SUMMARY: Shared SCSI w/o ASE

From: Alex Gorbachev <gorbachev_alex_at_bah.com>
Date: Tue, 03 Mar 1998 13:35:30 -0500

This time I have received a large number of replies to the original
question of whether it is possible to run a shared bus without ASE
subset installed. Based on these replies and the info from DEC, the
answer is definitely NO.

ASE contains kernel modules that prevent or provide tolerance to SCSI
bus resets caused by I/O request contention between the systems. HSZ
controllers may restart themselves in response to these resets.
Problems ususally appear at arbitrary times, about once a month in a
production-load environment, and may cause filesystem corruption,
failure to rebuild RAID sets, failure for dual-redundant controllers to
fail over and other nasty consequences.

Now that I am preparing to install ASE, here are some questions for
which I could not receive a reply from DEC. Any help would be greatly
appreciated:

- How to disable automatic reboots and automatic service relocation by
ASE?

- How to perform manual service relocation non-interactively from a
command line (script)?

Many thanks so far to:
Dr Marco Luchini <m.luchini_at_ic.ac.uk>
George Guethlein <GGuethlein_at_GiantOfMaryland.com>
Kurt Carlson <snkac_at_java.sois.alaska.edu>
Jirik Stephen x8373 <Sjirik_at_CitiPower.com.au>
Peter.Braack_at_degussa.de
"Spalding, Steve" <SSPALDIN_at_mem-ins.com>
Walker Steve <SWALKER_at_dsl.uk.ibm.com>
"Fliguer, Miguel" <M_Fliguer_at_miniphone.com.ar>

Original Post:

> I have deployed a pair of 2 4100 servers with Digital UNIX connected
> over a shared SCSI bus to a dual-redundant pair of HSZ50s with 4
> RAIDsets. The storage volumes are mounted each on only one system at a
> time. However, the systems perform a large number of SCSI bus resets,
> which cause problems every once in a while (one KZPSA resets the bus and
> the reset propagated in "waves" of error messages to the other system
> and the LUNs).
>
> DEC strongly feels that this is configuration related.
>
> The question:
>
> Is this never going to work and ASE is the only choice or can the
> systems become more tolerant using a kernel fix, etc. to prevent SCSI
> resets?
>
> Any info would be greatly appreciated.
> Alex Gorbachev
> gorbachev_alex_at_bah.com
Received on Tue Mar 03 1998 - 19:28:29 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT