Peculiar problem involving Windows 2003, VMWare Server, SQL Server 2005 and networking

Over the weekend I’ve built my new server – yes, the content you are reading right now is served from it. Perhaps more about this new server in another post. Back to the point. Host OS on the server is Windows 2003 R2 x32 and there is also a SQL Server 2005 running there. I’ve had x64 bit before but it is just too much trouble running it since the drivers and support situation. So, if you don’t need more than 4GB of RAM then I don’t see a compelling reason to go with 64 bits. Why I say “host OS”? Because I am running VMWare Server on top of it. And there are two guest OS running inside virtual machines: Another Windows 2003 that serves web content (it uses SQL Server located on host) and Windows Home Server that takes care of backups. So, after I’ve installed everything I fired up my web virtual machine and take a took a look at my blog – it was a no go. Virtual machine was working fine, just the Community Server wasn’t running. After turning off custom exception handling I ended with a asp.net exception reporting page which I was looking for. However the error was an odd one. It stated that connection with SQL Server (running on host) has timed out. Hm. I investigated further by creating a test application that reads a table from database. Running on my workstation it read the data just fine. But when running from within web server it read just first n rows (i.e. 20) and then it timed out, always at the same row – which was really puzzling. The same symptom appeared to any SQL Server client running within guest OS. So it was obviously a problem related to VMWare Server. Yet, if I’ve turned off Windows Firewall on the host my application worked even on guest OS – this fact deceived me to thinking that the problem is firewall related (perhaps it was in a way) – after half an hour testing with any possible firewall configuration I gave up. Since I knew it has something to do with VMWare Server I then started searching their forums. And soon enough I’ve found a solution (at the bottom of the thread):

Disable TCP Offload on the host

While the solution talks about disabling TCP Offload Engine on Windows, I’ve disabled TCP Offload LargeSend (it sounded enough similar to me and a good candidate for my problem) and it worked like a charm. This is how it looks on my computer:

image

Perhaps networking now consumes 0.00000000001% more of my CPU but at least it works fine. I am not sure whether this is a bug or not, I’ll contact VMWare anyway. Funny, the building and installing my server took less than troubleshooting this problem.

Leave a Reply