Volume Group Disk Problem

[Concert Global Networks]

I have a disk in a VG that's in an unavailable PV status. How do I change the status? pvdisplay does show that the PV should be in the VG.

[John Meissner]

is the vg active and available?

[LucianoCarvalho]

hi,
try to run vgsync.
vgsync vgxx.

regards

[James R. Ferguson]

Hi:

A 'vgdisplay' which returns "unavailable" for the VG status suggests an a deactivated volume group. Do:

# vgchange -a y

Regards!

...JRF...

[S.K. Chan]

Yeah but what does this shows ..?
# pvdisplay /dev/dsk/cXtXdX|grep Status
It should be consistent with the vgdisplay output ie ..
PV status ===> unavailable
Try to run ..
# vgchange -a y /dev/vgXX
to reactive the VG which this PV resides and see if it helps. If it doesn't you probably got a bad disk. Check with ..
# /etc/diskinfo /dev/rdsk/cXtXdX
and
# ioscan -fnC disk

[John Meissner]

james.... S.K...... exactly what I was getting at.

if the volume group is not available you will need to activate it with the vgchange command. I guess I should have included that in my first post

[Concert Global Networks]

Hey all,

Thanks for your help so far. Yes, the VG is active and pvdisplay shows the disk unavailable as well. Diskinfo returns a valid disk.I also tried a dd and the records in matched the records out, so I believe the disk is ok. Maybe the VGDA is corrupt? I've also tried deactivating and reactivating with no luck.

[James R. Ferguson · Accepted solution]

Hi (again):

So let's try reading the LVM header:

if [ "`xd -An -j 8192 -N8 -tc $DEV | xargs" = "L V M R E C 0 1" ]
then
INFO=`xd -An -j8200 -N16 -tx $DEV`
PVID=`echo $INFO | awk ???{print $1 $2}???`
VGID=`echo $INFO | awk ???{print $3 $4}???`
#
echo "$DEV : PVID = $PVID : VGID = $VGID"
else
echo ">>> Can't find LVMREC <<<"
fi

Regards!

...JRF...

[Helen French]

What's the status of PV in this command:

# pvdisplay -v pv_name

[Concert Global Networks]

Here's what I got after running the pvdisplay -v

root@rcptst01 / # pvdisplay -v /dev/dsk/c0t0d0
--- Physical volumes ---
PV Name /dev/dsk/c0t0d0
VG Name /dev/vgCPE
PV Status unavailable
Allocatable yes
VGDA 2
Cur LV 20
PE Size (Mbytes) 4
Total PE 38262
Free PE 47
Allocated PE 38215
Stale PE 0
IO Timeout (Seconds) default

... and so on. It did give the logical volumes but I wanted to keep this message short. James, the script is just hanging now...

[Helen French]

The unavailable PV status comes only when the communication fails between LVM and the disk. It can be either a HDD problem or LVM structure issue. From your outputs it's look like the HDD is working fine (check again the output of ioscan, diskinfo, dd, stm etc).

If you are sure that the disk is fine, then consider doing a vgcfgrestore to this specific disk:

# man vgcfgrestore

That will restore the VG structure back to the disk.

[Helen French]

Again, IMPORTANT: make sure that you have a clean backup of all the file systems and data on this disk and the VG before changing anything!

[Michael Tully]

A simple test of the disk should tell us something.

What does this do:

dd if=/dev/dsk/c0t0d0 of=/dev/null bs=1024k

[Helen French]

Michael,

First, congratulations on your new hat!

He checked the dd output and said it was okay on an earlier post.

Look back ... do not drink too much to celebrate your new hat level =))

[T G Manikandan]

I doubt some problem with the disk

With VG activated and PV status unavailable I would first suggest to check your
/var/adm/syslog/syslog.log for any scsi errors on the disk.
Also do check for the latest SCSI patches for the system.

[Concert Global Networks]

Hey all,

First of all, thanks for your help! vgcfgrestore is hanging. I've included a file with an excerpt of the syslog file. The latest XSWGR1100 patch bundle installed is B.11.00.49.3

[Concert Global Networks]

Oh, by the way, path 8/8.8.0
.255.0.0.0
is the path to the disk in question.

[S.K. Chan]

Take a look at my answer here. Somewhat the same scenario and the vgexport/vgimport seems to solve the problem.
http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xe50c237a4bc6d611abdb0090277a778c,00.html
Let us know if you have any question on that.

[T G Manikandan]

Please revert on the output of

#vgdisplay -v /dev/vg2
Do check the count of Cur PV and Act PV.
#vgdisplay -v /dev/vg1
#strings /etc/lvmtab
#ioscan -fnC disk

Please post the output


Thanks

[John Dvorchak]

This sounds eriely familiar to me. I had the exact same problem but the disk was on an EMC array. It seems that One, I should not have seen the power fail message and two I should not have seen the "can't write to the PVDA". A call to EMC provided me with a utility that repaired or masked the bad PVDA and I was able to deactivate the VG and vgcfgrestore the disk back into the VG. Status is now good. You didn't mention if this is a locally attached disk or a disk in a SAN etc.

[RolandH]

Check out these document.

Read the chapter about the "ghost disk"


Regards
Roland