In a SQL Server failover cluster environment, I can move the SQL Server role to node a and the quorum drive at node b. Although it's possible, is this a correct configuration? While I was Googling around, I noticed some people believe this is considered a bad configuration. But I can't find anything in SQL Server's documentation.
Add a comment
|
1 Answer
Much like the core cluster resources, it's not wrong per se if the resources aren't on the same node. Since the cluster handles all of this, and SQL Server knows nothing about it, it's more of a question about cluster configuration than SQL Server configuration.
Windows Server Failover Clustering (WSFC) can operate just fine with that setup. If there is a physical drive used as a witness, then all nodes need to be able to see said drive. Since the drive will be arbitrated by the cluster, the partition that has the ability to gain quorum should end up with the witness. This might take a few extra seconds if it isn't currently on a node that is part of the partition that gains quorum, however it shouldn't cause the cluster to go down.
This also doesn't mean that SQL Server will or won't stay online. For example, if there is a 3 node cluster where node 3 becomes partitioned, with SQL Server and a witness drive both running on node 3, nodes 1&2 will have quorum based on dynamic witness and dynamic quorum and this partition will win. This will cause the witness and SQL Server to move to one of the nodes in the winning partition. This also means SQL Server will have went down on node 3 and then brought back up on either node 1 or node 2.
answered Nov 20, 2022 at 17:58
