stderr: no
>         to_logfile: no
>         logfile: /var/log/corosync.log
>         to_syslog: yes
>         syslog_facility: local0
>         debug: on
>         timestamp: on
>         logger_subsys {
>                 subsys: AMF
>                 debug: off
>         }
> }
>
> and syslog is loging all levels of local0:
>
> local0.* -/var/log/corosync.log
>
> This this setting correct?
>
> Because with this setup, we do not have stderr from ocf linbit/drbd in
> corosync.log.
>
> Thanks
>
> Michal
>
> On 01/02/2013 01:00 PM, Andrew Beekhof wrote:
>> On Fri, Dec 21, 2012 at 11:16 PM, Michal Fiala <fiala at mfiala.net> wrote:
>>> Hallo,
>>>
>>> we use corosyng logging via syslog (to_logfile: no; to_syslog: yes;
>>> syslog_facility: local0; debug: on). Some OCF scripts do not use OCF API
>>> to execute commands. I mean function ocf_run, which capture STDOUT and
>>> STDERR. For example linbit/drbd uses its own function to execute
>>> do_cmd(), which does not capture STDERR. So you can not see STDERR
>>> output in corosync log file.
>>> I have discuss this problem on drbd-user mailing list and they mean,
>>> that capturing STDERR should be done by lrmd.
>>>
>>> Communication from drbd-user mailing list:
>>>
>>> Q: Is there a way how to capture stderr from
>>> /usr/lib/ocf/resource.d/linbit/drbd? I can simply patch ocf script, or
>>> better this could be a global, permanent fix from developers.
>>>
>>> A: We deliberately chose to do it this way.
>>> Besides the drbd.ocf predates ocf_run(), I think,
>>> and is supposed to also work on clusters not providing this.
>>>
>>> The cluster-glue lrmd used to capture both stdout and stderr,
>>> and logged it accordingly.
>>>
>>> The pacemaker 1.1.8 rewritten-from-scratch lrmd
>>> apparently does not (yet).
>>
>> Not true.  It is both captured and logged.  The severity depends on a
>> number of factors.
>> So you might need to enable debug to get stderr from actions that didn't "fail".
>>
>>>
>>> If you "fix" our do_cmd() in the way you propose to,
>>> you'll break the drbd ocf resource agent,
>>> because we use it in nested ways: X=$(do_cmd something else),
>>> and expect *exactly* the stdout of somthing else there.
>>>
>>> So don't do that.
>>>
>>> If anything, this needs to be fixed in the rewritten-from-scratch
>>> pacemaker lrmd.
>>>
>>> details see:
>>> http://lists.linbit.com/pipermail/drbd-user/2012-December/019416.html
>>>
>>> How can I solve capturing STDERR?
>>>
>>> Thanks
>>>
>>> Michal
>>>
>>>
>>> _______________________________________________
>>> 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
>>
>> _______________________________________________
>> 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
>>