Can anyone help me out. I have noticed recently that a host's CPU looks a little crazy. Similar to the attached "image1.png". I started moving every vm off of the host until I had one SharePoint WFE vm left which made the graph look like "image2.png" I have 2 sharepoint vm's that cause my Host to act this way. If I migrate all vm's off the host leaving either one of those SharePoint vm's the story is the same. However if i move the SharePoint boxes off of the host the CPU performance chart goes back to a more normal view like "image3.png" The environment have 6 host with 2 processor Sockets and 12 cores per socket. Hyperthreading is enabled. Most of the vm's have 2 vCPU's and 4 gb ram. However the SharePoint boxes are running 4 vCPUs and 16 gb RAM. "sharepointcpu.png" is an example of the cpu usage on the SharePoint vm. My question is this something a web front end would normally cause? Has anyone else seen this type of behavior? If so is it something to be concerned about? I also attache an esxtop output from the host.
Hi there,
everything seems all right in esxtop according to this post: Why is %WAIT so high in esxtop?
Can you provide us with more data from your OS? Is the CPU spiking in the OS as well or is it idle? This would be the main point to start from as we have only seen the hypervisor level so far. vCPUs traverse around Physical CPUs on the same NUMA Node where they can be best scheduled so maybe it's why you see so many spikes. Also I see that the VM is not that much loaded - can you try decreasing vCPUs to let's say 2 and repot back?
Check to make sure Sharepoint isn't "Crawling" your network to discover other sites.
I just checked and SharePoint is not Crawling
Hi there,
everything seems all right in esxtop according to this post: Why is %WAIT so high in esxtop?
Can you provide us with more data from your OS? Is the CPU spiking in the OS as well or is it idle? This would be the main point to start from as we have only seen the hypervisor level so far. vCPUs traverse around Physical CPUs on the same NUMA Node where they can be best scheduled so maybe it's why you see so many spikes. Also I see that the VM is not that much loaded - can you try decreasing vCPUs to let's say 2 and repot back?
OS CPU is pretty idle. I dropped the vCPU from 4 to 2 and it has seemed to calm the Host down. Thanks for your suggestion it seems like it has fixed it
