Re: [Bacula-users] jobs fail with various "broken pipe" errors

[Silver S. <si...@se...>]

On Tuesday 28 February 2012 08:26:50 Phil Stracchino wrote:
> On 02/28/2012 05:00 AM, Silver Salonen wrote:
> > I look at these lines:
> > 
> > 28-Feb 00:04 backupsrv-dir JobId 58414: Sending Accurate information.
> > 28-Feb 00:07 backupsrv-sd2 JobId 58414: Recycled volume "fbsd1-userdata-incr-6106" on device "device-fbsd1" (/mnt/backup/bacula/fbsd1), all previous data lost.
> > 28-Feb 00:07 backupsrv-dir JobId 58414: Max Volume jobs=1 exceeded. Marking Volume "fbsd1-userdata-incr-6106" as Used.
> > 28-Feb 00:40 backupsrv-dir JobId 58414: Fatal error: Network error with FD during Backup: ERR=Broken pipe
> > 28-Feb 00:40 backupsrv-sd2 JobId 58414: JobId=58414 Job="fbsd1-userdata.2012-02-27_23.05.02_42" marked to be canceled.
> > 28-Feb 00:40 backupsrv-dir JobId 58414: Fatal error: No Job status returned from FD.
> > 28-Feb 00:40 backupsrv-dir JobId 58414: Error: Bacula backupsrv-dir 5.2.3 (16Dec11):
> > 
> > ...and wonder why does DIR get a "Network error" after 33 minutes.
> 
> 33 minutes and 20 seconds, I'll bet.  That's 2000 seconds.
> 
> So what has a 2000-second timeout?

Yeah, that's what I meant by "Does DIR take up a new network session for that? It wouldn't be possible otherwise, would it?" - I don't think there's a persistent connection between DIR--FD while the job is running, ie. data is sent from FD to SD (it just wouldn't make sense).

On another job's log there are these lines:

27-Feb 23:19 winsrv1-fd JobId 58380: Generate VSS snapshots. Driver="VSS Vista", Drive(s)="EC"
28-Feb 01:37 winsrv1-fd JobId 58380: Error: /home/kern/bacula/k/bacula/src/lib/bsock.c:394 Wrote 8992 bytes to Storage daemon:backupsrv.url:9103, but only 0 accepted.
28-Feb 01:37 winsrv1-fd JobId 58380: Fatal error: /home/kern/bacula/k/bacula/src/filed/backup.c:1190 Network send error to SD. ERR=Input/output error

Which indicates that the job hast lasted over 2 hours when it finally fails to send some packets for some reason.

--
Silver