/stand online recreation

[raju_5]

My server,running on hpux 10.20 is having two disks in vg00 (root & root mirror). Recently I have noticed that the stand partition is not synhronised. When i tried to sync it, if failed saying IO error. I did an lvdisplay on to that lvol and found only the first extend is in the stale state(in the root mirror) and all other extends are syncronised. Thinking root mirror disk is having defects, I have replaced the disk and synchrinised back. But unfortunately I have the same problem again. Then I removed the morror on that particular lvol and did "dd if=/dev/vg00/lvol1 of=/dev/null bs=2048k" which terminated due to IO error(The actual problem was with the root disk).Now I can't afford to have one more downtime. Is there anyway that I can try distroying this volume(stand) and recreating it...??.
TIA

[Dietmar Konermann]

So everything except lvol1 is currently mirrored with all extents "current"?

First, take a backup of your /stand. Check if all files are readable, or if there are IO errors. You would need to get them from backup or re-create them later using mk_kernel.

The replace your root disk (which assumed to be the defective one). Synchronize the mirrors...

Now all lvols except lvol1 should be clean.

Then perform:

- umount /stand
(should work, no vital processes keep it busy)
- lvdisplay /dev/vg00/lvol1
(note the size of it)
- lvremove /dev/vg00/lvol1
- lvcreate -C y -r n -N -n lvol1 vg00
- lvextend -m 1 /dev/vg00/lvol1
- check with lvdisplay -v that lvol1 resides on the 1st extents of each disk only
- newfs -F hfs /dev/vg00/rlvol1
- mount /dev/vg00/lvol1 /stand
- restore your /stand backup
- check/fix lvlnboot setup

Regards...
Dietmar.

[raju_5]

Hi Dietmar

Thanks for the mail.

I have done so. But still I have the same problem. It seems that the root disk needs replacement.

[Dietmar Konermann · Accepted solution]

I assumed you would replace that...

Did you already some tests with it? E.g. reading from rdsk using dd(1)?

There may be some entries in the bad block directory of it.

How to check:

# dd if=/dev/rdsk/c0t5d0 bs=1k count=55 skip=9 | xd -t a
0000000 D E F E C T 0 1 nul nul nul nul nul nul nul nul
0000010 nul nul nul nul nul nul nul nul nul nul nul nul nul nul nul nul
*
55+0 records in
55+0 records out
000dc00

# dd if=/dev/rdsk/c0t5d0 bs=1k count=55 skip=73 | xd -t a
0000000 D E F E C T 0 1 nul nul nul nul nul nul nul nul
0000010 nul nul nul nul nul nul nul nul nul nul nul nul nul nul nul nul
*
55+0 records in
55+0 records out

This is an example of an empty BBDIR (both copies of it).

If you find entries there then you should replace the disk. Clearing the BBDIR or re-adding the PV to the VG (vgreduce/pvcreate/vgextend) could be other options... but you may only hide the root cause of your problem.

Regards...
Dietmar.