This post is more than 5 years old
15 Posts
0
174381
February 12th, 2014 16:00
CPU Ready & Capacity Availability Reporting
I have a question regarding CPU Ready and how it impacts Capacity Availability Reporting.
For example, in one of my clusters, there are a few over-sized VMs that get CPU ready from time to time.
The Summary Report from Foglight Standard shows that the CPU Ready Values for the cluster are as below:
I am wondering what this refers to – is it the maximum\min\average\latest CPU ready for individual VMs or across the cluster?
This makes forecasting capacity for new VMs tricky.
Due to CPU Ready being seen as a constraint by Foglight the forecast says there is no capacity in the cluster for new VMs. Yet on a per host level it suggests there is room for many.
I know the cluster level takes into account HA but is the cluster level forecast reporting zero because of the CPU ready status of a couple of VMs? (note there are circa 1000 VMs in the cluster)
My question is a little rambling, apologies, if you have a cluster of 1000 VMs and 10 of those experience CPU Ready – does this mean Foglight will automatically not suggest adding more VMs to that cluster?
bakins1
5 Posts
0
June 16th, 2014 10:00
CPU Ready on a VM in a cluster will limit the capacity availability to 0 in FVS.
CPU ready is a measure of time a virtual machine is waiting to run. It can occur any time a VM waits for an action to complete or the hypervisor takes control away from the VM.
CPU Ready can occur for a variety of reasons. It can occur if the Host is overly utilized, and there are never any CPU resources available for the VM. Another possible cause would be how that VM relates to the other VMs on the Host and the demand for resources of those VMs. VMware has a process that decides which VM gets the resources, and there are many variables in that calculation. This makes it very difficult (near impossible) to accurately predict how many more resources are needed to eliminate an alert for CPU Ready on a specific VM.
So, if we detect a CPU Ready alert (Peak or otherwise) , we set the additional VMs to be ‘0’ for the Host. The thought being that if there is a VM that is struggling to get resources ( alerting CPU Ready ) then that has to be addressed before we can say we can add an additional load ( more VMs) to a Host.
All of the virtual machines on the host will be affected and all will have performance problems. If DRS is enabled, then the problem will typically extend to all of the hosts in the cluster.
shaneirl
15 Posts
0
April 10th, 2014 08:00
... anyone?
shaneirl
15 Posts
0
April 30th, 2014 09:00
...
shaneirl
15 Posts
0
July 18th, 2014 18:00
Many Thanks for your reply!