topic Re: VMware App & CPU Ready in All Apps and Add-ons

VMware App & CPU Ready

chrislymanWMT1 — Tue, 07 Aug 2012 12:40:01 GMT

Hello All,

In the VMware app it reports CPU ready for a vCPU as a summation, but honestly I can't think of why anybody would want that stat as a summation.  Is there a way to pull the CPU ready times as a percentage?  

Also, I'd like to request that the next version of the app have CPU ready displayed as a percentage.  Using CPU ready is a very important metric to use when doing capacity estimation, in my opinion it's more useful than using utilization by MHZ (that's how the current app appears to be functioning).

Re: VMware App & CPU Ready

tfletcher_splun — Wed, 30 Jan 2013 23:38:17 GMT

So in the raw data of the splunk for vmware app the collection period is 20s. If you want to see the data as a percentage you can just change a search to provide you that data.

For example:

index=vmware source=VirtualMachinePerf SumRdy_ms=* | eval SumRdy_pct=SumRdy_ms/20000 | stats avg(SumRdy_pct) as AvgRdy_pct first(moname) as name by meid

Re: VMware App & CPU Ready

colinj — Thu, 20 Jun 2013 17:28:32 GMT

The trouble here is with VMware and not Splunk. VMware stores the ready time as a summation across all vCPUs in each VM.

Re: VMware App & CPU Ready

colinj — Wed, 03 Jul 2013 15:17:07 GMT

CPU ready time was just about the hardest thing for me to get my head around when I started working with vSphere and I'm still not always sure that I get it completely. The hardest part in understanding ready time is remembering that it is a summation of the ready times for all the vCPUs the VM has. However, when VMware talks about when you should worry they talk about ready times on a per vCPU basis. This means that you have to factor in how many vCPUs a machine has.

VMware says that 1000ms/vCPU should be the warning point and 2000ms/vCPU should be the critical point. Since CPU ready times are collected over a 20 second period (20000ms) these work out to 5% and 10% respectively. The search below is what I use to look at CPU ready times. For any given VM it will look up the VM in the inventory that Splunk generates so that it can divide the SumRdy_ms by the number of CPUs (numCpu). From that I can then get the max values and the average values for my normalized Sum Ready Time.

index=vmware source="VirtualMachinePerf" SumRdy_ms="*" perftype="cpu" moname="vmname"
| lookup TimeHierarchyVMSummary moname
| eval SumRdyPerCpu_ms=(SumRdy_ms / numCpu) 
| timechart max(SumRdy_ms) AS "Sum Ready Time" 
max(SumRdyPerCpu_ms) AS "Max Normalized Ready Time" 
avg(SumRdyPerCpu_ms) AS "Average Normalized Ready Time"