I am scratching my head with this one
We migrated around 200 vms from one cluster to another over the weekend and i noticed the max observed vm contention in vrops for the new cluster was at 28%. Looking into it all the VMs on one host in the cluster jumped from a contention of less than 0.5 % to anything from 6 - 30% after the migration
The strange thing that i cant explain is the total number of vcpus on all of the VMs on this one host is 24. The host has 2 x 16 core processors set to high performance. I cant understand what these VMs are showing as having high contention. There is nothing special about the VMs with limits or resource groups.
The source cluster is the same as the target the only difference is the new hosts are ACI connected.
I move a VM of the host to another and the contention drops. Move it back and it goes up again
Anyone have any ideas.
Oh forgot to say the max vcpu any of the VMs on that host is 4
Could you please review this article and ensure that you have correct power management settings both at BIOS and Hypervisor level - vXpress: Performance over Power : Make the right choice.
the bios is set to OS controlled and the hosts are set to high performance. This is the same across all 10 hosts in the cluster and if i move a vm of this one host the contention drops. when i move it back it jumps up again
So you see this behavior on one single node of that cluster.. Rest all dont show the contention?
Can you also check the CPU Latency counter in vCenter to see what it is reporting for this VM when it is running on that host.
Also, are these VMs running high Network Throughput or Storage I/O. When I say high, I mean very high 🙂