I no longer work at Microsoft, so please don't bother leaving a comment here or trying to contact me through my MSDN blog.
You can find my new blog at http://www.technologytoolbox.com/blog/jjameson. My new site also provides copies of all posts from my MSDN blog.
I've been a bad blogger this month.
Almost a month ago, I wrote a post about using Web standards with Microsoft Office SharePoint Server (MOSS) 2007, but I noted that there would be more to come on that subject in the near future. Well, almost a full month has passed and I still haven't posted "part 2" yet. Sorry about that.
As I sat down to write the next post in that series, I decided to take it in a different direction in order to make it more valuable to SharePoint developers. Let's just say that I've been working hard on the series and I hope to have the subsequent posts published by next weekend.
In the meantime, I wanted to share some lessons learned while performing an upgrade in the Jameson Datacenter.
A couple of weeks ago, I started receiving notifications (from Systems Center Operations Manager) regarding disk errors on my primary SQL Server computer (BEAST). The situation deteriorated and I soon found that one of the four drives in my RAID 1+0 array was offline. After a short sigh of relief that I hadn't lost any of my data, I added the drive back to the array and initiated a rebuild of the mirror. I also ordered a new drive to replace my backup drive and thus have a spare in case the problematic drive failed again.
When I received the new hard drive, I decided it was probably a good time to rebuild BEAST. As I noted in an earlier post, the server -- at least until a week ago -- was running Windows Server 2003 and SQL Server 2005. Note that BEAST also serves as the "data tier" for my instance of Visual Studio Team Foundation Server 2008. The "application tier" runs on a VM (CYCLOPS) on a different server.
Prior to initiating the rebuild, I found the following article on MSDN:
Note that the article is for TFS 2005, but based on my knowledge of the differences between TFS 2005 and TFS 2008, it seemed reasonable that the steps would be very similar, if not completely identical.
Also note that what I was trying to achieve with the rebuild was certainly "within the spirit" of this KB article, but wasn't strictly inline with what it prescribes. In other words, my goal was to take my TFS instance down while I rebuilt the backend data tier (using the same hardware but with the latest version of Windows Server and SQL Server). After restoring the backend SQL Server databases, I expected to be able to start up my TFS instance again without issue. I quickly discovered that expectation would not be met (at least not without a lot more work than I originally anticipated).
After ensuring that I had full backups of all SQL Server databases (as well as a backup of my BackedUp folder) on BEAST, I proceeded to rebuild the server with Windows Server 2008 R2. Note that I didn't find anything that stated TFS 2008 is supported on Windows Server 2008 R2 -- but then again, I didn't find anything that stated it wasn't supported on R2, either. It's not the kind of risk that I would recommend to a customer, but this is just my home lab after all.
I was less concerned with upgrading the version of SQL Server during the rebuild process, because in my experience the SQL Server team has done a great job in the area of restoring databases from one version to another.
Note that while my TFS instance was originally built with Team Foundation Server 2005, I upgraded this to TFS 2008 a couple of years ago and later applied TFS SP1 as well. This is important to note because in TFS 2008, there is no separate install for the data tier (hence my first deviation from KB 955601). I also encountered an issue when trying to manually process the TfsWarehouse database, but I'm getting ahead of myself, so I'll cover more about that in a moment.
After installing SQL Server 2008 SP1 and restoring my TFS databases, I found that most everything worked as expected in TFS except for the reports.
Note that I didn't attempt to restore the TfsWarehouse Analysis Services (OLAP) database from a backup, but rather rebuilt it using the SetupWarehouse utility, as prescribed in the aforementioned KB article:
This command completed successfully. However, when I attempted to process the TfsWarehouse OLAP database, I encountered an error that stated the service account that I had configured for Analysis Services (TECHTOOLBOX\svc-sql-as) could not login to the TfsWarehouse relational database.
