newfs and vgchange fails

[Robert W. Kieras]

I have a volume group in a wierd state. It has
one PV and one LV. When I do the command :
newfs -F vxfs -o largefiles /dev/vgnfs5ec/rlvol1 I get:
newfs (vxfs): Cannot open /dev/vgnfs5ec/rlvol1
newfs (vxfs): No such device or address

Vgdisplay looks normal. There are no collisions on vg numbers for group files. lvdisplay looks normal. I'm running MC/ServiceGuard and the volume group used to be in a package. I have deleted the package with cmdeleteconf -p nfs5ec. I would like to delete this VG and start over but when I do a vgchange -a n /dev/vgnfs5ec I get:
vgchange : Failed to notify clvm daemon about volume group deactivation - No such file or directory. Couldn't deactivate volume group "/dev/vgnfs5ec": No such file or directory.

The /dev/vgnfs5ec look normal.Vgdisplay shows the volumen group is active in exclusive mode.
How can I delete this volume group and clean up this mess ? Thanks for looking at this problem !

[Sanjay_6]

Hi,

First remove the cluster info from the vg, Unmount any lv's mounted from this vg and, do this

vgchange -a n /dev/vg_name
vgchange -c n /dev/vg_name

Now delete the VG,

vgremove /dev/vg_name

or

vgexport /dev/vg_name

Hope this helps.

Regds

[Jeff Schussele]

Try a vgchange -c y /dev/xxxxx
To make it cluster aware again.
Then try to vgchange -a n it.

HTH,
Jeff

[Jeff Schussele]

If you can't make it cluster-aware again & if you don't care what's on it - just nuke it with vgexport /dev/XXXXX

Jeff

[S.K. Chan]

Remove the cluster definition first by doing ..

# vgchange -c n /dev/

before you vgremove it.

[Robert W. Kieras]

Thanks everyone.

I tried vgchange -c y /dev/vgnfs5ec and got

vgchange: The volume group "/dev/vgnfs5ec" is active on this system. Cannot perform requested change.
Then I tried vgchange -a n /dev/vgnfs5ec and got
vgchange: Failed to notify clvm daemon about group deactivation - No such file or directory couldn't deactivate volume group "/dev/vgnfs5ec"

Earlier it tried blowing it away with vgexport or vgremove but when I recreate it I get the same behavior.

[melvyn burnard]

well the first question needed to be asked is who activated the vg? If it was activated in a package, try halting the package. what about the other nodes in the cluster?
you need to check and compare across the nodes in case there is a conflict.
You may have to resort to rebooting this node to free this confusion up, but first make sure that the vg will not get activated by any package that moves to an adoptve node.