We recently had a client inform us that a Windows Service we had written would occasionally appear to be running, but not actually be doing anything. Once the Windows Service was restarted, it would begin processing again. The service would get into this state after it had been running for awhile, but it was never the same amount of time. Sometimes it would run for days without issue, but other times the problem would appear after only 8 hours of operation. We had not made any changes to the Windows Service in a long time so we were a little baffled. We attempted to reproduce the error, but were not successful.
After much investigation it was discovered that the client had moved the Windows Service to a VMware Virtual Machine. We had the client move the service back to a physical machine and the problem went away. At that point we knew something in our service did not agree with VMware, but our service was written using Visual Basic 2005 and was not doing anything out of the ordinary (e.g. no calls to unmanaged code, no third-party libraries, etc.).
We finally narrowed the problem down to a System.Timers.Timer we were using to determine when the Windows Service should start processing again once it had completed its work and had gone to sleep. Apparently the System.Timers.Timer would stop firing at some point, which prevented our application from waking up. The timer would only stop firing if the service was run on a VMware Virtual Machine.
We were able to replace the System.Timers.Timer with a background worker thread (specifically a System.ComponentModel.BackgroundWorker control) that used a call to System.Threading.Sleep. This allowed our Windows Service to work on both physical machines and VMware Virtual Machines.
Some Lessons Learned
- .NET Timers and other time keeping techniques may not function properly in a VMware Virtual Machine.
- Additional testing needs to be performed to verify an application will work properly in a VMware Virtual Machine.
- The System.Threading.Sleep method does not appear to be affected by the issues that caused the System.Timers.Timer to stop firing.
VMware has an excellent article called "Timekeeping in VMware Virtual Machines", which can be found at http://www.vmware.com/resources/techresources/238. This article contains information that might explain why our .NET timer stopped firing.
Another valuable resource for .NET timers is an article written by Alex Calvo called "Comparing the Timer Classes in the .NET Framework Class Library", which can be found at http://msdn.microsoft.com/msdnmag/issues/ 04/02/TimersinNET/default.aspx.