After digging around a little, I discovered that the data source configured by TFS for populating the OLAP database (i.e. TfsWarehouseDataSource) is configured with the following security setting:
After ensuring the TECHTOOLBOX\svc-sql-as service account was added to the TfsWarehouseDataReader role in the TfsWarehouse relational database, I continued getting the error when trying to process the OLAP database. I even attempted to process the TFS data warehouse by invoking it from Warehouse Controller Web service (i.e. http://localhost:8080/warehouse/v1.0/warehousecontroller.asmx) but no luck there either.
I found that if I changed the Impersonation Info setting on TfsWarehouseDataSource to ImpersonateAccount (a.k.a. "Use a specific user name and password") and entered the credentials for the TFS reporting service (TECHTOOLBOX\svc-tfsreports) -- which is automatically configured as a member of the TfsWarehouseDataReader role in the TfsWarehouse relational database -- then the error no longer occurred.
However, I wasn't comfortable with leaving this change in place because it might cause other issues in the future, and it almost certainly puts the TFS installation into the dreaded "unsupported" category.
At that point, I decided to try a different approach.
Rather than simply rebuilding the data tier, I decided to rebuild the entire TFS instance (i.e. data tier and application tier) by following the steps in the following MSDN article:
Well, you can probably imagine what happened...
Yep...same error. Unable to process the OLAP database due to the following error:
Log Name: ApplicationSource: TFS WarehouseEvent ID: 3000Task Category: NoneLevel: ErrorKeywords: ClassicUser: N/AComputer: cyclops.corp.technologytoolbox.comDescription:TF53010: The following error has occurred in a Team Foundation component or extension:Machine: CYCLOPSApplication Domain: /LM/W3SVC/453528946/ROOT/Warehouse-2-129116740741730815Assembly: Microsoft.TeamFoundation.Warehouse, Version=220.127.116.11, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a; v2.0.50727Process Details:Process Name: w3wpProcess Id: 2660Thread Id: 3120Account name: TECHTOOLBOX\svc-tfsDetailed Message: Create OLAP failedException Message: Internal error: The operation terminated unsuccessfully.Internal error: The operation terminated unsuccessfully....OLE DB error: OLE DB or ODBC error: Login failed for user 'TECHTOOLBOX\svc-sql-as'.; 42000.Errors in the high-level relational engine. A connection could not be made to the data source with the DataSourceID of 'TfsWarehouseDataSource', Name of 'TfsWarehouseDataSource'.Errors in the OLAP storage engine: An error occurred while the dimension, with the ID of 'File', Name of 'File' was being processed.Errors in the OLAP storage engine: An error occurred while the 'File Extension' attribute of the 'File' dimension from the 'TfsWarehouse' database was being processed....(type OperationException)Exception Stack Trace: at Microsoft.AnalysisServices.AnalysisServicesClient.SendExecuteAndReadResponse(ImpactDetailCollection impacts, Boolean expectEmptyResults, Boolean throwIfError)at Microsoft.AnalysisServices.AnalysisServicesClient.Process(IMajorObject obj, ProcessType type, Binding source, ErrorConfiguration errorConfig, WriteBackTableCreation writebackOption, ImpactDetailCollection impact)at Microsoft.AnalysisServices.Server.Process(IMajorObject obj, ProcessType processType, Binding source, ErrorConfiguration errorConfig, WriteBackTableCreation writebackOption, XmlaWarningCollection warnings, ImpactDetailCollection impactResult, Boolean analyzeImpactOnly)at Microsoft.AnalysisServices.Server.SendProcess(IMajorObject obj, ProcessType processType, Binding source, ErrorConfiguration errorConfig, WriteBackTableCreation writebackOption, XmlaWarningCollection warnings, ImpactDetailCollection impactResult, Boolean analyzeImpactOnly)at Microsoft.AnalysisServices.ProcessableMajorObject.Process(ProcessType processType, ErrorConfiguration errorConfiguration, XmlaWarningCollection warnings)at Microsoft.AnalysisServices.ProcessableMajorObject.Process(ProcessType processType)at Microsoft.TeamFoundation.Warehouse.OlapCreator.ProcessOlapNoTransaction(Boolean schemaUpdated, UpdateStatusStore updateStatus, Server server, SqlTransaction transaction)at Microsoft.TeamFoundation.Warehouse.OlapCreator.CreateOlap(WarehouseConfig whConf, String accessUser, String dataReaderAccounts, Boolean dropDB, Boolean processCube)
...along with the corresponding error on the data tier:
Log Name: ApplicationSource: MSSQLServerOLAPServiceEvent ID: 3Task Category: (289)Level: ErrorKeywords: ClassicUser: N/AComputer: beast.corp.technologytoolbox.comDescription:Error during OLE DB operation. Error Code = 0xC1060000, External Code = 0x80040E4D: Login failed for user 'TECHTOOLBOX\svc-sql-as'. 42000.
