I have nasty problem that has occurred basically overnight.
I have a 18 server Citrix farm running on ESXi 3.5 cluster. The 10 host machines are all the same Dell R610 machines with 2 x 6 core processors in each. ram is 32GB per host.
The problem occured last night after modifying each of the Citrix guests to use 2 vCpu's instead of 1. Inturrupt times have gone through the roof, during daily usage the single vcpu inturrupt would rarely venture above 1-2% im now currently looking at the server perfmon stats which are showing % Interrupt Time as fluctuating between 0 and 44%.
Attached is an image from the Citrix management console showing the massive increase in interrupt.
Having performed some research across the googlesphere i came across this post http://communities.vmware.com/thread/237044?start=0&tstart=0
Following the advise given i have disabled any host power management functions and also disabled the "C1E" capability of the cpu. None of these changes appear to have had any effect on the guest systems cpu.
esxtop on the host machines doesn't appear to indicate any issues the %USED and %RUN values seem to match up ok and general system cpu usage is low. It seems the problem lies with the guest OS accessing the new vcpu's.
The guest OS is Windows 2003 fully service packed and updated.
I can only guess that the interrups will cause the system to underperform as the guest os waits for cpu availabliity.
Any help is greatly appreciated. Im scratching my head on this one.
Thanks
Hi mate, have you got any progress on this?
I'm having lots of issues with a citrix farm runing on ESXi as well, interruptions sometimes goes up to 70%, I'm quite frustrated
Nope gotten no further with this. The Citrix farm seems to be in good health except for the crazy high interrupt spikes.
I'm stuck for ideas on it really.
There must be someone out there running this setup successfully.
My servers are sort of working better now, but sometimes under heavy load high interrupt kick in going nearly to 80% of the cpu time but it is not happening very often
Does the Windows 2003 OS have the multiprocessor HAL?
So far I know it does
Please check out this KB - http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100371...
This doesn't look to apply for my case as the guest that is having interrupts inside the VM, at the host level I'm not seeing any problem. Has this applied to your case?
The workaround had resolved the issue at host level for us earlier.
http://support.microsoft.com/kb/309283
Check this