I recently had the need to move the quorum for a cluster to a new SAN drive. It’s a quite simple thing to do, but some of the options and locations of the cluster resources itself are a little hidden in Windows 2008. Let’s just walk through the basic steps.
The first thing you need to do is get the new drive added to the cluster and put into the proper group. If the drive is being presented from a SAN and you have multiple paths to that SAN, then you will need to setup whatever multi-pathing software you have. If you’re not sure, then an easy test is to open Window Disk Management and see how many new drives are there. If there are more drives there than you are supposed to have, then you need to setup your multi-pathing. Once you get that setup and reboot, you should now see the correct number of drives. If you are on a cluster, you need to do this on each node in the cluster.
Now that your drives are correctly presented to the OS, it’s time to bring them online and format them. You can do this from the node in the cluster that your services are currently running on and you only have to do this on that one node. Right click the drive in Windows Disk Management and select online. Now that the drive is online you need to right click it again and select initialize. Finally you need to format it by right clicking the drive and selecting New Simple Volume. Walk through the wizard giving it a drive letter and formatting it with a label. Don’t worry about what drive letter you assign, as we can change that later.
You are now ready to add the drive to the cluster. Open the Cluster Manager MMC, expand the cluster, and select storage in the left hand pane. In the right hand pane you will see all of your cluster groups with their assigned storage below the group. Select Add Disk in the actions pane and select the disk to add. You will now see a group called Available Storage where you will find your new drive.
At this point everything is now setup and ready for the move. In the left hand pane, click the cluster name at the top of the list. In the actions pane go to More Options and select Configure Cluster Quorum Settings. Follow through the wizard and the only thing you need to change is on the Configure Storage Witness page. On this page, simply select the check next to the new storage disk. You have now successfully moved your cluster quorum to another drive, but let’s verify it. Open Windows explorer and view the new drive. You should now see a folder called Cluster, which verifies the move. Just to be clear, what you have really done is not move the quorum but created a new one.
If you want to reuse the old drive letter (Typically Q:), then right click the old drive under the Available Storage group and select Change Drive Letter. Now you can right click the old drive and select delete, which will remove it from the Available Storage group in Cluster Manager. In order to assign your new drive to the previously used drive letter, simply right click it in Cluster Manager and select Change Drive Letter. The old drive letter will now be available, so simply select it. After changing the drive letter you now need to restart the cluster service or run the following command from an elevated command prompt to move it to another node.
cluster group “cluster group” /move:nodename
More Information – If you are moving your cluster quorum drive you might want to check out the following posts:
How to move the Master database
How to move the Model database
Thanks so much for posting this. I had created 2 Windows 2008 clusters using SQL 2008 on VMware ESX 4.0 Update 2 hosts some time back with RDM disks on an EMC storage array. I now need to migrate the quorum, MSDTC, and SQL data drive to another FC LUN (still using RDM disks). This post is exactly what I needed and the related MSDTC post. Thanks!
Thanks for the feedback! I’m glad it helped it you out. Good luck with your move and post back if you run into something interesting along the way.
So, this process can be executed while the cluster is up and running during normal hrs or do I need to do it during a planed outage time?
I would always do something like this during a planned outage.
Very well done…
Just a small question
Does it work for a Windows 2003 cluster with exchange Server 2003 as well?
Txs
Yes it does.
Good articles. Got some ideas from you regarding quorum and msdtc. However, the way I did storage migration for a 2008R2 cluster was:
1) show new drives to MSDTC and each SQL instance cluster groups
2) assigned temporary drive letters to each drive
3) offline each cluster group one at a time but online the drives on the offlined group
4) copy stuff from old to new drive with robocopy
5) remove drive letter from old drive and assign old drive’s letter to new drive
6) update the sql server (and msdtc) dependencies
7) finally online the cluster group
This exercise to each cluster group and you’re done.
Once I’ve done storage migration for a SQL cluster by just shutting down both servers, copying data from old to new LUNs with some imaging tool and then brought up the servers. It was kind of messy to get 2008R2 cluster manager to understand which drive is which but luckily there was only one instance and maybe 5 LUNs. Worked out okay. But I’m not doing that again… this robocopy migration is quite simple.
I am in the process of moving my windows 2008R2 active/active cluster from a older storage array to a newer storage. all the volume will be miragrated by the vendors replication tool. I just need the steps to move my quorum drive
If you are planning on shutting down the entire cluster, migrating the data, and then bringing the cluster back up on the new SAN then I would first change the quorum type. I would create a file share and change it to Node Majority and File Share. This way when you bring the cluster nodes back up they can find the Quorum drive and be happy. Then once you get the new SAN drives presented and ready you can just change the Quorum type back to Node and Disk Majority pointing to the new drive and do away with the share. If you have an odd number of nodes you could just use Node Majority instead.