Mailing List Archive

Cleanup after ugly shutdown
Hi all.

I recently encountered an incident where rsyslogd died in an "ugly"
manner. This was due to no fault of its own, but simply a matter of
running out of disk space, because an instance, running with a
disk-assisted memory queue, failed to reconnect to the syslog server,
and therefore didn't have anywhere to put the log data, except on its
own hard drive.

When rsyslog dies like this (or in any other way that prevents a decent
shutdown), it obviously do not write a .qi-file, indicating where it
left off, and thus all the data written to disk is not handled when
rsyslogd is started again. This behaviour is correct, as I see it, as at
least some of the data written to disk, almost certainly will be
inconsistent.

But still I would like the opportunity to manually process these data,
so that they eventually will be sent to the syslog server. What is the
smartest way to properly do this? Has it been considered to let rsyslogd
take a command line argument invoking such behaviour, or has someone
written a seperate tool to do something like this?

Any ideas are welcome.


Kind regards,

Jesper Nyerup.

_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com
Re: Cleanup after ugly shutdown [ In reply to ]
Hi Jesper,

one solution is to use the checkpoint interval. This ensures a queue
file is written every n-th time a message is enqueued. This is inside
the queue doc.

On the "restart queue" option, I agree that it would be useful, but I
unfortunately currently neither have time nor funding to implement it.
You may want to add this as a feature request to the bugzilla, so that
it is not forgotten if time arises...

Thanks,
Rainer

> -----Original Message-----
> From: rsyslog-bounces@lists.adiscon.com [mailto:rsyslog-
> bounces@lists.adiscon.com] On Behalf Of Jesper Dahl Nyerup
> Sent: Monday, December 08, 2008 9:06 AM
> To: rsyslog@lists.adiscon.com
> Subject: [rsyslog] Cleanup after ugly shutdown
>
> Hi all.
>
> I recently encountered an incident where rsyslogd died in an "ugly"
> manner. This was due to no fault of its own, but simply a matter of
> running out of disk space, because an instance, running with a
> disk-assisted memory queue, failed to reconnect to the syslog server,
> and therefore didn't have anywhere to put the log data, except on its
> own hard drive.
>
> When rsyslog dies like this (or in any other way that prevents a
decent
> shutdown), it obviously do not write a .qi-file, indicating where it
> left off, and thus all the data written to disk is not handled when
> rsyslogd is started again. This behaviour is correct, as I see it, as
> at
> least some of the data written to disk, almost certainly will be
> inconsistent.
>
> But still I would like the opportunity to manually process these data,
> so that they eventually will be sent to the syslog server. What is the
> smartest way to properly do this? Has it been considered to let
> rsyslogd
> take a command line argument invoking such behaviour, or has someone
> written a seperate tool to do something like this?
>
> Any ideas are welcome.
>
>
> Kind regards,
>
> Jesper Nyerup.
>
> _______________________________________________
> rsyslog mailing list
> http://lists.adiscon.net/mailman/listinfo/rsyslog
> http://www.rsyslog.com
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com