Ultra160 SCSI bus suspend / timeout errors

[Philip Kernohan]

We have an RP7410 connected to a disk array (non-HP but apparently compatible) via 2 x Ultra160 SCSI paths (5 metre cables).

The controllers are A6828-60101's. We had experienced some timeout errors on both SCSIs before now and both cables were reseated (carefully) and we increased the PV timeout from 30 seconds to 120 seconds at that time.

The system was running well for about 2 weeks and then yesterday we received the attached errors in syslog where the /dev/dsk/c4* path to the devices failed and was autoswitched to /dev/dsk/c9*.

On reboot of the server the same SCSI bus experienced a hang and that path is now unavailable. (Also in attached file)

diskinfo returns data but all PV/VG commands state that the /dev/dsk/c4* devices cannot be queried.

The array itself does not give a lot of detailed log information so I can't find anything from that end.

The ends of both SCSI cables are physically sound and I have only a simple connectivity tester to hand and so far that looks good. I have only one good Ultra160 cable and that's currently serving as the working path to the array that is now critical to development.

Patches up to and including June 2004 are loaded.

Questions are ...
- can anyone determine if this an HBA problem or a cable problem by looking at the log file attached?

- I have patches loaded up to June 2004 but there have been a number of patches associated with Ultra160 issues and may need others. I will search the patch database but wonder if anyone else has this same issue.

- is the PV timeout of 120 seconds what causes the 'delay' between bus being suspended at 08:47:00 and the autoswitch at 08:48:59? Are the VG/LVs still accessible during this time?

I think acquiring a new cable and scheduling maintenance to swap cards and cables is what it will eventually take but with 12+ days between failures this one might be tricky.

All help greatly appreciated.

Regards,
PK