2

This question already has an answer here:

We have an Availability Group configured. Total number of nodes in the cluster is 3 (odd). All are in same subnet. Also we do have file share witness. Everything is working fine for now.

My question is: do we really need file share witness? And what should the ideal quorum configuration be?

As per my knowledge, it is not needed as we already have an odd number of nodes, so node majority quorum configuration should be fine and is recommended.

SQL Version: Microsoft SQL Server 2016 (SP2) Enterprise Edition
OS Version: Windows Server 2016 Standard

marked as duplicate by Josh Darnell, Michael Green, Colin 't Hart, mustaccio, LowlyDBA Jan 22 at 17:02

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • 3
    No, it's not strictly necessary since you have an odd number of cluster nodes. However, it's important to consider your failure modes. Where are the nodes located in relation to each other? Do they share any other resources (same rack, same SAN?) Personally, I've always used a fileshare or cloud witness as a belt and suspenders approach even when I've been fortunate enough to have more than a 2-node cluster. – Jonathan Fite Jan 21 at 18:11
  • @JonathanFite Thanks for response. I didn't get the meaning of failure modes. Also it would be great if you can tell how nodes location in relation to each other can help us decide about quorum configuration. Sorry I am little naive when it comes to clustering. How having a witness is more safer approach...is it because even though nodes goes down..probability of file share going down would be less as it would have been created at a very reliable location in network?? – sachin-SQLServernewbiee Jan 21 at 18:41
  • @JonathanFite Also FYI, Dynamic quorum is also enabled on cluster. So I guess the current configuration is ideal??? – sachin-SQLServernewbiee Jan 21 at 18:45
1

In my opinion , based on version of windows you have and especially using always on configuration i would recommend having a quorum configured. In your case its File share which seems OK or else you have the option of quorum drive witness.

More on this is very neately explained by @Brentozar here

You don't have a choice as to whether or not to have quorum: Always On Availability Groups are built on top of Windows Server Failover Clustering, which requires quorum voting to understand who the primary is for a given AG at a given time.

I think what you might be asking is whether or not you need an old-school quorum drive (like a Q drive on a SAN). These days, Windows Server Failover Clustering (WSFC) can either use a quorum drive witness, OR a file share witness.

Windows Server 2012 R2 includes dynamic witness capabilities: if you have an even number of voters (like if 1 of your 3 cluster nodes goes down), then Windows will automatically add in your witness in order to be a tiebreaker. That's why you want to configure the witness (either a disk or a file share) ahead of time - Windows will automatically manage its voting rights whenever you have node failures or network failures.

Also refer to another part of the answer which should help you here

In a 3 node cluster you can setup node majority as your quorum mode and then you can survive the failure of a single node within the cluster . That is in most cases sufficient. If your cluster spans multiple sites you might want to fiddle with the quorum-vote weight depending on your failover plans

  • Just to be 100% clear, please let me know if my understanding is correct. Scenario 1: In this 3 node cluster with no shared storage, cluster will only be able to handle single node going down. If more than one goes down, entire cluster goes down. Scenario 2: Since we have 3 nodes cluster, no shared storage along with File Witness and Dynamic quorum enabled, the benefit we have with this config is that cluster will still be running even if two nodes goes down and one node and witness is up and running. Is my understanding correct? – sachin-SQLServernewbiee Jan 22 at 8:26
  • @sachin-SQLServernewbiee I have added different answer to your above questions. – Shanky Jan 22 at 9:02
  • @Shanky thanks for input. I got your point. So dynamic witness will only take the vote away of the node when it is purposefully shut down not in the case of unplanned shutdown? – sachin-SQLServernewbiee Jan 23 at 16:05
  • Yes you are correct – Shanky Jan 24 at 4:03
1

Creating Answer from comment:

In this 3 node cluster with no shared storage, cluster will only be able to handle single node going down. If more than one goes down, entire cluster goes down

No for 3 node cluster, assuming all nodes are voting, if one node goes down still you have 2 nodes(more than 50 % of votes/nodes) voting which is sufficient to keep WSFC up and running so WSFC will not come down if one of the node goes down.

Scenario 2: Since we have 3 nodes cluster, no shared storage along with File Witness and Dynamic quorum enabled, the benefit we have with this config is that cluster will still be running even if two nodes goes down and one node and witness is up and running. Is my understanding correct?

Yes, but let me add dynamic witness only protects from planned failovers or downtime. In your case, if 2 of the 3 nodes went down abruptly dynamic witness is not going to work and WSFC will come down( lack of quorum one node and FS witness is not greater than 50% of votes). The fact is simple, with sudden downtime registries are not updated to change witness dynamically hence dynamic witness is failed to perform its duties. Dynamic witness and quorum gives "last man standing" feature but subject to condition the nodes are gracefully shutdown

Not the answer you're looking for? Browse other questions tagged or ask your own question.