Some questions:
Is this one dump that's spanning the two tapes?
Or are you doing separate dumps to each tape?
Are the tapes being changed by hand or by a tape library/robot/stacker/
etc?
The reason I ask is that tape libraries take a certain amount of time
to do the tape movement and then the tape drive takes some more time
to load the tape and become ready. If the backup machine tries to
access the tape drive before it's ready you could see (and I have
seen) this sort of result. A common workaround is to add some fixed
amount of sleep time, or to use a loop that polls the drive and then
sleeps a few seconds before polling again in order to discover when
the drive is ready.
-Mitch
On Fri, 15 Dec 2000, Stephen Joyce wrote:
> On Fri, 15 Dec 2000, John W. Sopko Jr. wrote:
>
> > We have a strange intermittent problem doing AFS backups ever since
> > we moved to AFS 3.6. Has anyone else experienced this problem?
> > Any input would be appreciated.
> >
> > We do backups from our Sun E250's to Sun standalone DLT 7000 tape
> > drives. We are running Solaris 2.7 with AFS Base configuration
> > "afs3.6 2.3".
> >
> > Our full backups require 2 tapes. The first tape always completes
> > fine. The second tape always continues fine. But, sometimes the second
> > tape "freezes". That is the backups stop running on the second tape.
> > There are no messages in the butc window, we run butc
> > with a debug level of 1. There are no messages in the
> > /usr/afs/backup/TE_rmt_0u or TL_rmt_0u logs. There are no messages on
> > the console or the system log /var/adm/messages. The only clue is
> > in the backup prompt:
> >
> > backup> status
> > Task 1: Dump: 11033914 Kbytes transferred, volume
> > D1.prj.stm.da.gt.nM.fi.backup
> > Task 2: Dump: [drive wait]
> >
> > We start the entire dump over and it usually works the next time.
> >
> > I realize there is a new AFS 3.6 patch release. I read the
> > readme.txt and it did not indicate a fix for this. Any ideas?
>
> John,
>
> According to the AFS docs, "[drive wait]" occurs when "The operation is
> waiting for the specified tape drive to become free". (See
> http://www.transarc.ibm.com/Library/documentation/afs/3.5/unix/adm/adm12.htm
> and search for "drive wait")
>
> You don't have any processes which might be disrupting or restarting butc
> or otherwise trying to use the tape drive at the same time do you?
>
> FWIW, we have a similar setup next door in Phillips (3 E250 servers,
> sun4x_57, DLT 7000 drives, afs3.6 2.3, backups take multiple tapes...) and
> have not experienced this particular problem.
>
> I would be interested to know if installing afs3.6 2.5 solves your
> problem (or what the ultimate solution is)...
>
> Cheers,
> Stephen
> --
> Stephen Joyce
> Systems Administrator P A N I C
> Physics & Astronomy Department Physics & Astronomy
> University of North Carolina at Chapel Hill Network Infrastructure
> voice: (919) 962-7214 and Computing
> fax: (919) 962-0480 http://www.panic.unc.edu
>
>
