First, thanks to everyone who replied. Here's the original problem:
We're looking at upgrading our 3/50 to an ELC. The 3/50 currently
has a shoebox disk and tape drive hanging off it plus another disk.
I'm told that this isn't "officially" supported on the ELC but that it
can be hooked up using a SCSI2 -> SCSI3 cable and in at least once case
seemed to work OK (Thanks Robert!).
Has anyone had any experience doing this? Any problems, comments, warnings,
suggestions would be most welcome. Please e-mail and I will summarise.
*** And yes, that should read "SCSI-1 -> SCSI-2" cable. ***
The general consensus was that it's fine and quite a few people were
running in that setup without any problems at all. There were some problems
though (I've included those messages at the end.)
Here's some replies.
Simon Hackett writes:
>It's a SCSI-1 to SCSI-2 cable that you meant to say, not SCSI2->SCSI3.
>But otherwise, yes it works just fine. I did _exactly_ what you want to do
>about two weeks ago. I got a sun SCSI-1 to SCSI-2 cable via Sun Direct
>last week, and plugged my old shoebox (disk+60 mb tape) back onto the
>end of the SCSI chain after my ELC's new 200 MB disk (which runs three
>times faster than the old Adaptec+Micropolis disk combination in the shoebox)
>No problems, except for the kernel bitching that my old disk is a "non-CCS"
>disk, which I suspect means that it's noticing that the old disk doesn't detach
>from the bus after a command is issued to free it up for other operations. What
>this means in practice is less throughput on your SCSI bus when the old disk is
>active. If you don't hammer that disk normally, it doesn't seem to make much of
>a difference. If you hammer the old disk, and then try to go for something on
>the new disk at the same time, you can get your machine slowing down
>dramatically in bursts while the SCSI bus gets locked out by the old disk.
>But you really only see this when you're doing massive file operations (like
>recursively destroying a directory tree) on the old disk.
>Note that I personally found it worth reading the man page on "tunefs" and
>doing what it suggested with the "-a" paramete (run the test script
>it suggests to find a more optimal contiguous block factor). The default on
>the old disks seems to be 1, and the "new disk default" (7) works about
>30% faster for me.
>I did a very similar upgrade (3/50 --> SLC, shoebox with disk and tape
>plus Wren V in a separate box) several months ago and have had NO
>problems. Be aware the scsi device designations will be different:
> sd0 became sd3
> sd2 became sd1
>Current configuration on the SLC:
> 0. sd1 at esp0 slave 8
> sd1: <Micropolis 1355 cyl 1018 alt 2 hd 8 sec 34>
> 1. sd3 at esp0 slave 0
> sd3: <CDC Wren V 94181-702 cyl 2602 alt 2 hd 15 sec 30>
>As I recall, I did a backup/format/partition/newfs/restore on each disk but I
>don't know that this was necessary.
Marcel Bernards warns:
>Did not work in our case, at least with the smaller ones (70-141 Mb)
>The SCSI->ESDI converter is tooooo sloooooww to keep up with the
>ELC/SLC/SS1(+) SCSI interface.
>Lots off timeouts and resetting SCSI bus barfs and slow access.
>better is to buy a new SCSI disk (Wren VI 660 Mb ) and dump the
>ESDI Disk + SCSI->ESDI converter. Disk's are rather cheap these days.
>The Sun shoeboxes are fine big boxes to build in your own stuff :-)
>I don't know if Sun will support this :-) :-)
Simon Hackett in a another message:
>yep. the old sun drives do work (I think there would have been a lynching
>if they didn't). I have one other (very old) third party drive that didn't
>work due to apparent SCSI incompatibilities, I'll farm that out to
>my other Sun system ( a sun 2/120 !), that isn't so fussy. Reasonably
>modern SCSI drives (e.g. Wren's or Quantums or similar) tend to work on
Thanks to everyone who replied! I'm very grateful for your help.
Yuval Tamir <tamir@CS.UCLA.EDU>
bernards@ECN.NL (Marcel Bernards)
Bertil Roslund <email@example.com>
Simon Hackett <firstname.lastname@example.org>
Jon Mellott <email@example.com>
Mike Raffety <firstname.lastname@example.org>
email@example.com (Skip Gilbrech)
firstname.lastname@example.org (Steve Jay)
email@example.com (Jennine Townsend)
firstname.lastname@example.org (Osicki K.)
email@example.com (Robert Dal Santo)
-- Eric Halil |Internet/CSnet: firstname.lastname@example.org Prentice Centre |Bitnet: email@example.com University of Queensland 4072 |UUCP: uunet!munnari!cc.uq.oz!eric AUSTRALIA -- Phone: +61 7 365 3610 |JANET: firstname.lastname@example.org
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:22 CDT