SUN Managers:
This is a summary of the replies I received regarding "incomplete write-
retrying" error messages from one our SCSI-1 Drives.
My original Posting was as follows:
********************************************************************************
> On our SPARC/2 running SunOS 4.1.3 (no patches) one of our external SCSI
> drives reports system error messages which look as follows:
>
> ================================================================================
> Jun 5 16:50:18 felix vmunix: sd1: incomplete write- retrying
> Jun 5 16:50:20 felix last message repeated 4 times
> Jun 5 16:50:56 felix vmunix: esp0: Disconnected command timeout for Target 1 Lun 0
> Jun 5 16:50:56 felix vmunix: sd1: SCSI transport failed: reason 'timeout': retrying command
> ================================================================================
>
> The SCSI chain is close to the max length of 20' and additionally there is
> a mix of SCSI-I and SCSI-II devices on the same adapter. SCSIINFO reports
> the following:
>
> ================================================================================
> esp0: st1 tgt 5 lun 0:
> Asynchronous Clean CanReconnect
> <<< SCSI probe failed: device not accessible (media not loaded?) >>>
> esp0: sd2 tgt 2 lun 0:
> Synchronous(4.0MB/sec) Clean CanReconnect
> Non-removable Disk: CDC 94181-15 0293
> esp0: sd1 tgt 1 lun 0:
> Synchronous(4.0MB/sec) Clean CanReconnect
> Non-removable Disk: CDC 94181-15 5466
> esp0: sd0 tgt 3 lun 0:
> Synchronous(4.0MB/sec) Clean CanReconnect
> Non-removable Disk: SEAGATE ST1480 SUN0424 7516
> ================================================================================
>
> Anybody able to suggest the most likely source(s) of the problem?
********************************************************************************
Thanks very much to those who replied:
Bismark Espinoza <bismark@alta.jpl.nasa.gov>
Ashley Gilbert <ashley@india.ti.com>
Louis M. Brune <brunel@delver.iterus.org>
Carlo Musante <carlo@hub.eng.wayne.edu>
Ray Brownrigg <Ray.Brownrigg@isor.vuw.ac.nz>
Lee Ann Goldstein <lgoldste@ladc.lockheed.com>
Kevin Sheehan <Kevin.Sheehan@uniq.com.au>
Jeremy Hunt <jeremyh@chiron.nabaus.com.au>
John Elliot <johne@aiaitest.ed.ac.uk>
Paul Woods <woodsp@smtplink.Indigo.co.il>
Suggestion provided were:
1 - Shortening the chain (10' "better" than 20', account for internal cabling)
2 - Putting SCSI-1 & SCSI-2 devices on different chains
3 - ROM Version > 1.3 for better support when mixing SCSI-1 & SCSI-2
4 - Changing from Passive to Active Termination (Forced Perfect Terminator)
5 - Moving all SCSI-2 devices to higher IDs than the SCSI-1 devices
6 - Balancing disk usage between chains (if more than a single chain)
7 - Using special "twisted pair" cabling
8 - Convert to Differential cabling
The problem appears to have cleared up by shortening the total length of
cabling! Still NOT down to 10' but now well under 20'.
NOTICE: during the operation of replacing an external cable one of the
external "enclosuers" died and necessitated replacement of both its
internal ribbon cable and its power supply! Hence the improvement may NOT
be entirely due to the shortening of the cabling, but this was the most
frequent suggestion?
Lawrence Houston - (houston@felix.geog.mcgill.ca)
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:28 CDT