Quorum has been in Windows Cluster services for a while, but since Windows 2008 R2 and Server 2012 some great new features have been introduced.
Basically, Quorum is a feature of clustering which decides if the numbers of active nodes in a cluster is above the threshold of 50% else Quorum takes the cluster offline.
If you had an even number of nodes (e.g 4), and two of the nodes went down the Quorum would be exactly 50% in which technically the cluster will shut-down even though half the estate is fully working. For this reason the method of Node & Disk Majority would be implemented.
Node & Disk Majority
This scenario would only really be used when you have an even number of nodes as you are adding an extra ‘vote’ which comes from a disk, thus making it an odd number of votes. This disk can be called the ‘Disk Witness’ or ‘Quorum Disk’, which is a small cluster disk which is highly available so it can fail over between the different nodes.
So in this situation Node 2 has gone down, and without the disk witness the customer service would go offline. However as the disk witness is on-line and now on Node 1, in total it gets 3 votes out of a possible 5 and therefore the customer stays on-line.
Node & File Share Majority
This is very similar to the disk witness but is appropriately named the File Share Witness. A file share is set-up on a server in the same domain forest as the clusters which all the nodes have access to. A node will place a lock on that file share, but if the current FSW node fails then another node will lock out that file share.
In the same fashion as the disk witness, the File Share Witness has a vote and thus tips the balance if a sub-set goes down.
A good video which explains this is here
Server 2012 introduces a new feature to Quorum so that it is more flexible with regards to running maintenance on your nodes. If you have 4 nodes in sub-sets of 2 – 2, then running maintenance on servers could cause potential downtimes on the whole cluster.
Currently if you took down server 2,Quorum would treat this as 25% down and would continue working. However if an admin applied the same maintenance to a 2nd server, or if another crashed then Quorum would show this as 50% down and would shut-down the cluster service.
Now, in server 2012 this is alleviated by Dynamic Quorum which adds support for planned downtime. Basically when a server is taken down due a to a planned restart / shut-down then that node has its vote revoked until it comes back up again. So in the above scenario instead of it being 3 out of 4 votes, it would be 3 out of 3.
Once you get down to the scenario of only 2 nodes being up Dynamic Quorum does something different. It randomly removes voting from one of the servers, therefore only having 1 vote in total against the 2 nodes.
The benefit of this is that there is now a 50 / 50 chance of the cluster surviving another node failure. If server 4 dies then server 3 still has the only vote and thus is working with 100% of the votes and the cluster keeps going. If server 3 dies though then all the votes are gone and therefore the cluster shuts down.
Configuring Quorum on Server 2012
To be able to use Quorum we first need the feature ‘Failover clustering’. This can be added via the “Add Roles and Features Wizard’.