[2895] in SIPB-AFS-requests
Sending failed 1GB disk back to ADS
daemon@ATHENA.MIT.EDU (John Hawkinson)
Mon Nov 10 12:49:36 1997
Date: Mon, 10 Nov 1997 12:49:26 -0500
To: star-maintainers@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
Reply-To: jhawk@MIT.EDU, usenet@MIT.EDU
There was a 1GB ADS disk sitting in the office that Greg
seemed to think had come off of beacon, and was planned to go
on Charon.
I decided to co-opt it for an experiment with Solaris on beacon
(to see if it made better use of the multiprocessing capabilites).
While testing it on bobbi, I found it had a couple hundred bad blocks
that produced write errors, and could not be repaired (though the
defect map didn't seem full...). It also refused to format.
Greg then recalled that it had been taken out of service
due to disk errors. I'm not sure what service it was operating
from, as bobbi was unable to read a disk label, implying it wasn't
SunOS or Solaris (so presumably Ultrix and therefore AFS?).
Apparently this disk had been sent back to ADS previously (it appears
that we neglected to purchase advance replacement for it...), but they
couldn't find anything wrong with it. It's my assertion that if things
like this happen:
1) They replace it anyway (even if they cannot reproduce it),
because they like us as a customer.
2) We go to a vendor who can replace disks when we see failures,
regardless of whether they like us as a customer.
And we currently seem to be in neither of those states.
It's also disturbing to me that the disk wasn't documented in
/afs/sipb/admin/text/labels/disk-label.ps, and I had difficulty
finding information about it's failure in discuss meetings, or in
SIPB meeting minutes.
Info for the disk is:
Serial No.: 36897-6020
Model No.: ADCFP1080-STL (Conner 1080-S)
MIT Tag No.: *NONE*
Purchase date: 10 Jan 1996
I'm not sure why there is no MIT tag number.
The following entries in disk-label.ps lack tags:
/disk (DEC RZ55) def
/server (on Charon) def
/tag (MIT Tag Number - *NONE*) def
/disk (Falcon) def
/server (on Bloom-Beacon) def
/tag (MIT Tag Number - *NONE*) def
/disk (Falcon) def
/server (on Bloom-Beacon) def
/tag (MIT Tag Number - *NONE*) def
/disk (Falcon) def
/server (on Bloom-Beacon) def
/tag (MIT Tag Number - *NONE*) def
/disk (Falcon) def
/server (on Bloom-Beacon) def
/tag (MIT Tag Number - *NONE*) def
(Presumably we don't care about the RZ55). I'll verify
the taggedness tonight, unless someone does so first.
I'm not sure why they wouldn't be tagged, I guess someone
should follow up with the property office if it's accurate
(procedure?).
I opened RMA 13183 with ADS for this disk, if someone is heading
over to E40 from the SIPB office today I'd appreciate it if they
could deal w/ shipping (zwrite me for details). Expected turnaround
time is potentially long, as invariably it will require them shipping
it to Seagate/Conner, and we didn't purchase advance replacement [*].
Thanks.
--jhawk
[*] Some people have suggested we should stop purchasing advance
replacement and instead either have spare disks on hand or be prepared
to buy a new disk the moment we see a failure. I think this is a good
idea.
