We’ve had an ancient source control server, which were running Team Foundation Server 2005. Yes, that classic 1.0 version! Because the underlying OS, the SQL Server, the SharePoint framework as well as the hardware were becoming quite outdated, I decided to upgrade it to the latest 2013 version. Unfortunately you cannot upgrade from 2005 to 2013 in a single step (which is not a surprise because there were 3 versions between them), but first you have to upgrade to 2010, and then from 2010 to 2013:
Upgrading from 2005 to 2010 is a bit tricky, because you have to upgrade SQL Server and SharePoint as well, but you can find a good step-by-step guide and a best practices collection on the TFS Setup Support Team blog. The TFS Team also published a very useful article on upgrading WSS 2.0 to 3.0 for TFS.
In theory upgrading TFS is fairly simple: just uninstall the previous version, but leave the SQL databases in place, then install the new release and choose Upgrade in the install wizard which will update the SQL database schema to the latest version. In theory it does not look complicated, what’s more it looks simple if you use Tim Elhajj’s 61-page Upgrade Team Foundation Server 2012: The Ultimate Upgrade Guide.
However the practice can be a bit different.
First, none of the above blog posts call your attention that there is a serious bug in the TFS 2010 upgrade process which may cause inconsistencies in the database, so you have to be very careful with the upgrade. You can read more about it in Brian Harry’s TFS 2010 Upgrade Issue blog post which contains a link to the hotfix as well.
If was even more frustrating that after successfully completing the upgrade wizard I could not connect to the server. Different clients greeted me with different error messages, for example:
TF400324: Team Foundation services are not available from server MyServer\MyCollection. Technical information (for administrator): Unable to connect to the remote server.
TF205020: A connection could not be made to the following server: MyServer\MyCollection. This server was used during your most recent session, but it might be offline, or network problems might be preventing the connection. Contact the administrator for Team Foundation Server to confirm that the server is available on the network.
TF31002: Unable to connect to this Team Foundation Server
It even happened that Team System Web Access running on the same server could access the data, while remote Visual Studio clients could not connect to it. According to Network Monitor the communication went between the hosts flawlessly, but the server mostly returned HTTP 4xx and 5xx errors (which were visible in the IIS log anyway). It also happened, that I could connect to the server and manage the work items, but could not access the version control and the file history data.
Particularly interesting was that the clients could successfully connect to a new collection I created on the upgraded server, which meant that there were no problem with system level settings (eg. port, firewall, certificate), but the issue was somehow related to the upgraded database. I even tried the Best Practices Analyzer from the TFS 2013 Power Toys, but it could not find any relevant issue neither on server, nor on collection level.
Because I ran out of ideas, I went down to database level. I monitored the queries with Profiler and looked into the tables via Management Studio. I compared the data of the freshly created, working collection database with the records in the database of the upgraded collection. Finally I have found the tbl_ServiceDefinition table which seems containing the URL endpoints of the TFS .ASMX webservices. It was interesting that while all cells contained data in the working database, the table in the upgraded database contained several NULLs in the RelativePath column. So I took a deep breath, created a snapshot of the VM, and manually updated the records in the tbl_ServicesDefinition table. I changed the values of the RelativeToSetting, RelativePath and IsSingleton fields in those rows where RelativePath was NULL, and I could find a matching row in the working database based on the GUID value in the Identifier column. This is my final version:
On the clients I cleared the Visual Studio Team Explorer cache folder (C:\Users\username\AppData\Local\Microsoft\Team Foundation), added the server again, tried to connect aaaaaaaaaaaaaand it worked!
I have to add that this is a total hack! It is not a good solution, not supported and not recommended, it just works on my machine. If you have similar issues, you probably better to contact with Microsoft via the official support channels.
After this using the Configure Features wizard and manually updating the process templates was not that terrible at all.