Hi All
I am currently having issue with 8+ of our esxi hosts in relation to CPU Ready Time.
I know this is because a lot of the VMs have had vCPUs over subscribed and I am getting a plan together to reduce the vCPUs across the board which will hopefully reduce the ready times.
There is something I found though that concerns me though, our monitoring software (Proactivenet) throws out an error message everytime the CPU Ready Time goes above 5% on any of the VMs. I have been collating this data over the past week, I thought I would see that the hosts with the most vCPUs assigned within them would have the most alerts, however this is not the case:
| HOST | Overall Errors | CPU Ratio |
| HOST1 | 0 | 0.75 |
| HOST2 | 0 | 1.41 |
| HOST3 | 4 | 1.91 |
| HOST4 | 0 | 2.16 |
| HOST5 | 0 | 2.25 |
| HOST6 | 40 | 2.33 |
| HOST7 | 40 | 2.58 |
| HOST9 | 56 | 2.75 |
| HOST10 | 56 | 2.83 |
| HOST11 | 0 | 2.91 |
| HOST12 | 35 | 3 |
| HOST13 | 107 | 3.33 |
| HOST14 | 146 | 3.33 |
| HOST15 | 0 | 3.5 |
| HOST16 | 17 | 3.58 |
| HOST17 | 2 | 4.08 |
HOST17 has four vCPUs to every 1 physical CPU yet it shows next to no errors, the same for HOST15. Yet host 4 with less vCPUs assigned overall is having a major issue in relation to CPU Ready Time.
I wondered if it was the monitoring, but it checks out fine, I have also taken into consideration vMotion when collating the information. All host servers are the same spec, the only difference between host13 and 15 is the fact it has one VM more but yet would that matter with less vCPUs being used overall anyway.
This doesn't affect the fact that we will still reduce the reduction of vCPUs across the board. But it niggles at me as why some servers seemingly have an issue and some that should seem to have little at all.
Thanks
Matt