Can somebody provide an in depth explanation to how the disk witness and file share witness assign a vote to a node?
2 Answers
There is a good article on how cluster quorum works in Windows Failover Clustering here.
to summarise, for File Share or Cloud Witnesses, a witness log file is maintained with state information about the cluster, witnesses and quorum voting. Each node has a cluster service running that periodically checks in with the other nodes and the witnesses (depending on dynamic voting, node weighting etc) to validate the state and health of the cluster. If a node cannot communicate with other nodes, it will then verify connectivity with the witness, assess the witness log (if possible) and proceed from there with either taking control of the cluster or not.
For example, assume we have two nodes, Node A and Node B, and a File Share Witness (Witness A), with Node A the primary node in the cluster. If Node B, during a periodic health check, cannot communicate with Node A because a switch went down then it will try to verify cluster health by contacting the cluster witness (Witness A) and reviewing the witness log to determine health state.
If Witness A can still communicate with both nodes because it has a different path to Node A than Node B does, this information will be in the witness log and Node B will not try to assume the role of the primary node. Node A will also see this information in the witness log and continue normal operations.
On the other hand, if Witness A has also lost communication with Node A, this information will be in the witness log and Node B will try and assume the role of the primary node. Since Node A will not be able to communicate with Node B or Witness A in this scenario, its cluster service will have shut down to prevent split-brain from occurring.
A Shared Disk witness works almost identically, except the witness is highly-available storage connected to the cluster, so instead of a witness log the shared disk maintains a copy of the cluster configuration database on the disk instead of just a witness log. Behaviorally, the witness and node interactions are the same.
-
That was a brilliant answer thank you. So the disk witness and file share witness pretty much act the same way by communicating through a file or log file but disk witness has an extra thing where it keeps a copy of the cluster configuration database on the disk. Is that correct?– UjjiCommented Nov 28, 2019 at 14:02
-
Pretty much. The disk witness doesn't use a log file, it stores the witness log information in the cluster configuration database, but for all intents and purposes, they behave the same.– HandyDCommented Nov 28, 2019 at 22:33
-
-
1Not a problem. Feel free to accept the answer if it was helpful.– HandyDCommented Nov 28, 2019 at 22:37
-
Does disk witness (entire database) have any advantage over fileshare (log file)?– variableCommented Jan 4, 2023 at 6:44
I have one scenario for DC-DR activity.
Node 1 - Primary replica ----------------- DC-1 Node 2 - Secondary replica ----------------- DC-2 Node 3 - File Share Witness ----------------- DC-1 -- Standalone VM (Protected VM)
If we shutdown primary replica which is in DC-1 then secondary replica will become primary which is in DC-2. Now if we shutdown File Share Witness and it will take 15-30 minutes to move DC-2. During movement of File Share Witness node what will be the status of Availability Group and DC-2 replica. And once File Share Witness node moves successfully on DC-2 then will it up the primary replica which is in DC-2 or there will be any manual intervention required or my cluster will be down.
Please suggest..
Thanks in Advance..
-
If you have a new question, please ask it by clicking the Ask Question button. Include a link to this question if it helps provide context. - From Review Commented Jul 12, 2023 at 20:27