We have a five node windows 2012r2 stretched cluster, running SQL. Three nodes are on our production site and two nodes are on our DR site. We have multiple SQL instances running in production, we have SQL replication running over to our DR site, which is running multiple SQL instances too.
We have configured our three production nodes with quorum votes, the DR site does not have any votes.
So we have three quorum voters at our production site, the two DR nodes have no votes.
If two of the Quorum voters fail, we would expect the cluster to fail, as the votes left (i.e. 1) is less than half.
We have not configured any disk witness.
In a full DR scenario, we are planning to manually configure a disk witness and assign votes to the DR nodes.
We are puzzled why our cluster continues to function, when two of our three quorum voters are down. i.e. two of our production nodes go down, the third one continues to function.
The cluster validation report states "The quorum will be able to sustain failure of two nodes", I am reading "the quorum" being the nodes with votes, if this is the case, then the validation report is telling us, correctly, how the production nodes behave, when two go down, but we are still puzzled with this behavior.
Can anyone explain how the cluster stays up when two of three voters go down?