adding a new cluster

[chrif]

Hello,
In a cluster of two nodes, with MC/ServiceGuard 11.14 and HP-UX 11i is there a particular way to add a cluster !!
I already tried to do that as it is explained in the documentation but it do not work (either by using sum or by using command line )
error message is :
unable to determine a unique identifier for physical volume /dev/dsk/cXtYdZ on node NODE_NAME use pvcreate to give the disk an identifier.
but the problem is that this disk (adress) is unavailable for the moment (VA controller 1 out ) but we can access to this disk using the other controller.

help needed
Thank you.

[Armin Feller]

Hi,

I understood that you tried an "cmapplyconf" and got the message:

Error: Unable to determine a unique identifier for physical volumn ...

There are two ways to solve that:

a) Please make the disk available again.

b) reduce the disk from the volume group
# vgreduce -f /dev/vg_name

Regards,
Armin

[Christopher McCray_1]

Hello,

Your 1st question:

In a cluster of two nodes, with MC/ServiceGuard 11.14 and HP-UX 11i is there a particular way to add a cluster !!

You can't add a cluster to a cluster. Clusters are independent of each other. You can, however, add a node to a cluster.

The message is probably due to the fact that disk is in the lvmtab and serviceguard is expecting to be able to get information on it and can't. Do as suggested and vgreduce the disk out of the vg

Hope this helps

Chris

[Rajeev Shukla]

If you say you can access that disk by another controller is the alternate path configured. If not configure the alternate path for the VG with this disk or else make this disk available again and then try.

Rajeev

[Christopher McCray_1]

Hello again,

Your disk has two paths.


#vgreduce /dev/dsk/ vg??

This removes the disk on controller 1, leaving the other disk. When youget the controller fixed, readd with vgextend.

export the new lvm information and update the other nodes

# vgexport -v -p -s -m /tmp/vg??.map vg??

on the other nodes after copying the mapfile:

# vgexport vg??
# mkdir /dev/vg??
# mknod /dev/vg??/group c 64 0x??0000 -- the minor number must be the same on all nodes.
# vgimport -v -s -m /tmp/vg??.map vg??

Done. Just be sure to vgexport/vgimport after the controller is fixed and you have readded it to the vg on the original node.

Hope this helps

Chris

[chrif]

Hello,
I asked a wrong question !!!
her is the real problem
In a cluster of two nodes, with MC/ServiceGuard 11.14 and HP-UX 11i is there a particular way to add a package!! (2 packages are already there)

[chrif]

rehello,
are you sure tha MC\secviceguard needs the tow(all) physical physical on the volume group to add a pkg ???

[Christopher McCray_1]

Hello again,

If it cannot query all physical volumes that are supposed to be shared/present in clustered volume groups, it will complain.

As for your first question:

http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000062922821

http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000062686627

These and other well-done docs may be found in the technical knowledge base using the keyword "uxsg" or "umcsg"

Hope this helps

Chris