I/O error on backup

[Midrange Team]

We have a SAN config (FC-AL) consisting of several HP Hosts, 7700e disk and a scsi/mux (4 dlt drives). When we reboot a host on the loop, and a backup is running on another host id the loop, we recieve an I/O error on the backup. I assume this is due to a scsi reset upon lip of the rebooted host. Any ideas?
Cheers

[Patrick Wessel]

Peter,

There is the possibility that the loop initialization during the reboot is responsible for the IO error. Where do you get the IO error? And do you find any 'strange' message in the syslog-file of the box which reports the error (SCSI reset, link failure?.)?

[Midrange Team]

The I/O error occurs on the host that is being backed up. No messages in the syslog or dmesg to indicate a FC failure or scsi reset.

Cheers

[Patrick Wessel]

That's strange, I thought that the system should recognize the IO error in another way. Have you checked the status of the FC host bus adapter before and after an IO error? Just to get an idea what happened on the loop. fcmsutil /dev/fcms? stat prints you the statistic of the interface that belongs to the devicefile fcms? There are just a few values of interest.

Loss of signal Count
Bad Rx Char Count
Loss of Sync Count
Link Fail Count
Bad CRC Count
num_resets_initiated
num_resets_completed

Can you see any significant changes of these parameters after a your system got the IO error?

[Pamm Martinez]

Your problem is with the arbitrated loop. Any reboot is sensed in all systems attached to the switch. This causes backup to abort. This is a known issue. Turn off arbitrated loop in the switch. The servers need to be able to handle this. Try L and N class servers. K's can't talk non-arbitrated loop.