cmclconfd: Cluster attributes have been removed from Volume group

[Ravilov]

Hello!

When we try run cmapplyconf -v -C on our running cluster (HP-UX 11.11 SG A11.16), process cmclconfd remove cluster attributes from LOCK VG vg_fe_out. After that, we can't
restart package without run vgchange -c y manually. Is it bug or SG feature?

Thanks.

P.S. Sorry, my English.
May 13 13:12:41 holmes CM-CMD[14372]: cmapplyconf -v -C cmclconf_newvg.ascii
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg00 for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Volume group /dev/vg00 has no cluster attributes
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_arc for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Volume group /dev/vg_fe_arc is configured exclusive
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_ora_bak for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_omni for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_ora for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_ora_arc for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_out for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Volume group /dev/vg_fe_out is configured exclusive
May 13 13:12:44 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_iof for node holmes
May 13 13:12:44 holmes cmclconfd[14374]: Volume group /dev/vg_fe_iof is configured exclusive
May 13 13:12:45 holmes cmclconfd[14374]: Querying volume group /dev/vg_fe_ora2 for node holmes
May 13 13:13:10 holmes cmclconfd[14374]: Cluster attributes have been removed from Volume group /dev/vg_fe_out for node holmes
May 13 13:13:10 holmes cmclconfd[14374]: Updated file /etc/cmcluster/cmclconfig.tmp for node holmes (length = 46352).
May 13 13:13:10 holmes cmclconfd[14374]: Updated file /etc/cmcluster/cmclconfig.tmp for node holmes (length = 0).
May 13 13:13:10 holmes cmclconfd[8383]: Updated file /etc/cmcluster/cmclconfig for node holmes (length = 23176).

[Duncan Edmonstone]

The VG containing the lock disk needs to be active when cmapplyconf is run, so that the cluster lock information can be checked/written onto the LUN.

However when cmapplyconf is run, the cluster isn't active (at least in your case I'm assuming that is the case), so the VG can't be activated in cluster exclusive mode.

If you look at the notes for creating a cluster, you'll see this:

http://docs.hp.com/en/B3936-90079/ch05s06.html#ciidhhja

HTH

Duncan

[Ravilov]

We need to add new volume group on our cluster without stoping SG (24x7). I was mistaken, vg_fe_out is't the lock vg. In other words, I don't understand, why cmclconfd remove cluster attribute from vg_fe_out.

[johnsonpk]

Hi ,

For adding a new Vg into running cluster you need not to do cmapplyconf. You can mark the vg as cluster aware by executing
vgchange -c y (for exclusive access) or
vgchange -S y -c y (for shared access)

and then you can use those Vgs for the package

Thanks!
Johnson

[Ravilov]

Why cmclconfd remove cluster attribute from vg_fe_out? This is my main question.

[John Bigg]

It is probably because you have not included it in the bottom of the cluster ascii file listing the volume groups. If it is not listed here, cluster attributes will be removed on each cmapplyconf. Add it to the ascii file and you should be fixed.

[Armin Kunaschik]

If the volume group is not listed in the cmclconfig.ascii then it won't get the cluster flag set! In fact, any volume group with cluster flag (and no entry in cmclconfig.ascii) will lose the cluster flag each time you apply a new cluster configuration.
So keep the cluster ascii config current when adding/removing volume groups!

My 2 cents,
Armin

PS: Please asssign points if you find answers useful!