[699] in Info-AFS_Redistribution

home help back first fref pref prev next nref lref last post

tapeconfig-values for a Sun-Exabyte combination?

daemon@ATHENA.MIT.EDU (Hakan Torbjar)
Fri Apr 3 10:58:41 1992

Date: Fri, 3 Apr 92 14:25:11 GMT
From: Hakan Torbjar <torbjar@utc.chalmers.se>
To: info-afs@transarc.com

I have spent some time in an attempt to do a multi-tape dump
(2390146 KB from 96 volumes) to an Exabyte EXB-8200 on a Sun (sun4c_411).
The Sun is an AFS client via the NFS/AFS translator, but that
shouldn't matter.

When everything goes almost well(*), the dump is aborted because of an
end-of-tape condition.  I have run `fms', which reported reasonable
values as far as I can tell:

    fms test started
    wrote 130440 blocks
    Tape capacity is 2137128960 bytes
    File marks are 1908957 bytes

and I use the 15%-rule to create the tapeconfig file:

    1816559616 2195300 /dev/rst1 0

Is this maybe a problem with this equipment setup, or can someone give
a better value for the tape capacity reduction factor?

	Regards,
	-- H}kan Torbj{r


(*) When it's worse, the dump stops with an I/O error trying to write
an EOF mark and then the Exabyte doesn't respond to neither `mt'
("device offline") nor the button.  When this has happened, the dump
has been previously terminated for other reasons (once by end-of-tape,
once the xterms died 2.5 hours into the dump; butc was restarted in
both cases and the tape rewound).  The dump was restarted and seemed
to proceed correctly for roughly 45 minutes before the write-EOF error
occurred and the drive got inaccessible.  I think this points towards a
tapedrive problem.  The system has logged two messages around the time
of the problem:

    esp0: Disconnected command timeout for Target 5 Lun 0
    st1:     SCSI transport failed: reason 'timeout': giving up

According to the SA that takes regular /etc/dump backup, this has
never happened to him.  Could this be some strange interaction between
`butc' and the device SW/HW?  And as they say, has this happened to
anyone else?

--
    H}kan Torbj{r           : Dept. of Computer Engineering     : qui me amat,
    torbjar@utc.chalmers.se : Chalmers Univ. of Technology      : amat et
    (+46) 031-72 16 66      : 412 96  G\TEBORG, Sweden.         : canem meam

home help back first fref pref prev next nref lref last post