cluster lock vg

[N Ward]

I'm recreating a cluster at a DR exercise and I have got a LUN presented to both nodes, for the purposes of creating the vg_lock.

I created vg_lock. It has too many paths for LVM apparently.. but thats another issue, I'm presuming it will just ignore some (maybe) When I run cmcheckconf it complains vg_lock is not active, but it is on both nodes. Does it need to be to run cmcheckconf?

Do I need to choose the device showing at the top of vgdisplay output for vg_lock in the cluster config file (i.e. not one of the alternate paths). I chose the SAME device name for each entry when I created the file. Node 1 is showing c12t4d7 Node 2 is showing c38t4d7.

[Mel Burslan]

I know this is not the answer you are looking for but knowing the pain of dealing with SAN volumes as cluster lock vg's, can I kindly suggest using a quorum server on a non cluster member server accessible from bith cluster members ? qs (quorum server) process is very simple to set up and a very low overhead process. Once set up properly, can serve multiple clusters. If it goes down, while the cluster is running, it is no big deal. You just restart it.

When you mention DR sites, you most probably are dealing with a SAN infrastructure that is not what you are accustomed to, hence proven to be more headaches that you bargained for. If you are keen on using the vglock, you need to work with the DR site's SAN admin and a very sharp one at that, to sort out which paths are actually need to be used, as the back-end of the SAN (type/model of fabric, SAN storage brand/model) has a lot to do with how the storage is presented your system. It is quite hard if not totally impossible, to give you an answer that is not speculation at best.

[N Ward]

I'm not able to do that unfortunately. I have to use the SAN since that is how it is done back at base. Plus we have been assigned a certain amount of equipment to match our prod environment. Thanks for the information though.

[Vishu]

vg_lock(shared VG) should be active on the node while doing cmcheckconf, cmapplyconf commands.

Also, as i know, both nodes must have the same device for their cluster_lock_PV and they should show the same in both. if node1 is showing c12t4d7, then node2 must show the same.

[Duncan Edmonstone]

>> When I run cmcheckconf it complains vg_lock is not active, but it is on both nodes. Does it need to be to run cmcheckconf?

vg_lock need only be active on the node where you run cmcheckconf...

you certainly don't want it active on both nodes as that implies potential corruption. (a cluster VG is only ever active on one node of a cluster unless you are using SLVM with Oracle RAC, even then the lock VG needn't be)

again though I'm surprised at the number of people who have an entirely seperate VG just for a lock disk... this isn't MSCS with its quorum disk! You don't need to waste LUNs from your disk array on this. Any disk thats in a cluster VG can be used as the cluster lock and still have data on it and be used in the normal way.

HTH

Duncan

[N Ward]

cmcheckconf now works, and after making vg_lock cluster aware I was able to run cmamcpplyconf. Many thanks....