Looks like it was the OOM reaper killing the task. Thanks for pointing
me in the right direction Brian!

David Gersting
Linux Systems Administrator
WVU Information Technology Services

On 12/12/16 1:13 PM, Brian Bouterse wrote:
> I think the celery worker is experiencing a segfault or maybe it's
> being killed by the OOM. If the OOM is killing it there would be log
> evidence. If it's a segfault, with Python a segfault is unlikely, so
> this is probably a segfault while calling to the system using
> subprocess which Pulp does in various places. I haven't looked in the
> publish code of platform and rpm to look for subprocess usage but that
> would probably hint at the problem. To really debug something like
> that you would want capture a coredump. I think celery has the ability
> to capture coredumps, but I've never done it.
>
> The pulp-smash tests for publish showed they were working. Is it
> possible that this could be an environment issue? Is it possible to
> reproduce the issue on separate hardware to rule that out. If it is
> reproducable, I recommend opening a bug [0].
>
> [0]: https://pulp.plan.io/projects/pulp/issues/new
> <https://pulp.plan.io/projects/pulp/issues/new>
>
> -Brian
>
> On Mon, Dec 12, 2016 at 11:49 AM, David Gersting
> <[email protected] <mailto:[email protected]>> wrote:
>
>     Hello everyone,
>
>     I've been banging my head against the desk for a while on this
>     one, and
>     could use the group's help.
>
>     I have a rather large repo (OEL 6's base repo with 36,684 RPMs)
>     that I'm
>     trying to mirror locally to speed up our os patching, and every time I
>     try to publish the repo the task fails just after the "Publishing
>     Delta
>     RPMs" step starts. After some digging it seems to me that the
>     worker is
>     timing out. Has anyone else seen this and/or know how I can fix it or
>     increase the timeout for this task?
>
>     I've attached the full shell output for anyone who wants it, but the
>     error message I'm seeing from the worker is:
>     # journalctl --unit=pulp_worker-5
>         *SNIP*
>         Dec 12 10:48:19 *HOSTNAME* pulp[1403]: celery.worker.job:ERROR:
>     (1403-27776) Task
>     
> pulp.server.managers.repo.publish.publish[e3d25854-757c-40af-8979-d0b7287263ed]
>     raised unexpected: WorkerLostError('Worker exited prematurely:
>     signal 9
>     (SIGKILL).',)
>         Dec 12 10:48:19 *HOSTNAME* pulp[1403]: celery.worker.job:ERROR:
>     (1403-27776) Traceback (most recent call last):
>         Dec 12 10:48:19 *HOSTNAME* pulp[1403]: celery.worker.job:ERROR:
>     (1403-27776)   File
>     "/usr/lib64/python2.7/site-packages/billiard/pool.py", line 1171, in
>     mark_as_worker_lost
>         Dec 12 10:48:19 *HOSTNAME* pulp[1403]: celery.worker.job:ERROR:
>     (1403-27776)     human_status(exitcode)),
>         Dec 12 10:48:19 *HOSTNAME* pulp[1403]: celery.worker.job:ERROR:
>     (1403-27776) WorkerLostError: Worker exited prematurely: signal 9
>     (SIGKILL).
>         Dec 12 10:48:21 *HOSTNAME* pulp[49191]: py.warnings:WARNING:
>     (49191-27776)
>     /usr/lib64/python2.7/site-packages/pymongo/topology.py:74:
>     UserWarning: MongoClient opened before fork. Create MongoClient with
>     connect=False, or create client after forking. Se
>         Dec 12 10:48:21 *HOSTNAME* pulp[49191]: py.warnings:WARNING:
>     (49191-27776)   "MongoClient opened before fork. Create MongoClient "
>         Dec 12 10:48:21 *HOSTNAME* pulp[49191]: py.warnings:WARNING:
>     (49191-27776)
>         Dec 12 10:48:22 *HOSTNAME* pulp[49191]:
>     pulp.server.async.tasks:INFO: Task failed :
>     [e3d25854-757c-40af-8979-d0b7287263ed]
>
>
>
>     Any help would be much appreciated!
>
>     --
>     David Gersting
>     Linux Systems Administrator
>     WVU Information Technology Services
>
>
>     _______________________________________________
>     Pulp-list mailing list
>     [email protected] <mailto:[email protected]>
>     https://www.redhat.com/mailman/listinfo/pulp-list
>     <https://www.redhat.com/mailman/listinfo/pulp-list>
>
>

_______________________________________________
Pulp-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/pulp-list

Reply via email to