[ejabberd] ejabber crash

Tiaan tiaanh at gmail.com
Fri Nov 18 15:12:53 MSK 2005


Thanks for your response.

You seem to be unwilling to isolate what causes the problem. If you
would, you would experiment with 1 node running another OS, same
ejabberd version/config. Then you'd see if the other OS would exhibit
the same problems.

We are busy testing on BSD, using kernel polling to see the effect on CPU
usage. But this specific problem is not easy to replicate, it runs fine for
2 weeks before going into this state.

So it's good to monitor that you can make a TCP connection, or actual
XMPP connection.

I wrote an monitoring utility that logs into jabber, etc, if it can't login
within a timeout its sends an text message our mobile phones. I know about
any issues immediatly :)

I agree, it would be strange for ejabberd to restart the VM, but the wording
of the messages put some doubt in my mind.

On 11/18/05, Andreas van Cranenburgh <andreas at unstable.nl> wrote:
>
> You seem to continue to be dealing with a Redmond related problem :)
>
> I'm curious why the OS needs to be from Redmond, but that's not my
> business of course.....
>
> On Fri, Nov 18, 2005 at 01:14:13PM +0200, Tiaan wrote:
> > I got the following error messages in the Windows event logs, logged by
> > ErlSrv.
> >
> > ejabberd: Restarted erlang machine.
> > ejabberd: Erlang service restarted
> >
> > Is this ejabber restating the erlang VM, or the erlang VM restarting
> > ejabberd ?
>
> [Without much erlang experience] I'd say it's the latter. Think
> hierarchically :)
>
> ejabberd is a child process running in the VM of erlang, so it would be
> a bit weird, if it were the former. Remember, ejabberd == non-native
> erlang code, erlang == native (C code right?). (correct me if I'm
> wrong).
>
> > This is the second time it happened, the servers was under heavy load
> both
> > times when it occurred. It seems like when one crashes, the second one
> > crashes a few minutes later. Crash dumps were created, "erl_crash.dump".
> I
> > had to stop both, and then start them up again to get it working again.
>
> You seem to be unwilling to isolate what causes the problem. If you
> would, you would experiment with 1 node running another OS, same
> ejabberd version/config. Then you'd see if the other OS would exhibit
> the same problems.
>
> Ok, you wouldn't be sure yet, since it may be the OS, or the particular
> erlang binary.
>
> ejabberd is supposed not to need monitoring (eg. monit, like I described
> in another thread), but it _has_ happened once or twice on unstable.nl<http://unstable.nl>
> (ever since I run it, so like 1 or 2 years), that it simply stopped
> accepting connections, without crashing! Very nasty. That wasn't with my
> current install though, it wasn't with the official Debian packages!
>
> So it's good to monitor that you can make a TCP connection, or actual
> XMPP connection. See jabbernet.dk <http://jabbernet.dk>'s site for a good
> example, with mrtg!
>
> > Is there a dump analysis tool available for erlang to analyze the dump
> > file ?
>
> Should you be asking ejabberd? Maybe an erlang community can help you
> much better with that particular erlang related question?
>
> --
> Andreas [ http://unstable.nl | xmpp:andreas at unstable.nl ]
> [ callto:ils.seconix.com/andreas at unstable.nl<http://ils.seconix.com/andreas@unstable.nl>]
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (GNU/Linux)
>
> iD8DBQFDfbtiAcXr2MrY86MRApJTAJ9RjExepMThoyRlD7uTG/nY1KemqACgqOvB
> r8Z5u6lqbI7vzchfDL1iqT8=
> =ou9b
> -----END PGP SIGNATURE-----
>
>
> _______________________________________________
> ejabberd mailing list
> ejabberd at jabber.ru
> http://lists.jabber.ru/mailman/listinfo/ejabberd
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jabber.ru/pipermail/ejabberd/attachments/20051118/8f72cede/attachment.html


More information about the ejabberd mailing list