topic SCSI Resets in Operating System - HP-UX

SCSI Resets

Angela Swyers_1 — Wed, 18 Aug 2004 15:45:18 GMT

I am having a lot of SCSI resets appear in my syslog file:

Aug 18 12:01:36 acorn vmunix: SCSI: Resetting SCSI -- lbolt: 826160, bus: 4
Aug 18 12:01:36 acorn vmunix: SCSI: Reset detected -- lbolt: 826160, bus: 4
Aug 18 12:05:47 acorn vmunix: SCSI: Resetting SCSI -- lbolt: 851260, bus: 4
Aug 18 12:05:47 acorn vmunix: SCSI: Reset detected -- lbolt: 851260, bus: 4
Aug 18 12:08:21 acorn vmunix: SCSI: Resetting SCSI -- lbolt: 866660, bus: 4
Aug 18 12:08:21 acorn vmunix: SCSI: Reset detected -- lbolt: 866660, bus: 4
Aug 18 12:08:34 acorn vmunix: SCSI: Resetting SCSI -- lbolt: 867960, bus: 4
Aug 18 12:08:34 acorn vmunix: SCSI: Reset detected -- lbolt: 867960, bus: 4
Aug 18 12:08:49 acorn vmunix: SCSI: Resetting SCSI -- lbolt: 869460, bus: 4
Aug 18 12:08:49 acorn vmunix: SCSI: Reset detected -- lbolt: 869460, bus: 4

These tend to happen during any type of disk to disk backup. It appears, and this may be a total coinscience that my available memory drops quite a bit when this happens and doesn't seem to get cleared up. It eventually gets to a point where I must reboot my server.

Anyone have any suggestions? My patches are al up to date.

Re: SCSI Resets

Michael Tully — Wed, 18 Aug 2004 19:01:39 GMT

lbolt messages tend to mean there is some form of timeout.
You may want to look at the pvtimeout of your LUN's. If it is set to the default you may wish to change and see the result. This cahnge can be done on the fly.

pvdisplay -v /dev/dsk/cxtydz

pvchange -t 180 /dev/dsk/cxtydz

You might also look at your buffer cache as well. The general consensus is usually in the 300-500Mb range. (dbc_max_pct)

Re: SCSI Resets

Jeff Schussele — Wed, 18 Aug 2004 20:13:41 GMT

Hi,

Besides what Tully states, another leading cause is a device on the bus in distress.
Whenever *any* device on the bus resets due to a HW fault - the entire bus resets. So it's possible you're getting indications of an impending HW failure. Could be a drive or it could be the HBA itself. Unfortunately the msg is not indicating the device - the numeric value is simply a date value - specifically seconds from 1/1/70. You should scan the syslog.log for any other SCSI msgs in search of the device.

Rgds,
Jeff