Cluster LOCK issue

[srinimar]

Hi,

I need help on this cluster lock information. in cluster config file we will mention the device disk name here some thing like below one how to we know this is the specific disk we need to mention here. how to know that..

FIRST_CLUSTER_LOCK_PV /dev/dsk/c4t5d2

[Kevin Wright]

cmquerycl will create that for you, otherwise you can use any disk you like, just ensure it is the same PV referenced from each node, the path may be different from each node.

[Steven E. Protter · Accepted solution]

Shalom,

cmquerycl list the node names so it knows what systems you want in the cluster

cmquerycl -n node1 node2

See the man page for more options.

The lock disk just needs to be there on shared storage. It does not use a lot of space, it determines who wins and remains running when the cluster heartbeat fails.

SEP

[srinimar]

Thnx for that info.

and

i am planning two node cluster. so is it compulsory to have cluster lock. or can I go without cluster?

or

a quorum server.

[Steven E. Protter]

Shalom,

I believe you have a choice with a two node configuration. You can go either way. Cluster lock does not normally work with three node or greater.

SEP

[A. Clay Stephenson]

No, on a 2-node cluster, a cluster lock is absolutely mandatory. However, I would advise you to seriously consider a Quorum Server rather than a lock disk. The QS is much more flexible and is really considered the better way to resolve cluster locking these days. A little Linux PC makes a dandy QS and the QS software is free.

[srinimar]

Hi,
while activating cluster lock. which VG we need to define here....how to know that

#vgchange -a y vgXX

-----
Do i need to re-boot the server while setting up cluster? I mean at any point of time? from starting to ending?

[A. Clay Stephenson]

No, there is no need to reboot. You can vgexport -p and vgimport while the vg is not cluster aware and then do the vgchange -c y later. That operation need only be done on one node and because this is written to the LVM metadata on the disks, once done the change will be effective on all nodes which attemp to activate the volume group.

[melvyn burnard]

there is no need whatsoever to reboot any server while doing the cluster creation, but once done it is good practice to reboot to confirm the cluster starts correctly and any packages do what they are supposed to do.
As for the cluster locking, for a 2 node cluster it is mandatory, whether you use a lock disc or a Quorum Server. For a 3 or 4 node cluster a lock disc is optional but recommended as is a QS, and for 5 nodes and above you cannot use a cluster lock disc, you must use a QS if you choose to have a lock facility.

[srinimar]

yes i will do this later but i want to know

#vgchange -a y vgXX

which VG do i need to select while doing this..as i will have a lot?

and can you give me how to test the cluster setup i mean all the PKg's are failovering to other node or not and if i have any issues how to roll back it??

[melvyn burnard]

I suggest you read the managing Serviceguard manual for ideas on testing etc.
As for which VG, well if you are using a cluster lock disc, then the VG containing this disc MUST be activated with vgchange -a y at the time of doing the cmapplyconf, and MUST appear in the cluster ascii file.
Any other VG's can either be done at the same time, or marked as cluster aware when creating the package, using vgchange -c y

[srinimar]

Cluster lock issues is resolved now.. I have created it. Thanks