At this point, I suspected a bug due to the fact that I was using Windows Server 2008 R2 on my backend SQL Server, because I had carefully followed each step in the TFS install guide for a dual-server deployment. I then rebuilt BEAST again, but this time used Windows Server 2008 (instead of Windows Server 2008 R2).
Fast forward a couple of hours and -- yep, you guessed it -- same error. Crikey!
However, at this point, I firmly believed that I was in a "supported configuration" and therefore decided to spend some more time troubleshooting the problem. [I felt confident that reverting back to Windows Server 2003 and SQL Server 2005 would also resolve the issue, but I viewed this as taking a step backwards and would use it as a last resort.]
I discovered that just prior to the error in the event log shown above, the following event was logged:
Log Name: ApplicationSource: MSSQLSERVEREvent ID: 18456Task Category: LogonLevel: InformationKeywords: Classic,Audit FailureUser: TECHTOOLBOX\svc-sql-asComputer: beast.corp.technologytoolbox.comDescription:Login failed for user 'TECHTOOLBOX\svc-sql-as'. Reason: Token-based server access validation failed with an infrastructure error. Check for previous errors. [CLIENT: 192.168.0.101]
After researching this a little bit, I suspected that SQL Server was attempting to use Kerberos to impersonate the Analysis Services service account. Following KB 917409, I then registered a Service Principal Name (SPN) for the Analysis Services service on BEAST and ensured the proper settings in Active Directory to enable delegation.
However, I still couldn't get it to work. At that point, I decided to just "punt" the issue and change the service account for Analysis Services to run as NetworkService instead of a domain account. Consequently, I had to add NT AUTHORITY\NETWORK SERVICE to the TfsWarehouseDataReader role in the TfsWarehouse relational database.
At this point, I confirmed that the TFS data warehouse could be successfully updated (using the Warehouse Controller).
You might be wondering why I chose to use a domain account in the first place. It's my understanding that this is a best practice (and I believe it's required when configuring a cluster) and it's the way I've always done it. It was definitely the way BEAST was configured prior to this TFS rebuild. Note that the TFS install guide says to:
type the name of a domain account or NT AUTHORITY\NETWORK SERVICE in Account Name for every service.
Thus I thought that I should be able to use a domain account for Analysis Services.
As a sanity check, I also found the following in SQL Server 2008 books online:
You can choose to run an instance of Microsoft SQL Server Analysis Services in the security context of many different accounts. However, we recommend that you use a domain or local user account as the logon account for Analysis Services.
Continuing the process of validating my new installation of TFS, I created a new test project (based on the MSF Agile template) and confirmed that I could view the reports from the Team Explorer window in Visual Studio. I also confirmed that I could browse to the project team site (i.e. the SharePoint site).
Note that after all this work, I was really only at the beginning of the overall process. More specifically, I had just completed the following step:
1. Install Team Foundation Server in the new environment and make sure it is operational. For detailed instructions...
I continued following the MSDN article and restored my database backups (except for the ReportServer and ReportServerTempDB databases -- but I'll get to that in a moment).
