VG problems following restore

[John Warminski]

After Restoring from Ignite tape, two volumne groups are missing, that part I would expect.
Issue is inconsistancy bewtween /dev, ioscan, and volume group

There is a new drive in /dev/dsk/c2t2d0 ...

The drive that appears in the ioscan and the old vg returns an error if I try to write to it.

Has anyone run across this ?

Any suggestion on an approach to this ???

Restoring the vg will incorporate basically a bad devise ( I think ). Also concerned about the new spurios device .

Following is the editted ioscan, vgcfgrestore, andls of /dev/dsk with the results of cat to
the original drive

ioscan -funC disk
Class I H/W Path Driver S/W State H/W Type Description
==============================================================
........

disk 12 16.9.0 disc3 CLAIMED DEVICE SEAGATE ST15150W
/dev/dsk/c2t9d0 /dev/rdsk/c2t9d0
/dev/floppy/c2t9d0 /dev/rfloppy/c2t9d0
disk 13 16.11.0 disc3 CLAIMED DEVICE SEAGATE ST15150W
/dev/dsk/c2t11d0 /dev/rdsk/c2t11d0
/dev/floppy/c2t11d0 /dev/rfloppy/c2t11d0
************ The Disk in question *******************
disk 14 16.12.0 disc3 CLAIMED DEVICE SEAGATE ST34573WC
/dev/dsk/c2t12d0 /dev/rdsk/c2t12d0
/dev/floppy/c2t12d0 /dev/rfloppy/c2t12d0
******************************************************************
disk 15 16.13.0 disc3 CLAIMED DEVICE SEAGATE ST34573WC
/dev/dsk/c2t13d0 /dev/rdsk/c2t13d0
/dev/floppy/c2t13d0 /dev/rfloppy/c2t13d0
disk 16 16.15.0 disc3 CLAIMED DEVICE SEAGATE ST15150W
/dev/dsk/c2t15d0 /dev/rdsk/c2t15d0
/dev/floppy/c2t15d0 /dev/rfloppy/c2t15d0
disk 10 52.2.0 disc3 CLAIMED DEVICE TOSHIBA CD-ROM XM-3301TA
/dev/dsk/c1t2d0 /dev/rdsk/c1t2d0
/dev/floppy/c1t2d0 /dev/rfloppy/c1t2d0
disk 11 52.6.0 disc3 CLAIMED DEVICE HP C3010M1
/dev/dsk/c1t6d0 /dev/rdsk/c1t6d0
/dev/floppy/c1t6d0 /dev/rfloppy/c1t6d0
Hewie >

Hewie >vgcfgrestore -n vg02 -l
Volume Group Configuration information in "/etc/lvmconf/vg02.conf"
VG Name /dev/vg02
---- Physical volumes : 5 ----
/dev/rdsk/c2t9d0 (Non-bootable)
/dev/rdsk/c2t11d0 (Non-bootable)
/dev/rdsk/c2t12d0 (Non-bootable) ***** drive in ioscan *****
/dev/rdsk/c2t13d0 (Non-bootable)
/dev/rdsk/c2t15d0 (Non-bootable)
Hewie >

ls -l c2*
brw-r----- 1 bin sys 28 0x02b000 Jan 20 12:25 c2t11d0
brw-r----- 1 bin sys 28 0x02c000 Jan 13 2003 c2t12d0
brw-r----- 1 bin sys 28 0x02d000 Jan 21 12:41 c2t13d0
brw-r----- 1 bin sys 28 0x02f000 Jan 20 12:25 c2t15d0
brw-r----- 1 bin sys 28 0x022000 Jan 20 12:14 c2t2d0
brw-r----- 1 bin sys 28 0x029000 Jan 20 12:25 c2t9d0
Hewie >cat /etc/hosts > c2t12d0
sh: c2t12d0: Cannot create the specified file.
Hewie >

[Helen French]

Is c2t12d0 actually a part of vg02? Did you add a new drive or change the LVM configuration after your last Ignite recovery tape creation? Also, if you can describe the problem little more, that would be helpful.

Another thing is 'cat /etc/hosts > c2t12d0' is not a right way to check the disk. Check the following commands:

# pvdisplay /dev/dsk/c2t12d0
# diskinfo /dev/rdsk/c2t12d0

If you find out which LV is holding on c2t12d0 (from pvdisplay), write to that file system:

# cat /etc/hosts > /file_system/testhosts

Also, use files in /tmp or other log files while testing things instead of configuration files from /etc. A reverse "redirect" can erase your hosts file.

