On May 17, 2021, at 2:03 AM, Ryan Schmidt <[email protected]> wrote: > On May 16, 2021, at 17:57, Daniel J. Luke wrote: >> On May 16, 2021, at 10:48 AM, Christopher Nielsen wrote: >>> I’d bet the hypervisor is spending more time on scheduling and pre-emption, >>> than actual processing time. >> >> This is something we could actually measure, though, right? Then we don't >> have to just speculate (and if we do determine that a config change needs to >> happen, we can use the actual measurements to help us optimize the >> configuration). > > What would be your suggested methodology?
I'm not an ESXi expert but after a quick look through some of their documentation it looks like there's an `esxtop` command that can show some information. More info here: https://kb.vmware.com/s/article/1005362 (some google searches seem to indicate that when the oversubscription is a problem, it's usually because ESXi is waiting for 'enough' cores to be available to start all of the vCPUs - and that this used to be much worse in older ESXi versions, but can still be a problem). This post also has information that looks useful: https://www.heroix.com/blog/vmware-vcpu-over-allocation/ -- Daniel J. Luke
