sysruth@helios.toronto.edu (Ruth Milner) (08/01/89)
Here is the promised followup concerning major problems running CDC Sabre
disks on Iris 4D systems (at least).
I received only a few replies (which would seem to indicate that the number
of systems running the Xylogics 754/CDC Sabre combination is quite small).
Most of them were from others who have experienced the same problems we have.
One of them was from SGI, and has been very helpful.
The current method of flagging bad blocks by setting a bit in the block's
sector header is actually the Xylogics spec. SGI has a version of fx which
I believe uses this method where possible, but in the situation where the
defect is in the header itself, handles it differently - and successfully.
At the moment I have only a standalone version, but that will be enough
to do our reformatting.
I strongly recommend that those of you who are still having problems getting
your Sabres working contact your local office, explain the problem, and
get them to get hold of a copy of this for you. If your system is on software
support, it should make no difference whether you bought your disks from SGI
or not (we didn't, though the 754 is theirs). The problem is in the software,
not the hardware. If it isn't covered, well, you can try, but getting software
patches free from any vendor is generally dependent on buying software support.
You could probably arrange to buy it, I don't know for sure. Don't be
discouraged if they haven't heard of it; it will probably take some research
on their part. I do not know for sure whether this has a bug number assigned
to it.
As a side note, some of you may be interested in the resolution of the
problems with our Exabyte. We are hanging this directly off the SGI SCSI
interface, by means of a SCSI extender cable from SGI, daisy-chained off
the tower containing the 1/4" tape and small disk. We were told before we
purchased the system that this would work, but even at 3.1F it doesn't. It is
necessary to either apply a special options tape, or upgrade to 3.1G or 3.2.
These versions have proper support for the Exabyte.
Finally, some advice based on a lot of frustration :-) :
If any of you are still trying to run SMD disks at 3.1C, you should upgrade
to at least 3.1D. There was a lot of added support for SMD in this release,
as well as some NFS fixes. You should also make sure that your 754 is at
least rev. 2.4. Prior to this there were problems which cause the system to
hang under heavy disk loading. Upgrading will require reformatting any
existing disks, though it need not be done immediately as the new revs can
read the old format. This is why we have to go through it.
For those of you thinking of purchasing the 754/Sabre combination, you will
want to make sure you have all the right bits and pieces ahead of time. One
of the people I heard from has tried 5 Sabre 1230's and had this problem with
every single one of them. We have seen it on both of our two. So the chances
of it happening to others are probably high.
Hope this helps some people :-).
--
Ruth Milner UUCP - {uunet,pyramid}!utai!helios.physics!sysruth
Systems Manager BITNET - sysruth@utorphys
U. of Toronto INTERNET - sysruth@helios.physics.utoronto.ca
Physics/Astronomy/CITA Computing Consortium