While subsequently attempting to attach the SharePoint content database (for the TFS project team sites), I encountered an error because my TFS server was previously running Windows SharePoint Services v2 but was now running WSS v3. When I installed TFS in my "new environment" I had to first integrate SP1 into the installation of Team Foundation Server (i.e. a prerequisite for supporting SQL Server 2008 and Windows Server 2008). In other words, installing SP1 onto an existing TFS 2008 doesn't upgrade the environment to WSS v3. However, installing TFS 2008 with SP1 integrated will install WSS v3.
In order to upgrade my SharePoint content database, I restored it to a VM running WSS v2, ran the Prescan.exe utility, and subsequently backed up the updated database. I was then able to restore the new backup on BEAST and attach it to the WSS v3 instance running on CYCLOPS.
The last part of the move process was to restore the TFS reports. Keep in mind that when I said earlier that I confirmed that I could view the reports, I was referring to the test project that I created.
However, I was wary of simply overwriting the ReportServer database on BEAST because my backup was from SQL Server 2005 and BEAST now had a freshly rebuilt SQL Server 2008 instance of that database (created by the TFS install). Perhaps it's just me being a little paranoid or some form of "analysis paralysis", but what if the SQL Server team made changes to the Reporting Services schema or stored procedures between SQL Server 2005 and SQL Server 2008?
Instead, what I wanted to do was simply recreate the TFS reports for all of my various projects using the new Reporting Services instance that I had created. Note that I had not created any custom reports that I needed to worry about getting back.
This is when I discovered that TFS 2008 doesn't support the ability to easily recreate the reports for a project. It appears to be a somewhat frequent request and something that has been considered as part of the TFS Power Tools but not something that has been formally addressed by DevDiv at this point (at least not for TFS 2008).
Downloading the process template and subsequently uploading the RDL files appears to be a supported operation, but I have 13 TFS projects and when you multiply this by the number of reports for each project (~20), this quickly becomes quite a burden to perform manually.
Fortunately, I found the following KB article:
Unfortunately, the tool referenced in this KB article (TFSUploadReports.exe) was originally built for upgrading from TFS 2005 Beta 3 to TFS 2005 RTM. Consequently, it references version 18.104.22.168 of the various Microsoft.TeamFoundation assemblies (and my freshly rebuilt TFS server only has version 22.214.171.124 of these assemblies). I first attempted to use assembly binding redirection (to force the old utility to use the new assembly versions). I succeeded in redirecting the three assemblies that I thought would suffice, but then encountered additional errors and therefore quickly gave up on that effort (before attempting to dive into Fusion logging).
Instead I fired up Reflector on TFSUploadReports.exe and observed that it really wasn't doing anything too complex -- essentially just reading the report configuration file and then uploading the RDL files using the SOAP interface for SQL Server Reporting Services.
Putting aside any concerns about copyright infringement or licensing issues for this specific scenario, I copied the code from Reflector into a new console application project in Visual Studio, added a Web Reference to http://cyclops/ReportServer/ReportService2005.asmx?wsdl, and made a few tweaks to get the code to compile (e.g. changing ReportingService references to ReportingService2005). After a little debugging, I discovered that I also had to change the URL of the Web service proxy:
private static void CreateReportServerProxy()
string reportsServerUrl = GetReportsServerUrl();
//m_proxy = new ReportingService();
m_proxy = new ReportingService2005();
//m_proxy.Url = reportsServerUrl;
m_proxy.Url = reportsServerUrl.Replace(
m_proxy.Credentials = m_tfs.Credentials;
This was necessary to avoid the following error:
System.InvalidOperationException: Client found response content type of '', but expected 'text/xml'.The request failed with an empty response.
I was then able to quickly upload the reports for each of my 13 TFS projects (in about 3-4 minutes).
I'm happy (and relieved) to say that my TFS instance is now back up and running and seems to be no worse for wear -- given all that it's been through. In fact, I'm actually glad to have the environment upgraded from Windows Server 2003 and SQL Server 2005 to Windows Server 2008 and SQL Server 2008 -- as well as all the experience that I gained from it.
So, in closing, here are some thoughts to consider: