3

I am planning to create FCI + HA solution in the same datacenter, but running into some problems. I'd really appreciate if someone can provide some guidance.

Here is the current configuration (All nodes are in same datacenter, Win 2012 R2, SQL 2012):

WINCluster1 SQL1 - Part of FCI instance SQLFCI1 (S drive for data, L drive for log, Q drive for quorum) SQL2 - Part of FCI instance SQLFCI1 (S drive for data, L drive for log, Q drive for quorum)

The above is a standard SQL FCI cluster and it works great.

SQL3 - Standalone box for HA (Separate S drive for data, separate L drive for log, No Q drive)

I added SQL3 to WINCluster1 and setup Node Majority for quorum (Removed Q drive disk witness from quorum) as I have odd number of nodes.

I removed SQL3 from possible owner for SQLFCI1.

Then, I setup AlwaysOn Availability Group AG1, with SQLFCI1 as Primary replica and SQL3 as secondary replica. The setup is Asynchronous with Manual failover and readable secondary.

When I use SSMS to failover between SQLFCI1 and SQL3 (back and forth), it works fine. When I use FCI under WinCluster1 to failover between SQL1 and SQL2 for SQLFCI1, it works fine.

But, when I shutdown SQL1, the SQLFCI1 fails over to SQL2, which is perfect, but SQL3 goes offline. I can remote into SQL3, click on drives, etc, but the cluster resources goes offline.

Same thing happens when my network engineer unplugs NIC cards, simulating NIC fail scenario, from SQL1. And both cases happens when SQL1 is online but SQL2 is down or NIC unplugged.

Here is the error message under event viewer: "The cluster service is shutting down because quorum was lost. This could be due to the loss of network connectivity between some or all nodes in the cluster, or a failover of the witness disk."

And another strange thing is, when I shutdown both SQL1 and SQL2, under WinCluster1, the current owner for SQLFCI1 says SQL3, even though there is nothing shared between SQL3 and SQL1, SQL2. SQL3 is just standalone box. It brings down the Windows cluster and nothing works. Once SQL1 and SQL2 come up, it starts working.

Hope this explains the current configuration and issue. Thank you so much for your assistance in advance.

Have a wonderful day.

  • I'm not even going to try to analyze the entire description of your environment (that's a lot of words). But to maintain quorum you know that more than half of the quorum votes have to be online, yes? Exactly half means cluster down. – Aaron Bertrand Aug 20 '14 at 18:35
  • Can you verify that you've properly removed the disk witness (though I would add it back in and use Dynamic Quorum)? It almost sounds like when the disk goes down no other node can see it and panics. Losing a single node out of 3 wouldn't do it. – Sean Gallardy Aug 20 '14 at 19:47
  • This link has some good info on determining how to setup the quorum and voting: sqlservercentral.com/Forums/Topic1329593-2799-1.aspx Sounds like you probably could have just left the quorum disk alone and just adjusted who can vote. – Alf47 Aug 21 '14 at 19:58
  • Thank you all for replying. I have removed every node and disks from the windows cluster, and added back in. First I used node majority with 3 modes (node1, node2, node3). That didn't work. Then I tried node1, node2, and witness disk, but that didn't work either. Is it possible to add alwayson on sql failover cluster (node1, node2) and HA node3? Just to test, I created another windows cluster for node4 and node5 as standalone boxes. I added witness disk for that cluster and it worked. So, the issue might be sql failover cluster and HA setup. May be that setup is not recommended by Microsoft. – Raj Aug 22 '14 at 19:03
  • I've not set up HA the way you are doing it. Each node in my cluster was also a replica. It sounds like you don't have a 3 node cluster, you have a two node cluster. Adding in SQL3 for HA doesn't make it a node, from what I understand. Since you have an even number of nodes, I would try moving your quorum drive off the cluster and onto another server. Try using WSFC quorum mode Node and File Share Majority – Steve Mangiameli Jan 16 '15 at 17:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.