Scsi addressed changed

[John Liden]

Hi all
After a power failure last weekend, my SAN attached disk's on just one server decided that the old addresses of C9txdx and C7txdx would rather be C3txdx and C5...

No HW was added or deleted.
Why did they change?

[Steven E. Protter]

Some hardware may have been damaged.

insf -e

might put things right, but you'll have to rmsf the drives you don't like first.

Also this is a good time to make sure you have a good make_tape_recovery tape on hand.

SEP

[John Liden]

insf -e didn't do much. Changed the following in /dev
rw-rw-rw- 1 bin bin 31 0x000000 Nov 17 10:19 lan0
drwxr-xr-x 2 bin bin 57344 Nov 17 10:20 pts
crw-r----- 1 bin sys 255 0xffffff Nov 17 10:20 rroot
brw-r----- 1 bin sys 255 0xffffff Nov 17 10:20 root
crw-rw-rw- 1 bin bin 207 0x000000 Nov 17 15:27 tty

any other ideas?

[Steven E. Protter]

Its time to figure out what the power failure did to your system.

Your box going up and sown should not change the SCSI adderss of the devices. Even if this were daisy chained scsi devices the impact would not be this.

If you had two fiber cards and one of them blew out during the power fail, then obviously the scsi address would have to change. The scsi address is based on the bus position of the scsi or fiber card.

Is it possible that due to the failure there was a switch in cable position or a change on the fiber switch?

This has become an investigation. You'll probably figure out what happened by talking to people involved in the recovery.

SEP

[Suresh Patoria]

Hi,

You check in the SAN software which address u define in the software

i mean LUN address if it changed then put it back to the previous address

Thanx

[Sunil Sharma_1]

Hi,

There could be 2 reason for this.
1.it's possible some device got failed becoz of that your device address cahnged.
2.You might add some new device.

Sunil

[G. Vrijhoeven]

Hi,

If your switch is in fabric mode and ou change the port number on the switch or the domain number on the switch your instance number ( Ct?d? will change.


Gideon

[John Liden]

All good answers but none of these things happened. There is only one fiber card in use and no cabling changes or hardware changes occured. All of the other devices in the server are seen by ioscan, and nothing is unclaimed or missing. I am not sure if that eliminates the failed board issue or not.

[ketan_5]

Looking at history and detail you provided there are more chances that somebody tried to do R&D on your server. Anyway If you have latest recovery tape , will solve the problem.

[Jakes Louw]

Have you checked the port assignment on the SAN switch itself? If the switch config changed or rebuilt itself, then your CTD allocation from the switch would change.

[Francis_12]

Hello,

>Why did they change?

You did not mention anything about your attached disks (3rd party storage ?), your eventual switches (HP brocades ?) and the servers identities (model, pdc, etc...). So, it's pretty hard for us to diagnose it like that.

If you have no relevant informations in the syslog.log, you certainly might find helpfull informations in the switches logs (a supportshow for instance on brocades might reveal something).

Hope this helps, Bye.

Francis DERDEYN - HP-UX ASCE.

[John Liden]

The configuration is a N4000-55 with fabric to a Brocade 3800 to a Hitachi 9970.
I checked syslogs and brocade info and nothing is pointing to changes either in hardware, software, drivers, zoning or port reconfiguration or error on the components.
Would the power hit on the GSP cause anything like this?