thanks for the link.
according to the link demand and entitlement metrics are max by provisioned metric.
I tried to sum up all CPU provisioned to all VMs on the cluster plus overhead and I have something around 137 Ghz which is lower than demand (159.19 Ghz) in the above picture.
so how can demand be higher than total provisioned CPU ?
I am really confused
Because a VM can demand more than it has
Think of it like this. And these are in my mind the 3 most important metrics for cpu performance
Usage = The amount of CPU the VM is using (vrops counter includes the VM overhead)
Demand = The amount of CPU the VM is demanding
Contention is the difference between the two
you can have a VM using 20% CPU but demanding 70% so you contention is 50%
In this case the VM may be running slow and have performance problems ( depends on the users and what the vm is) adding more CPU wont improve the performance but just add to the bad performance because out of all the cpus assigned to the VM it is only demanding 70% so you have two may cpus assigned to your VMs or to many CPUs on the cluster
On the other side of that you can have a VM with 100% cpu usage and a demand of 120% This would indicate that the VM does not have enough cpus assigned to it.
In your case all your VMs are demanding more that you physical have. This could be down to one VM miss performing or just to many VMs on the cluster. Again i would check the VM that is in red in the bottom line
Ok I will check the red VM and reply.
thanks for all these explanations.
I checked the virtual machine in red and found that it had 107% CPU workload.
According to above advices I think this virtual machine needs more CPU.
I also noticed DRS is on with some affinity and anti affinity rules on the cluster.
can this last fact also be the root cause ??? because total demand is less than CPU capacity of the cluster.