[Sridhar Bhaskarla]

Hi,

Are you saying that this disk was not there before and is appearing now all of a sudden after the ignite restore?.

How about a 'diskinfo /dev/rdsk/c2t12d0'? Does it give any information?.

Looks like the vgcfgrestore is successful. Did you run 'vgchange -a y vg02' later after importing it to see if it is getting activated?.

-Sri

[Helen French]

To add, use 'dd' and 'stm' commands to check your hard drive:

# dd if=/dev/dsk/c1t12d0 of=/dev/null bs=256k

# stm (or cstm or mstm and select the device you want to test)

[Todd McDaniel_1]

John,

First you cannot merely cat file > c2t12d0... not sure what that is... you need to do a dd to the device to test it not a cat.

Now, I dont think I know enough yet...You say all disks are visible.. run a diskinfo and pvdisplay on each of them. If you get errors they are not part of a VG.


I first have a question why your VGs are missing? What caused them to be missing?

I would restore via ignite, you already did, and then vgexport vgX to destroy them and then restore using the mapfiles you have in /etc/lvmconf/*


What disks appear in your mapfiles for these VGs?

Did you have a hardware path change? This may be why you are seeing a new path. Are all your disks visible that are listed in the mapfiles for each VG?

Please elaborate.

[John Warminski]

The issue is that I need to rebuild the volume group(s) .. since the restore didn't rebuild them .

Found that pvdisplay couldn't identify the
physical volume or volume group

physical volume "/dev/dsk/c2t12d0" belongs.
pvdisplay: Cannot display physical volume "/dev/dsk/c2t12d0".

Wanted to be sure the drive was present, thus ioscan .

vcfgresotre was a sanity check.

....... moving on .....
diskinfo on c2t2d0 and c2t12d0
both show can't open nodevice or address


At this point also found the new
but seemly writeable c2t2d0.

[John Warminski]

c2t2d0 was not there before the restore.
c2t12d0 was the original device ....
( note creation date in ls )

moving on ...

dd fails on both devices .... think I'm going to call this a bad drive .

[Sridhar Bhaskarla]

Hi John,

Did you change any scsi ids before you start restoring the OS?. Do you have a drive at the scsi id 2?

Well. My inclination goes towards bad disk only. But the new disk that wasn't there before bothers me suspecting a conflict with your disk subsystem. What backend do you have?

-Sri

[Helen French]

It might be a bad drive. Check with 'stm' before you call the hardware support. Also, make sure that the device files are poitning to the correct device. Check and verify the disk's SCSI ID if accessible (ID 12 for c2t12d0). If you think you need to recreate the device files use 'insf -e'

[Todd McDaniel_1]

John,

Have you checked any logs? I am sure there is an ignite log which will show why your ignite rebuild of your VGs failed...

Does syslog.log show anything strange?

You are surely in a restoral mode now. I would rebuild your VGs and restore from a good backup of your data.

That seems to be all you can do now. Maybe create the VGS again and then run the ignite restoral again? That might work...not sure since the devices are different, but I think it shouldnt matter.



------------------------------------------------
I wish you had been able to create vgexport -s on your VGs so it wouldnt matter what the disk name is... -s assigns a VGID to the export and identifies the phsyical disks without regard to the device names. That would have made this very easy...

I usu create these with vgexport -s -p -m... so that I have copies of my config in a stable env.

[John Warminski]

Made no scsi ID changes.

Subsystem is an HP2896A wide-scsi on a Jamaica enclosure.

Went ahead and did insf for the hardware path just in case ....

dd fails .... diskinfo hangs, so I'm betting on the bad drive principle .

As we say in Texas ya'll for the prompt reponses .

[John Warminski]

Again ... thanks so much .

Ya'll are really quick so my replies are a little out of synch.

Todd,

In reply to your note and this sounds pretty stupid, where would I find the ignite log .... the syslog looks ok

Generally do the vgexport, however, did it after the ignite tape was made ... since the whole adventure began because a vg00 drive died I didn't have a place to begin.

[Todd McDaniel_1]

It usu would be under /var/opt/ignite I think... Im not logged in my boxes to check now.