Lock Disk Requirments

[Gene Kornacki_2]

I have a two node cluster with the latest MCSG software on 11.0.

My cmquerycl returns nothing for possible lock volume:

Possible Cluster Lock Devices:

LVM volume groups:

/dev/vg_ops1 Node1
Node2

and the rest of my shared disk in the same fasion.

What is the requirment of the lock vg? I edited my config file to plave it on what I'm calling vg_ops4 here is the current vgdisplay:

VG Name /dev/vg_ops4
VG Write Access read/write
VG Status available, shared, server
Max LV 255
Cur LV 3
Open LV 3
Max PV 16
Cur PV 2
Act PV 2
Max PE per PV 18432
VGDA 4
PE Size (Mbytes) 4
Total PE 17362
Alloc PE 1344
Free PE 16018
Total PVG 2
Total Spare PVs 0
Total Spare PVs in use 0

I've tried recreating the cluster from scratch, but the problem is cmquerycl doesn't see a possibility for a lock vg.

Anyone?

[Christopher McCray_1 · Accepted solution]

Hello,

This is an existing cluster and you lost the cluster lock disk? If so, check this link:

http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000062686621

Did the syslog show any errors? What were the errors? Did you include both nodes in your cmquerycl?

Hope this helps

Chris

[Ashwani Kashyap]

Usually the shared disk with the least seek time i.e the disk that can be searched/looked up or queried in the least possible time is selected as the lock disk because time is crucial in obtaining the lock in case of a cluster tie up . This disk is usually suggested by the cmquerycl command itself .

The disk can be a data disk or for that matter any other disk as long as it is seen by both the nodes .

So you can just choose any disk that you deem fit and put that in your config files and apply them to all the nodes and it still will work . THere shouldn't be any issues .

[Christopher McCray_1]

Also try this link:

http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000062686663

Chris

[Gene Kornacki_2]

NODE_NAME piggy
NETWORK_INTERFACE lan0
STATIONARY_IP xxx.xxx.xxx.xxx
NETWORK_INTERFACE lan1
HEARTBEAT_IP 111.111.111.111

FIRST_CLUSTER_LOCK_PV /dev/dsk/c5t14d0

NODE_NAME smelly
NETWORK_INTERFACE lan0
STATIONARY_IP xxx.xxx.xxx.xxx
NETWORK_INTERFACE lan2
NETWORK_INTERFACE lan1
HEARTBEAT_IP 111.111.111.112
FIRST_CLUSTER_LOCK_PV /dev/dsk/c4t14d0



This is what is in my config file and the error in my syslog.log are

ARNING: Cluster lock on disk /dev/dsk/c4t14d0
is missing!

[Christopher McCray_1]

Given the error you provided, you should look into the first doc I posted.

Chris

[Gene Kornacki_2]

Chris,

The cminilockcl is great.

Thanks, I've created it but I suppose it will be about an hour before I find out if it worked. I can't see why it wouldn't.

And thanks to others who replied. It's tough getting this and Real Application Cluster going wihtout HP or Oracle support, but it's going and this appears to be the only error. What a great combo!

[Stephen Doud]

Hi Gene,

If the cmquerycl is performed on a node that already has a cluster applied, then it is likely that there are no non-shared, non-clustered VGs left to choose from. If the present cluster is no longer needed, perform a 'cmdeleteconf' to purge the cluster and remove the cluster references from the disks.

If this is a new cluster, cmquerycl will succeed in finding a cluster lock disk (or any shared disk) ONLY when each node has /etc/lvmtab set up to see the shared VGs.

This requires the Admin to build the VGs on one node and vgimport them on the other nodes.
When each node can "see" the shared VGs, then cmquerycl can "see" them and identify the common VGs for selection of a cluster lock VG/disk.

-s.