0

In the past I have used a two node cluster setup. All the services (SQL, File Server, App Server, DTC) were passed back and forth within the cluster as normal.

Now I take a different approach. I like to build my 2 node cluster and ONLY install the Hyper-V Role and Failover Cluster Feature. Then create about 6 servers that migrate freely across the 2 nodes as needed. With this setup all "services" are now just standalone servers.

I was under the impression I install and setup the DTC service to account for the nodes throwing the SQL service from one node to the other in the middle of an SQL transaction. This way the SQL transaction/request can recover when the service comes back up on the other node it was sent to.

Now that I only have ONE SQL server service running and all node hopping is done on a lower level (via VMs) do I even need the DTC service anymore?

Arvo Bowen
  • 795
  • 5
  • 15
  • 33
  • A Hyper-V Failover Cluster safeguards against a cluster node failure. It doesn't safeguard against VM failure, application failure, database failure, etc. It looks to me like you've taken one step forward and two steps back. – joeqwerty Mar 26 '16 at 13:37

1 Answers1

1

No, you do not. Obviously nothing will ever use the DTC in the setup you made, as all there is running in the cluster are virtual machines.

THAT SAID: You also gave up "a lot" of high availability by going to moving VM's. There always is a chance an image turns corrupt. Even with the 2015 coming Storage spaces Direct that would not help.

Which is why SQL Server has long moved from shared files in a cluster to AOAG (Always On Availability Groups) where every instance keeps it's own copy of the database. Removes storage as single point of failure (which includes corruption).

TomTom
  • 50,857
  • 7
  • 52
  • 134
  • Using a virtual environment I have never seen "an image turn corrupt". The way I have it set up I don't see a chance of the VM going down (at least due to hardware). With live migration running smooth and there being at least 2 nodes running there should always be redundancy. That was the design model. So I don't fear the VM going down at any point, that's not what the question was about. I was simply asking if I needed DTC anymore. Since you did answer that with the first sentence you get the check. – Arvo Bowen Mar 26 '16 at 14:24
  • ... I'm happy with machine's being "paused" and moved to another physical node and those few seconds of a pause or downtime but If I ever get into a situation where the SQL server is being hammered 100% of the time I will look into AOAG. That definitely sounds like a smoother running SQL server environment. – Arvo Bowen Mar 26 '16 at 14:24
  • Well, wait until you see it. I have. I have both, lost a complete Raid in Hyper-V because the disc structure got totally bonkered, and a database. In both cases, defective write back cache memory of the Raid controller was the main factor - not a lot the OS can do when the raid controller is writing garbage. – TomTom Mar 26 '16 at 14:35