I have the exact same behavior, with the same load, my vm's CPU usage on the host increase all the time. After two month it is now 5x the normal amount. Shutting down the vms and restarting them bring back a "normal" CPU usage. Restarting the guest OS is not enough, the vmware guest process needs to be stopped and a new process created.

There is defenetively something very wrong here.

RHEL4 host with Linux and Windows server 32bits guests (1 CPU per guest).

I have exactly the same problem. I graph my host's CPU usage with Cacti, and here are my results:

Has anyone found a solution, or evern a workaround to this problem? I've just bought a replacement CPU for the host server - faster clock speed and Intel-VT is supported, so I am planning to migrate all VMs to 64-bit, to see if that has any effect.

Still seeing this here as well. Here are our Nagios graphs for CPU usage. (Where it goes back to zero are when we have rebooted the machine - the missing data in Sept. is when we tried disabling Nagios)

Note that the number of VMs has not increased, nor has the load placed on the individual VMs.

Yep.  Same problem here.  When I was on 1.x with the same number of VMs and the same server my host stayed up for months (almost a year!) with no issues. Now with server 2.x it looks like I may have to hard reboot  (power on/off) the VMs every 2-4 weeks as my CPU usage starts climbing dramatically.  This is the only fix that works.  The funny (or frustrating thing) is that the VMs themselves CPU usage will report 0-5% or so, but the vmx process on the server reports 25%.  This seems to indicate a problem with vmware server itself.  Sounds like a big bug and reduces my confidence in server. And I used to LOVE vmware server pre 2.x days  

BTW my layout is 64bit RHEL 5.x on the host, 16GB RAM, 8 VMs with mix of 32/64bit Windows and Linux.  Both my production and development server show the same symptoms and they are different hardware.

One of my graphs (using Zabbix - this is user %CPU over time):

Please let us know your results.  It would interesting if the problem was a result of mixing 23/64 bit guests - not a whole lot of help for those of us that have to run 32 guests, but at least we have a possible culprit.

To vmware :  Is there any official method of reporting bugs to vmware besides hoping they see this in the forums?  I'll gladly provide a detailed report if it helps get this fixed.

It doesn't seem like this problem is isolated enough to warrant having to pay to fix it.  It seems like everyone's problem.  Are you not having this problem, LucasAlbers?

HelgeAlg, I don't think that issue is related (or at least we do not see it here). Memory usage remains consistent for us at 99% almost all the time.

CPU on the other hand continues to grow at a constant rate.

You can see at the start of week 41 we stop one VM, another at the end of week 41 and yet another at the start of week 42. And yet, growth continues at a rate of around 10% a week.

This is with no load change on the individual VMs at all!