Bacula-users

Re: [Bacula-users] multiple mtx-changer scripts

2008-07-30 02:44:51
Subject: Re: [Bacula-users] multiple mtx-changer scripts
From: "Timo Neuvonen" <timo-news AT tee-en DOT net>
To: bacula-users AT lists.sourceforge DOT net
Date: Wed, 30 Jul 2008 09:42:58 +0300
"Mark Nienberg" <gmane AT tippingmar DOT com> kirjoitti viestissä 
news:g651m3$cv4$1 AT ger.gmane DOT org...
> Timo Neuvonen wrote:
>
>>> P.S. Why?  Because I have modified the "load" command to use "btape" to
>>> read the
>>> bacula label on the volume after loading it.  If it fails, then it 
>>> unloads
>>> and
>>> reloads the tape and tries again. This is an attempt to automatically 
>>> fix
>>> the I/O
>>> problem that happens about 10-20% of the times bacula changes a tape in
>>> the middle of
>>> a backup.  Note that it is not a timing problem.  It happens even after
>>> "wait_for_drive" returns ready.  Obviously, when working at bconsole,
>>> there are some
>>> commands that I have to perform on tapes that have no labels, so I need 
>>> a
>>> different
>>> "load" command.  But don't get bogged down thinking about all this, just
>>> think about
>>> the question above, OK?
>>>
>>
>> Could you tell something more about this problem, what hardware you have
>> etc?
>>
>> I have experienced some, possibly similar problems with Exabyte 1U 10 
>> tape
>> autochanger. No way to reproduce the problem, but happens quite often 
>> when
>> tapes need to be changed in real work :-(
>
> Yes, I have an Exabyte 1U 10 tape changer with a VXA 320 drive in it.  The 
> problem
> looks like this:
>
> 19-Jul 03:32 khyber-sd JobId 2235: 3305 Autochanger "load slot 6, drive 
> 0",
> status is OK.
> 19-Jul 03:32 khyber-sd JobId 2235: 3301 Issuing autochanger "loaded? drive 
> 0" command.
> 19-Jul 03:32 khyber-sd JobId 2235: 3302 Autochanger "loaded? drive 0",
> result is Slot 6.
> 19-Jul 03:32 khyber-sd JobId 2235: Error: block.c:995 Read error on fd=3 
> at file:blk
> 0:0 on device "VXA3drive" (/dev/nst0). ERR=Input/output error.
> 19-Jul 03:32 khyber-sd JobId 2235: Error: block.c:995 Read error on fd=3 
> at file:blk
> 0:0 on device "VXA3drive" (/dev/nst0). ERR=Input/output error.
>
> When I first got the changer I saw other errors too, but those were all 
> related to
> timing problems with mtx-changer not waiting long enough (specifically the
> wait_for_drive subroutine).  I have since resolved all of that by tweaking
> mtx-changer and the Max Changer Wait setting in the device resource.  But 
> the error
> above still happens.  As you say, it is not reproducible.  Generally I can 
> use
> bconsole to unmount and then mount the tape again and it will proceed. 
> But this is
> putting a strain on my Saturday mornings, since my full tape backup runs 
> Friday night!
>
> Since I have never been able to find the cause of the problem I have 
> decided instead
> to try to work around it as I described in my initial message.  In effect, 
> I want the
> mtx-changer script to do what I have to do manually (unmount and mount 
> again).
>
> I do daily backups to disk so my tape backups only happen once a week, and 
> since the
> problem doesn't happen every time it can be a month or so before I have 
> real world
> results of the effectiveness of my revisions.  So if you are interested in 
> testing my
> script, let me know.
>

Mark,

The problems you are experiencing look very much the same that I have. I'm 
very interested in testing the changes you have made into the script. I 
couldn't follow from this mail thread if it included all the changes, but 
can you email your script to me (or include it as an attachment to mailing 
list message, I don't know if the list system will support it).

I can not promise any immediate results, since I have the same problem like 
you: while testing, everything seems to work ok, but real tape changes in 
the middle of the night tend to fail. Due to my environment, this happens 
typically only 1-2 times a month.

Pls. also let me know what other changer-related settings you have in the 
conf files, like the MaxChangerWait that you mentioned.


Regards,
Timo



-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>