Tag Archives: MSDTC

SQL 2008 MSDTC cluster problems, when coordinators collide

Well, today I’ve just finished our (dual) production fail-over cluster (almost). All that is left is to remove one of the clustered drives and carve it into 5 LUNS. The reason being is that in SQL2008/ Windows 2008 it is very easy now to run the MSDTC resource in the clustered groups, in fact, it is recommended but unfortunately drives weren’t made available for this. Why not you might ask? We’ll in a Windows 2003 and SQL 2005 cluster setting up the MSTDC was a bit of a pain. However, it wasn’t too hard to get working using the Quorum drive as a dependency as so that is what we did -this is then shared across all cluster groups. In Windows/ SQL 2008 clustering it won’t let you do this (using Quorum) for obvious good practice reasons as so that option is not open to you. However because doing it properly it no longer a technical issue in terms of difficulty it is better we try and do it properly! The reason I can’t run the MSDTC in one of my existing clustered drives is because I don’t/ can’t get this working using the dual cluster integration (which over the next days or weeks I hope to document in this blog), therefore like the Quorum drive I am going to accommodate this by having separate non mirrored (SAN MirrorView) disk resources within each clustered group that can house the MSDTC.
The fifth LUN is obviously so I can have the same (although ever so slightly smaller) disk resource to put back into one of the cluster groups replacing the LUN that was taken.

The reason why this MSDTC doesn’t seem to work is no doubt in some way related to the cluster group name SID account since once DR fail-over has taken place this is all that has really changed. Anyway needless to say the clustered MSDTC resource fails if using the original one, so due to time constraints I have had to compromise and move one.

Hopefully, in the next post I’ll document and post my screenshots on creating a cluster, and then I’ll follow this up with how I do the dual cluster DR using active/ active on all nodes. I think you are going to like it.