how to analize a raw volume group/disk

[Michele (Mike) Alberton]

Hi !

We have a set of disk drives assigned to Sybase 11.x and defined as raw disks.
Sybase db check is showing an issue on one block of the drive.
Is there any way to analyze the status of the drive at a Unix level ?
I believe fsck does not allow that since the disk is not using hfs or vxfs.

Could you help me ?

Thanks in advance,

Mike

[A. Clay Stephenson · Accepted solution]

About the only thing you can do is to read every block in the raw disk like this:

dd if=/dev/rdsk/c5t2d0 (or maybe /dev/vg03/rlvol4 if using LVM) bs=256k of=/dev/null

This will read everything in your raw volume/disk in 256k chunks and output it to the null device. This is a read-only operation and is safe. Dd will output a status when finished. Man dd for details.

[Wodisch]

Hi Mike,

if Sybase server had a hardware problem with the drive, you'll find en entry (many lineS) in your local "syslog" (/var/adm/syslog/syslog.log).
That will tell your controll, drive, and block, IIRC :-)

The "check" with "dd" will enter an additional entry...

HTH,
Wodisch

[Michele (Mike) Alberton]

Thanks Folks !

Actually sybase tells me a lot about where to look for (virtual disk No. and block No. which I can map to drive/lvol).
I got these info both from a dbcc checkdb and from Sybase errolog when trying to access that part of the drive

I'll take a look with dd and wait for the status.

Do you know why lvdisplay -v on that lvol does not show any stale partition ? (everything looks available there).


Thank you very much for the hints !

Cheers,
Mike

[Wodisch]

Hi again,

"lvdisplay" will only tell you about an "LE" being *stale* if that LV is actually mirrored - if it is not, you won't see "stale"...

Sorry,
Wodisch

[Michele (Mike) Alberton]

Hi !

I used dd and actually I got back a read error after a number of records.
That's enough for me, just wanted to thank you for the quick response.

Ciao !


Mike