Re: [Bacula-devel] [Bacula-users] bacula-sd volume data errors during copy from disk to tape

[Pasi K. <pa...@ik...>]

On Mon, Feb 09, 2009 at 04:58:02PM +0100, Kern Sibbald wrote:
> These kinds of errors are not good as I said before.  Something is going 
> wrong.
> 
> It seems to me that you are writing to network mounted Volumes. Please remind 
> me if this is the case.  If it is, I am not willing to spend any time on this 
> unless you can reproduce it on a locally mounted Volumes.  
> 
> Bacula does a lot of seeking around (forwards and backwards) on Volumes, and 
> network mounted filesystems in my experience (both Samba and NFS) do not 
> properly implement standard Unix filesystem calls (Samba does opportunistic 
> locking so data is inconsistent on the two sides, which is absurd IMO; and 
> NFS at a minimum does not implement ftruncate; one or both are or were 
> limited to 4GB addresses -- who knows what other horrors are in their code).
> Doing a simple copy or a simple sequential read will surely work, but I'm not 
> sure they can handle Bacula.
> 

I'm using ext3 on iSCSI LUN. So all the filesystem operations work just like
with locally attached storage. 

I don't have any SCSI/iSCSI errors in dmesg/syslog.. 


> See below ...
> 
> On Monday 09 February 2009 15:52:52 Pasi Kärkkäinen wrote:
> > On Thu, Jan 29, 2009 at 06:01:52PM +0100, Kern Sibbald wrote:
> > > OK, thanks for the feedback.  From what I read, the problem is now
> > > resolved. If it is, OK no need to respond.  If it is not, please let me
> > > know.
> >
> > Actually it seems I'm unfortunately still getting these errors.. now it's
> > with disk volumes created and also being copied (to tape) with the same
> > Bacula 2.5.29 version.
> >
> > Log of the copy job:
> > http://pasik.reaktio.net/bacula/debug/bacula-copy-job-volume-data-errors.tx
> >t
> >
> > (Also attached to this mail).
> >
> > Original job:
> >
> > *list jobid=9126
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> > | JobId | Name    | StartTime           | Type | Level | JobFiles |
> > | JobBytes       | JobStatus |
> >
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> > | 9,126 | server1 | 2009-02-07 08:03:07 | B    | F     |  209,761 |
> > | 35,221,140,175 | T         |
> >
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> >
> > Copied job:
> >
> > *list jobid=9178
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> > | JobId | Name    | StartTime           | Type | Level | JobFiles |
> > | JobBytes       | JobStatus |
> >
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> > | 9,178 | server1 | 2009-02-07 08:03:07 | C    | F     |  209,761 |
> > | 35,259,190,600 | T         |
> >
> > +-------+---------+---------------------+------+-------+----------+--------
> >--------+-----------+
> >
> >
> > So.. at least the amount of files is the same for the original job and the
> > new copied job.
> >
> > I'm also wondering about the "Termination: Copying OK" when there are
> > _errors_ during the copy?
> 
> Bacula believes that it has recovered from the Error.  If the SD Errors line 
> does not show some count, please report it as a bug.
> 

  SD Errors:              0
  SD termination status:  OK
  Termination:            Copying OK


-- Pasi