José,
I don't have any reason to believe this is specific to Bacula on Debian, so I
figured you likely would just kick this along to Kern; it just happens that at
the point I got fed up enough to file the report, it was easier to use
reportbug on the command line than visit bugs.bacula.org :-)
I certainly will upgrade to 1.36.3 as soon as it transitions to testing, but in
this case I may go ahead and pull it from unstable as bacula has been pretty
good to me :-) and I don't know if it will get caught up in the sarge freeze or
whatever. Frankly I doubt it will do me any good, as I don't recall seeing any
traffic regarding this on the lists and all of the previous 1.36 releases have
had identical behavior.
My best guess still is that this got triggered by my kernel 2.6 upgrade and
some sort of behavior change perhaps in the mt driver, but it completely stumps
me how nobody but myself could see this problem, and why btest doesn't have
problems appending but bacula-sd does.
I am more than game to assist with debugging, since I can reproduce the problem
at will -- I just don't know what sort of logging etc. would be most useful (or
how to enable it).
Thanks again for your hard work in packaging Bacula for the Debian universe,
Scott
mailto:[EMAIL PROTECTED]
-----Original Message-----
From: José Luis Tallón [mailto:[EMAIL PROTECTED]
Sent: Tuesday, May 10, 2005 6:15 AM
To: R. Scott Bailey; [EMAIL PROTECTED]
Subject: Re: Bug#308377: btest passes but bacula-sd cannot append under Linux
2.6.x
Hi, Scott!
R. Scott Bailey wrote:
>Package: bacula-sd
>Version: 1.36.2-2
>Severity: normal
>
>
>I noticed this change in behavior when I upgraded my system from kernel
>2.4.27 to 2.6.10. Attempting to position to the end of data on a tape in
>order to append backups no longer works:
>
><unmount>
><mount>
><run something>
>09-May 16:02 hamster-sd: Volume "S12021" previously written, moving to
>end of data.
>09-May 16:04 hamster-sd: ahbase02.2005-05-09_16.02.45 Error: Unable to
>position to end of data on device "/dev/nst0". Err=dev.c:494 ioctl MTEOM
>error on /dev/nst0. ERR=Input/output error.
><etc>
>
>Everything works fine as long as I don't position to the end of the
>tape; for example, I can purge the volume and then bacula is happy to
>reuse it, writing from the beginning, and I can continue appending jobs
>until it fills up, just like I expected. Restores (where I position into
>the middle of the data) work fine too. I have run btest and it claims
>that everything is fine (and particularly the append tests all pass).
>
>This problem started specifically when I upgraded my Linux kernel from
>2.4 to 2.6. Having spooling enabled or disabled does not affect the
>problem. Upgrading through the various 1.36.* versions of Bacula did not
>affect the problem.
>
>
What can i do for you w.r.t. this bug? Shall i forward it to Kern?
In which ways do you think it is specific to Debian's compilation of Bacula?
I am about to upload Bacula-1.36.3 very soon... this *might* help.
>Here is the device stanza from my /etc/bacula/bacula-sd.conf file:
>
>Device {
> Name = TZ89
> Media Type = TZ89
> Archive Device = /dev/nst0
> AlwaysOpen = yes
> RemovableMedia = yes
> RandomAccess = no
> Spool Directory = "/var/spool/backup"
> Maximum Spool Size = 10000000000 # A little below 10G
> Maximum Job Spool Size = 2000000000 # A little below 2G
>}
>
>
>
Thanks for the feedback. It is much appreciated.
J.L.
