ServiceGuard cluster & volume group failover

UNIX HP-UX
1

I have a 2-node ServiceGuard cluster. One of the cluster packages has a volume group assigned to it. When I fail the package over to the other node, the volume group does not come up automatically on the other node.

I have to manually do a "vgchange -a y vgname" on the node before the package will come up. If I fail it back to the original node, I once again have to issue the vgchange command manually before the package will start.

What am I missing? I see some cluster related options in the vgchange man page but I don't understand if/how/when to use them.

What do I need to do for this volume group to automatically come up when I fail the package over?

TIA! :slight_smile:

2

And not the other packages?

??

Is this VG a new VG? and you forgot to add in your <packagename>.cfg file ?

3

The other packages do not have VG's assigned to them, only this one.

It is a new VG, and it has been added to the packagename.cfg file.

The VG comes from SAN disks that are presented to both nodes. Everything works fine, except for the fact that a manual vg activation has to occur any time we move the package from one node to the other.

4

in the file /etc/cmcluster/yourpackage.cfg:

# List of cluster aware LVM Volume Groups. These volume groups will
# be used by package applications via the vgchange -a e command.
# Neither CVM or VxVM Disk Groups should be used here.
# For example: 
# VOLUME_GROUP          /dev/vgdatabase
# VOLUME_GROUP          /dev/vg02

VOLUME_GROUP            /dev/vg20
VOLUME_GROUP            /dev/vg21
VOLUME_GROUP            /dev/vg22
...

From vgchange man page:

You should use vgchange -a e VGXX

5

^^ I think that's what I needed. Thx!! :b:

6

Ok, I just got a chance to test this. Here's what I had to do:

primary node: vgchange -a n vgname
secondary node: vgchange -a n vgname
primary node: vgchange -c y vgname
primary node: vgchange -a e vgname

And now it's failing over properly! yay!

thx again!