Load Balancing Team Foundation Server 2008 (or as close as you can…)
The goal of this post is to discuss some options that may be helpful for anyone who is working on, or trying to, implement a high availability scenario with TFS. The fact of the matter is TFS (in its current versions) itself is not able to load balance. At this point in time, TFS cannot share it’s date tier among TFS instances. In other words, TFS does not support load balancing on the app tier. The best you can hope for a warm stand-by strategy.
A typical TFS implementation resembles the following:
 | As you can see from the illustration, TFS as a “system” is actually comprised of multiple “sub-systems” namely; the TFS service stack, Windows SharePoint Services, and SQL Server Reporting Services. Optionally, for purposes of example, Team System Web Access will be included since it is typically installed at the App Tier level. The data tier may or may not be on the same server…for the purpose of this post, it is irrelevant other than to again re-emphasize that the TFS data tier can only utilize one TFS App Tier at a time.
Now I’ve already told you that TFS 2005 or 2008 cannot be load balanced. I’ve also told you that TFS as a system is comprised of multiple “sub-systems”.
So here’s where we are going with this…the TFS service stack cannot be load balanced but everything else can be.
WSS, SSRS and TSWA can be load balanced. Knowing and configuring this will allow us to achieve a semi-load balanced failover ready solution.
|
So what else do we need to do to get us as close as we can to a load balance solution? Let first look at what a load balanced TFS implementation would look like.
In the above example all parts of both TFS instances are hot with the exception of the secondary APP Tier’s TFS service stack. So what has to happen to make the failover as quick and painless as possible? What we need is to have a mechanism to ensure that 2 things occur: 1) the TFS web services on the Primary App Tier are stopped and 2) we can change the data tier to look at the Secondary App Tier as the new Primary App Tier. If we can implement this type of solution, then we have achieve a very rapid warm standby failover strategy.
How we would complete this will be addressed in the next post, so stay tuned!
-Kevin