Ok, so this is not a real blog post – I’m just testing to see that I can get Word 2007 to post to my blog…..
If you’re reading this, then it worked!! J
So after much wailing and gnashing of teeth, our TFS setup seems to be pretty stable now. We have a real server on the way for it, and we’re waiting till thats up and running before we go full on with migrating from VSS and Clearcase to TFS.
But for now we’re still using TFS from a Virtual Machine. Through much pain I think I’ve figured out the best way to set up the VM so that it can survive the worst that crappy hardware and dodgy server admins can throw at it…So far it’s survived two unscheduled reboots without needing to be rebuilt, so I’m pretty happy with it…
Here’s a brief rundown on how I’ve got it configured:
TFS is installed as a single server instance, in a Virtual Server 2005 VM. Our VM is using 1GB of RAM, and Virtual Server reserves up to 100% of one CPU (The host server is a 4-way Xeon box, so we have 8 logical CPUs in total).
I’ve got two Virtual Hard disk drives, set up to use a virtual SCSI controller, rather than as IDE drives (this seems to be the more stable way, and is faster). The data drive is set up as a fixed size drive as well (20GB), which again improves resilience and performance.
We’ve got SQL Server backups running for the TFS, Reporting Services and Sharepoint Services databases, with the files stored directly on a share on the Host machine, rather than having them saved to a folder on the VM itself.
The SQL backups run as follows: Full DB backups at 1am every Saturday; Differential backups at 1am every other day; Transaction log backups run every 4 hours, starting at 3am every day. I’ve scheduled these like this to avoid contention with a couple of built in jobs – when they ran in tadem with the other jobs, the backups appeared to finish successfully, but when I tried to restore from them, SQL Server would report that the database “was corrupt when the backup was made” for the databases where the TFS admin jobs were running – I’ve not seen this happen since I set the SQL backups to run at a different time to the TFS jobs.
Finally, I also have a backup clean-up job that runs at 2am every Sunday, to keep the number of backup files, and the space they take up, down to a minimum. I keep 4 weeks worth of backups online at any one time.
