We have an issue with Windows XP SP3 VMs on VMWare ESXi 4.1.0 (the free version): They sometimes seem to "pause" for several minutes. This happens rarely (maybe once a week per VM, we don't always notice), but still is an issue for us. It happens for three different but similar VMs on three different hosts with pretty different hardware.
I have the feeling that the "pausing" is not actually the CPU blocking, but probably the harddisks, but not 100% sure. The servers have one IDE disk (C:) and one SCSI (D:) and it might be either of the two. I have seen scheduled tasks not starting for up to 9 minutes and then running normally again with normal speed. This is not a load issue, the VMWare hosts have mostly average load and the VMs in question already have reserved CPU resources plus high priorities for CPU and disk. I did get a chance to see this happen live and the VMWare host load was actually below average at that time.
The Windows boxes run mainly MySQL, Tomcat, Cygwin stuff, Java + R applications, VMWare tools, Elusiva Terminal Server pro, FileZilla FTPS server, Nagios client. Not sure if the issue might be caused by any of that software (e.g. Elusiva).
Trying to debug this, there was nothing visible in Windows Event log, other logs in C:\Windows, VMWare events etc. Unfortunately the vmware.log file on the host ended with "Log throttled" for all three instances weeks ago. We found that we ran into 2 known and documented VMWare bugs there: The VMWare tools write lots on bogus messages in the vmware.log causing the throttling VMWare KB, plus VMWare never unthrottles the log (at least so far despite VM reboots). We now applied the fix for the first bug (VMWare tools log level error setting), but so far no unthrottling despite VM reboots.
I know this is not much to pin down the issue and that may also be the reason why I so far didn't find anything related on the web or on ServerFault, but maybe some of this rings a bell with someone? Or please direct me to what more info to post. I hope that the vmware.logs get unthrottled eventually (can't easily restart the hosts at the moment, which may help).
Thanks a lot for any input!