SCSI error

[gerjam]

Hi HP Gurus,

I encountered some problems yesterday in one of my servers. My application was very slow so I rebooted the server using command "shutdown -ry 0". The server didn't go up and it hanged. I performed power cycle and it started ok. I checked on the syslog and found the ff:

Oct 13 16:43:12 above message repeats 19 times
Oct 13 16:43:02 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:43:12 above message repeats 19 times
Oct 13 16:43:12 vmunix:
Oct 13 16:43:12 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:43:23 vmunix:
Oct 13 16:43:23 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:43:33 vmunix:
Oct 13 16:43:33 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:43:43 vmunix:
Oct 13 16:43:43 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:43:54 vmunix:
Oct 13 16:43:54 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:44:04 vmunix:
Oct 13 16:44:04 vmunix: SCSI: Read error -- dev: b 31 0x005000, errno: 126, resid: 2048,
Oct 13 16:50:57 vmunix:

I checked for hardware failures but didn't see any. No amber light on the disks. Any idea about my problem will be very much appreaciated. Btw, we are using K-class platform. Thanks in advance.

[John Poff]

Hi,

It looks like your error messages are complaining about the device c0t5d0. Your power cycle might have fixed it for now, but I would be suspicious of that device. Usually when these kind of errors pop up and then disappear, they reappear on me at 2:00am on a weekend. :)

Here is a link to a post that discusses how to decipher the device field in the error messages:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=180363

JP

[Sridhar Bhaskarla · Accepted solution]

Hi,

Looks like there are intermittent failures on the disk c0t5d0. A quick way to find the device is to do ll /dev/dsk |grep 05000

Since these are only Read errors, there may not be any data corruption. However, if you see Write Errors, then you would need to unmount all the filesystems (pvdisplay /dev/dsk/c0t5d0) and do an 'fsck' on them.

If they keep on occuring (/var/adm/syslog/syslog.log), then I suggest you take a backup of these filesystems and replace the disk and restore the data.

-Sri

[Alexander M. Ermes]

Hi there.
I would get a new disk and mirror the old one to it. If the original disk fails, break the mirror and your system should still be ok.
Rgds
Alexander M. Ermes

[HGN]

Hi

The error relates to the disk currently that is read error & will not be a problem like the data getting corruputed but it is better to get the disk replaced ,the replacement can be done by different ways mentioned by others in this thread.

Rgds

Gopi

[Prashant Zanwar_4]

ls -altrR /dev | grep -i "0x05000" | grep -i ^b

this will be giving you the device the syslog is reporting on.

More ways to analyse on this device would be:

pvdisplay -v /dev/dsk/c5t0d0
diskinfo /dev/rdsk/c5t0d0

echo "selclass type disk; info; wait; infolog" | cstm | more

And also if you can have outage, just try unmounting the filesystem on the disk and do fsck, and mount them back. You will have a confidence taking steps..

Thanks
Prashant