how to recover a VG with its devices files lost?

[stephen peng]

dear all,
I've got network issue to cause MC ServiceGuard halted and server rebooted, after it's reboot,a package could not startup cause it could not mount a specific filesystem.one VG configured in the package lost its device files for no reason including lv* and rlv*(just one lv in that vg), vg's configuration on other server was not a correct copy. there was no map file backup for this vg, how could I achive to recover this vg?

thanks a lot!

[Kapil Jha]

Do you have old lvmtab file, you can check the disk from lvmtab

else
restore the volume configuration from
/etc/lvmconf directory.

man vgcfgrestore
for more information.

What HPUX and service guard?

BR,
Kapil+

[rariasn]

Hi Stephen:

Restore files /dev/vglost/* from backup.

Rgs,

[Elmar P. Kolkman]

Or do a vgscan to find out which disks belong together in a VG and then do an import on the disks that most likely hold the VG you are missing. The only thing, before doing the vgimport, you have to do, is make sure that those disks are not in use by an active cluster package.

[Elmar P. Kolkman]

(That, of course, only works if you have the device files for all disks... If not, do an ioscan -fnC disk, followed by insf (no options!))

[stephen peng]

the VG was there, only that it missed its lv related device files.

[Viktor Balogh]

>the VG was there, only that it missed its lv related device files.

try to activate the VG and see if those LV devices will get created.

[stephen peng]

it is not an option to create lv device files by activating vg

[Viktor Balogh · Accepted solution]

So, this was happened on the node on which the vg in question is deactivated? take a map file with "vgexport -s -p -m vgname.map vgname" and re-import it on this node.

[chris huys_4]

Hi Stephen,

So you're saying that a vgdisplay -v /dev/ still shows the logical volume with the "disappeared for no reason" /dev/lv* and /dev/rlv* device files ?

Like victor says, vgexport/vgimport would be the proper way to get the device files back, without specifying a mapfile as a backup seems also not have been taken, so afterwards you will need to rename the device files again that dont have a "standard" lvol* name.

If you have a ll output of the /dev/ directory in question, which a good sysadmin should have ;) , you can also manually mknod the device files back into existence, preserving the minor and major with which the originally /dev/lv* /dev/rlv* were created.

Greetz,
Chris

[Emil Velez]

you can use mknod to create the device files for the LVs.

mknod c 64 0xXX000D /dev/vg01/rlvol9
mknod b 64 0xXX000D /dev/vg01/lvol9

you need to know the volume group minor number XX and the LV number in the volume group

The name of the LV can be anything but the minor number will be used to access the correct LV device in the VG.

[Kapil Jha]

As stated earlier if you have no information about LVs,
just make sure you have /dev/VG_name directory
and group file inside this directory and import the VG.

#/sbin/vgcfgrestore -n vg_name -l
to check

and /sbin/vgcfgrestore -n vg_name PV1 PV2 ....

BR,
Kapil+

[stephen peng]

I would try to vgexport that vg and then vgimport it