topic Re: Bad Disk in Operating System - HP-UX

Bad Disk

Bill Pontius — Wed, 31 Oct 2001 16:09:57 GMT

I've run the command below and have determined that I have a block block on the /dpss fs which span the h/w paths of 52.3.0, 52.4.0 and 8.5.0. How do I find which is the bad disk. Also for my fyi, after the disk is replace, I should add the disk to the vg01, check to see that /dpss filesystem is assigned to correct logical volume and then restore all of /vg01.
/dpss/hp$fsck /dev/vg01/lvol21
** /dev/vg01/lvol21
** Last Mounted on /dpss
** Phase 1 - Check Blocks and Sizes

CANNOT READ: BLK 670272
CONTINUE? y

CANNOT READ: BLK 670272
CONTINUE? y

CANNOT READ: BLK 670272
CONTINUE? y

FAILED READ OF BLOCK #670272, RETRIED 2 TIMES
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
SUMMARY INFORMATION (INODE FREE) BAD
BAD CYLINDER GROUPS
FIX? y

** Phase 6 - Salvage Cylinder Groups
90 files, 0 icont, 1135636 used, 2255489 free (121 frags, 281921 blocks)
DISK MEDIA PROBLEMS ENCOUNTERED!
BAD BLOCKS WERE FOUND ON THE DISK.
***** FILE SYSTEM IS NOT CLEAN -- DISK MEDIA PROBLEMS ENCOUNTERED *****

***** FILE SYSTEM WAS MODIFIED *****
/home/root$dd if=/dev/vg01/lvol21 of=/dev/null bs=1024k
dd read error: I/O error
654+1 records in
654+1 records out

Re: Bad Disk

Roger Baptiste — Wed, 31 Oct 2001 16:18:59 GMT

hi,

do an lvdisplay -v /dev/vg01/lvol21 |more

and look for stale extents.
The disk corresponding to
the stale extent should be
the faulty one.

Once you identify it, do
a pvdisplay on it to make
sure it is indeed not accessible. Also check it
doing an ioscan -fnCdisk
and see whether it shows
a NO_HW corresponding to it's
path

-raj

Re: Bad Disk

Sanjay_6 — Wed, 31 Oct 2001 16:21:34 GMT

Hi Bill,

Have you tried "lvdisplay -v "

hope this helps.

thanks

Re: Bad Disk

harry d brown jr — Wed, 31 Oct 2001 16:21:35 GMT

Try using stm and exercise the three disks, then view the logs.

live free or die
harry

Re: Bad Disk

Uday_S_Ankolekar — Wed, 31 Oct 2001 16:22:54 GMT

Hi,

Two way you can check about the bad disk.
do a ioscan on the disks
ioscan -fnC disk

and run a lvdisply -v /dev/vgnn/lvolnn and look for any stale

GoodLuck,
-USA..

Re: Bad Disk

Bill Pontius — Wed, 31 Oct 2001 20:37:33 GMT

When I did the lvdisplay ..., all was current. The pvdisplay found the 3 disk for /dpss but all had stale pe 0 - two of the disk showed all current and the other one was current or free. ioscan showed all disks without a NO_HW. stm just sat there and the only thing I could do was cnt k, and then got trapped in the help menu.

Re: Bad Disk

Frank Slootweg — Thu, 01 Nov 2001 12:49:19 GMT

An "lvdisplay -v" will display the device names (/dev/dsk/...) for the PVs used for this LV.

Then, for each device file, do a dd(1) read test:

dd if=/dev/rdsk/... of=/dev/null bs=64k

(note rdsk, not dsk)

Probably one of these will give an I/O error. Then do a "lssf /dev/rdsk/..." to determine the hardware path, do an "ioscan -f -H ...", etc..

Re: Bad Disk

Bill Pontius — Thu, 01 Nov 2001 13:46:37 GMT

Frank,
The dd using rdsk did the trick. I now have the hardware path needed. Thx.