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

Disk Failure

Tiago Marques_2 — Wed, 07 Aug 2002 18:54:20 GMT

Hi!

I've executed the pvdisplay command at my disk c1t0d0 and got the message:

pvdisplay: Warning: couldn't query physical volume "/dev/dsk/c1t0d0":
The specified path does not correspond to physical volume attached to
this volume group
pvdisplay: Warning: couldn't query all of the physical volumes.
pvdisplay: Couldn't retrieve the names of the physical volumes
belonging to volume group "/dev/vg00".
pvdisplay: Cannot display physical volume "/dev/dsk/c1t0d0".

Is there a physical failure with this disk??

Is there another test that I can do to test the disk??

Thanks.

Re: Disk Failure

Sajid_1 — Wed, 07 Aug 2002 18:58:04 GMT

Is this c1t0d0 was part of vg00 or are you trying to add a new one ?

Is the disk comes in:
# ioscan -fnC disk ?

Is the disk shows somethin here:
# diskinfo /dev/rdsk/c1t0d0

Did you check the disk with stm or dd:
# stm
# dd if=/dev/rdsk/c1t0d0 of=/dev/null bs=64kb

Re: Disk Failure

Jeff Schussele — Wed, 07 Aug 2002 18:58:17 GMT

Hi Tiago,

Try this command

#diskinfo /dev/rdsk/c1t0d0

Note that you must use the raw disk "rdsk"

If this returns nothing OR O size OR O bytes/sector, then there is a definite problem with the disk.

HTH,
Jeff

Re: Disk Failure

A. Clay Stephenson — Wed, 07 Aug 2002 19:00:28 GMT

Almost certainly. It appears that vg00 is mirrored since you are still up although vg00 could consist of multiple unmirroed disks as well.

You could have inadvertently overwritten the LVM data structures on the disk. A good safe test would be to do this:

dd if=/dev/rdsk/c1t0d0 bs=256k of=/dev/null.

If that command completes sucessfully then the disk is ok and a vgcfgrestore, vgchange -a y, and a vgsync should get you back.

I'm betting however on a failed disk drive.

Re: Disk Failure

S.K. Chan — Wed, 07 Aug 2002 19:04:19 GMT

You can also run cstm's exerciser on it. Do this last after you've done what have been suggested by others.
# cstm
cstm> map
cstm> sel dev
cstm> exercise
cstm> exeractlog
cstm> exerfaillog

Re: Disk Failure

Tiago Marques_2 — Wed, 07 Aug 2002 19:06:26 GMT

This disk's part of VG00, actually it's mirrored with another disk. This disk already exist in the machine, I'm just checking his status after a building move of the company.

I tried the command diskinfo /dev/rdsk/c1t0d0 and the response is:

SCSI describe of /dev/rdsk/c1t0d0:
vendor: IBM
product id: DMVS
type: direct access
size: 0 Kbytes
bytes per sector: 0

The STM recognize the disk but doesn't generate any information about that...

Re: Disk Failure

Sajid_1 — Wed, 07 Aug 2002 19:10:35 GMT

hi,

It look like a failed disk, if the device name you are giving is right. Replace the disk and reconfigure it (vgcfgrestore).

gl,

Re: Disk Failure

Jeff Schussele — Wed, 07 Aug 2002 19:26:03 GMT

Hi Tiago,

Before you replace the disk, I would suggest that if it's in a hot-swappable enclosure, you should reseat it. Just pull it out & plug it back in securely. Wouldn't be the 1st time a move has unseated a drive.

Rgds,
Jeff

Re: Disk Failure

Sanjay_6 — Wed, 07 Aug 2002 19:38:18 GMT

Hi,

Looks like a failed disk to me. The disk is at c1t0d0. Try Diskinfo,

diskinfo -v /dev/rdsk/c1t0d0

If this works, then try to reactivate the root VG to which this seems to belong,

vgchange -a y /dev/vg00

then again do a pvdisplay,

pvidsplay -v /dev/dsk/c1t0d0

or go a vgdisplay,

vgdisplay -v /dev/vg00

See if the disk is shown as available in vgdisplay.

If none of these generate the desired output, replace the disk.

Hope this helps.

regds

Re: Disk Failure

Vincent Farrugia — Thu, 08 Aug 2002 05:40:56 GMT

Hello,

You say you started your server again after a building move.

I assume this server was up and running for a very long time before you decided to switch it off for the move to take place.

Harddisks sometimes fail to spin after being on for a very long time. I think that's what happened to you.

Try to remove the disk, shake it for a few seconds, then back on. Try to listen to it spinning. If it does not spin, you have to replace it. You have mirroring so no big deal.

HTH,
Vince

Re: Disk Failure

Tiago Marques_2 — Thu, 08 Aug 2002 10:36:37 GMT

I can't shutdown the machine at this moment, is there any problem if I remove the disk with the machine running? And then I will put that again to executed the tests again.

Re: Disk Failure

Vincent Farrugia — Thu, 08 Aug 2002 11:57:00 GMT

Hello,

The hot-swap action of harddisks is supported only on a handful of servers, and, before removing, you have to prepare the system LVM-wise. It's not like a RAID disk array where you can remove disks without anything.

Please tell us what is your server.

Vince

Re: Disk Failure

Tiago Marques_2 — Thu, 08 Aug 2002 12:16:51 GMT

I didn't understand what you mean by a "handful of servers" but what I understand about your explanation, it's better I shutdown the server and proceed with the "shake"of the disk, and then put that again and restart the machine.

Ok?

Re: Disk Failure

Sajid_1 — Thu, 08 Aug 2002 13:09:51 GMT

Hi,

If the disk is in a HASS box or Disk array which is hot swappable, then you can remove and install it back (or install new one) when the system is online. This will not create any problem if you are removing the right disk.

But if the disk is not hot-swappable, then you should shutdown the server to test your ideas.

Re: Disk Failure

Vincent Farrugia — Fri, 09 Aug 2002 05:00:13 GMT

Hello,

You can shutdown your server and do that yes, but if you have one of the newer servers, like the rp54xx or rp74xx, this is not needed. You can actually remove the disk online and replace it online. Before that, though, you have to prepare your system in such a way as not to recognise the disk. In your case, remove mirroring and remove the disk c1t0d0 from lvmtab.

HTH,
Vince