EnterpriseTeam
Contributor
Contributor

Max MV Memory Contention

Jump to solution

For some reason, in my single vRealize instance monitoring 1 cluster of 6 hosts running 5,5 u3b, I never see anything but zero for the memory contention counter, which translates into dashboards that are probably somewhat deficient when generating an overview of cluster health. Am I being daft and missing something here, or is this metric simply not being collected for some reason? Other metrics seem perfectly realistic, e.g. CPU and Disk Latency (see attached screenshots).Summery.jpgCluster_Performance.jpg

0 Kudos
1 Solution

Accepted Solutions
sxnxr
Commander
Commander

Looking at both screenshots i would say you dont have any memory contention that is why it is reporting at 0. I have clusters with 200 VM on them with 0 memory contention and other with 50 VMs and 20% contention ( this is by design for a lab environment)

Now looking at the dashboards you should look at the vCPUs assigned to the VMs and look at stripping back the number allocated to Vms as the CPU contention is high. If you look at the demand compared to usage i would guess you will see demand is higher than usage which is not great. Also the dashboard is showing max vm cpu contention as with any max value is worst case.

Below is an example of contention causing a vm to run slow because the vm needs more (demand) that what it is getting (Usage). In my case this is being caused because we have to many 8 vcpu vms on the host. The hosts are 2 x 8 core so anything over 2 x 8vcpu will cause me contention.

cpucon.png

View solution in original post

0 Kudos
2 Replies
sxnxr
Commander
Commander

Looking at both screenshots i would say you dont have any memory contention that is why it is reporting at 0. I have clusters with 200 VM on them with 0 memory contention and other with 50 VMs and 20% contention ( this is by design for a lab environment)

Now looking at the dashboards you should look at the vCPUs assigned to the VMs and look at stripping back the number allocated to Vms as the CPU contention is high. If you look at the demand compared to usage i would guess you will see demand is higher than usage which is not great. Also the dashboard is showing max vm cpu contention as with any max value is worst case.

Below is an example of contention causing a vm to run slow because the vm needs more (demand) that what it is getting (Usage). In my case this is being caused because we have to many 8 vcpu vms on the host. The hosts are 2 x 8 core so anything over 2 x 8vcpu will cause me contention.

cpucon.png

0 Kudos
EnterpriseTeam
Contributor
Contributor

Quite surprised I don't have any memory contention as I would have expected this to show up (and other external tools seem to be suggesting such as well). Many thanks for the assigned vCPU advice as well. Hopefully with this information I can present a case for considering expanding the number of hosts in the cluster.

0 Kudos