Choose and configure the quorum of a failover cluster on Windows Server 2012 / 2012 R2

5.4. Quorum not reached

Remember the quorum configuration ? This defines which of your nodes and/or your witness disks / shares have voting rights and therefore in which case your cluster will continue to work or not.

5.4.1. Dynamic Quorum (since WS 2012)

Since Windows Server 2012, Microsoft has created a feature called Dynamic Quorum (which is enabled by default).
You will see in this example what it changes.

Since we have an odd number of nodes in our cluster, we use the "Node Majority" quorum mode.
This implies that a majority of voters (of nodes in this case) must be online for the cluster to continue to function.
In summary, we must have at least 2 nodes out of 3 available nodes to keep our cluster running smoothly.

So, the loss of a server is not a problem at the moment.

In addition, the service is still hosted on our 2nd server.

Suppose this 2nd server goes down, this causes ... the migration of the role on the last server of the cluster.

Oddly enough, our cluster continues to function when we should have 1 vote out of 3. Which would mean that our cluster should have stopped.
Yet, this one still works.

It's due to the "Dynamic Quorum" feature that appeared with Windows Server 2012 and which is enabled by default.
Indeed, thanks to this feature, each node that breaks down loses its right to vote.
Currently, we therefore have a quorum of 1 on 1. This is why our cluster continues to operate without problems.

Of course, if the last node fails, it will cause the cluster to shut down.

5.4.2. Classic quorum (not dynamic)

If you want to see what your cluster would do without the dynamic quorum, be aware that this feature can be disabled with a simple powershell command :

PowerShell

(Get-Cluster "Cluster name").DynamicQuorum = 0

Which gives in our case :

PowerShell

(Get-Cluster "iw-cluster").DynamicQuorum = 0

We put our 3 servers online.

The role is running on server 1.

Stopping a server doesn't change anything since we still have 2 votes out of 3 (nodes).

The cluster continues to work correctly on server 1.

But, when the server 2 breaks down, we have only 2 votes out of 3.
As a result, we no longer have the majority and the cluster manager can no longer obtain information about this server.

A little later, we see that the cluster has disappeared because the failover cluster manager can not connect to your server.
Although there is still a server available. But not enough votes to get the majority.

As you can see, only the first server is available, but the other servers of the cluster, as well as the cluster itself, are no longer accessible.

5.4.3. Cluster fallback

To restore your cluster, restart all servers of the cluster, and then click "Connect to Cluster" (if needed).

The name of the last cluster to which you were connected should appear in the "Cluster name" box.
Click OK.

As you can see, there is a critical event in recent cluster events.

During automatic shutdown of the cluster (due to loss of quorum), this event was logged :

Plain Text

The cluster service encountered an unexpected problem and will be shut down. The error code was '5008'.
ID : 1556

The 3 nodes are now in service.

To remove the event, click : Reset Recent Events.