Disk diagnostics and problem solving

[Doug_3]

Our db back up utility is now aborting. The db vendor is claiming we have a disk error or problem.
I see nothing in syslog/dmesg. Also, when I run STM I see only 'non-medium errors' on three of the disks. I was unable to see what this meant in the STM help files.

The question is what is the best way to determine the disk error, type, etc. The db vendor wanted me to run dd, but I have never done that as a diagnostic tool. The disks were created as raw. Ideas on running either fsck or stm's built in diagnostic tools?

Thanks in advance,
Doug

[Patrick Wallek · Accepted solution]

You can use dd by doing something like:

# dd if=/dev/rdsk/c?t?d? of=/dev/null bs=2048k

If dd terminates abnormally then you have a disk problem.

DO NOT use fsck if these are raw volumes. fsck is designed for filesystem use only and may screw them up.

[A. Clay Stephenson]

fsck will be absolutely useless because there is no underlying filesystem. STM can be more useful but be carefule that you do not do destructive exercises that write to the disk.

DD is actually a good tool for this.

There are two levels of I/O that need to be examined. If you are using LVM then it is possible that a single LVOL might span multiple disks.

dd if=/dev/vg02/rlvol1 bs=256k of=/dev/null

OR (to access a disk directly)
dd if=/dev/rdsk/c2t5d0 bs=256k of=/dev/null

I would choose the one that matches your backup scheme.


Man dd for details.

[Caesar_3]

Hello!

Check the disk with
dd if=/dev/rdsk/cXtXdX of=/dev/null bs=1024k
see if all the block readed.

Also check the disk with diskinfo if you
will get the info about the disk.

This way you will see if any hardware problems
You can use the sar to check the I/O of disk.

Caesar

[Doug_3]

Thanks all.