vg00:one disk failed, and some LVs on it not mirrored.

[stephen peng]

dear all,
I've got a situation, there were two disks in vg00, and os based LVs were all mirrored, but others were not, somes were even cross two disks. now one disk failed, the LVs on it could not access the filesystem on it hung. I reboot the system using alternate disk, how could I remove the non-mirrored LVs(on the failed disk and across two disks) to clear the configuration of vg00. I tried to run lvremove, it posted:
lvremove: Couldn't query physical volume "/dev/dsk/c2t0d0":
The specified path does not correspond to physical volume attached to
this volume group

thanks a lot!

[singh sanjeev]

replace the failed disk
#vgcfgrestore -n vg00
#recover the FS data from backup..

or you have no intention to replace the failed disk...

[stephen peng]

since one disk was failed at vg00, so the mirrored root LV's status were stale. we had application demand to free disk space on other disk, so someone backup /etc/lvmtab and vgscan, then, since c2t0d0 was unable to query, it was kick out from vg00, and he could use lvremove to remove all those massive LVs, then it raised another problem, the root LVs remain stale status.
1, why those massive LVs could be lvremove after vgscan?
2, how to tune the status of root LVs to sync?
could someone please explain these to me? thanks!

[stephen peng]

can I restore the former /etc/lvmtab, replace the failed disk and vgcfgbackup? pretty doubt about it.

[Aneesh Mohan · Accepted solution]

I belive you are facing problem with a ghost disk.You can remove failed PV using PVkey in that case.

#lvdisplay -v -k /dev/vg00/lvol1

#lvreduce -k -m 0 /dev/vg00/lvol1 1 (Substitute 1 with stale key in your case)

#vgreduce -f vg00


#mv /etc/lvmtab /etc/lvmtab.org

#vgscan -v

#vgchange -a y

#lvlnboot -R


Aneesh

[R.O.]

Hi,

Take a look at this document from HP:

http://docs.hp.com/en/5991-1236/When_Good_Disks_Go_Bad_WP.pdf

Regards,

[stephen peng]

guysï¼
I don't know whether the definition of "ghost disk" at the WP fit my situation. At WP, /etc/lvmtab still contains the failed disk, but in my situation, /etc/lvmtab does not contain the failed disk, could I lvreduce those LVs still ?

[R.O.]

Hi,

If your disk is completely dead (it does not appear neither in lvmtab, nor in vgdisplay -v vg00 - pvdisplay -diskinfo, you can replace it (if it is hot swappable) and do the vgcfgrestore and further steps.

Regards,

[stephen peng]

R.O.,
how can I vgcfgrestore a pv to a vg when this pv is not in /etc/lvmtab?

[R.O.]

Hi,

If you cannot vgcfgrestore, you always can vgextend the disk and re-mirror the lvols with lvextend.

Regards,