SCSI Async Write Error

[CCIL]

I am getting the following message in dmesg output in a Virtual partition of HP-9000 N class server

SCSI: Async write error -- dev: b 31 0x040300, errno: 126, resid: 2048,
blkno: 5260124, sectno: 10520248, offset: 1091399680, bcount: 2048

This VPAR boots from external storage device VA7100
Pls suggest solution for the same.

[Armin Feller]

If there is surely no HW problem, then check if there is set lvtimeout (lvdisplay). If you are working with filesystem on this lvol there should NEVER be set a timeout !!!

Regards,
Armin

[Dietmar Konermann]

Hi!

The scsi driver clearly indicates that some I/O requests timed out (did not complete withing configured pvtimeout, 30 secs by default). The errno 126 is EPOWERF/ETIMEDOUT.

Typically this is caused by hardware problems.

Best regards...
Dietmar.

[CCIL]

no lvtime out is set.
the device is /dev/dsk/c4t0d3 & it is vg00 for this vpar

Thanks & Rgds
Sarvesh

[Eugeny Brychkov]

Device is at address c4t0d3. It looks like VA7100 LUN.
First of all, check with increasing this PV timeout: 'pvchange -t 180'.
Could you please attach to your next reply (in one attachment) 'ioscan -fn' and 'armdsp -a' outputs? I would like to check your VA7100 storage subsystem
Eugeny

[Armin Feller]

Please check the disk, it seams to be a hardware problem.

Regards,
Armin

[Steven E. Protter]

Typically I get this kind of messaging with an lbolt erorr.

lbolt means the drive is dead, even if it is in fact still walking, forgive the metaphor.

You can increase the timeout on your drive as specified above, but powerfail is powerfail, and you may get an lbolt soon.

I have a utility that checks for lbolts and sends an email out at 4 a.m. That is attached, you may want to start running this so you can get your data off before the drive dies.

Other possible causes:

Problem with pins or SCSI card.

Problem with a drive cage(these devices provide power and data transfer, typically to a number(5) of internal drives).

Another dead drive on the SCSI chain.

A little story.

On a D320 server with 20 conventional scsi drives and 5 vdisks accessed via a fibre card, one of the old scsi disks essentially melted down. When hardware removed it, it was too hot to touch.

When the drive died however, we lost access to the vdisks hung off the fibre card and every scsi device hung down the chain.

Interesting.

Anyone ever have a disk drive catch fire? That's how hot it felt.

Steve