[2093] in SIPB-AFS-requests
rosebud2 cables replaced
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Aug 10 00:10:36 1995
Date: Thu, 10 Aug 1995 00:07:53 -0400
From: Greg Hudson <ghudson@MIT.EDU>
To: sipb-afsreq@MIT.EDU
I replaced all the SCSI cabling on rosebud2. (The disk-to-disk cables
came from jtkohl; the terminator and disk-to-Decstation cable came
from bwmelans. All three cables were in their packaging, although the
terminator might have been used before for all we know.)
The outage took about twenty minutes, since I ran into a problem with
a bent piece of metal on the Seagate 4GB enclosure and had to fight
with it to straighten it out far enough that I could plug in and clip
on a cable correctly.
vos creates on rosebud2 f still take about 25 seconds (as oppsed to
1-3 seconds for any other partition, including rosebud2 e). So the
Seagate slowness problem is not, apparently, a SCSI bus cabling
problem.
The next step is to move all the data off rosebud2 rz4 (the Seagate
4GB disk), find a way of reproducing the Seagate 4GB disk slowness
that's not related to AFS operations, do timing tests on rosebud2
relative to its internal disk, move the Seagate 4GB disk to another
Decstation, and see if it's slow there relative to the other
Decstation's internal disk. This will tell us whether it's a host
adaptor problem or a problem with two disks. In the former case,
we'll swap the CPU board; in the latter case, we'll get RMAs for both
disks.
We have about two weeks to get this done. Tom, can you please start
bulk moves off rosebud2 e and f as soon as is convenient?
(Incidentally, in a conversation with Tom, I indicated that the
slowness problems were intermittent. This was due to my not knowing
about the difference in speed between vicepe and vicepf for vos create
operations. As far as I can tell, dd read tests don't indicate any
slowness problems with any partitions, but the vos create slowness is
very consistent on vicepf.)
