Hi We have 2 cluster,
Cluster 1 have hosts with Intel(R) Xeon(R) CPU E5-2600 V3,v4 CPUs , Intel 2600 series and Intel gold 5100 series.
Cluster 2 Have hosts with Intel gold 5200 Series
Cluster 1 EVC disabled & Cluster 2 have EVC enabled with casecade lake.
Can we do share nothing vmotion from Cluster 1 to Cluster 2 without downtime.
if not why? if yes then how?
Hey, hope you are doing fine:
I would take the following into consideration:
EVC requires the CPUs to be the same brand (intel) and family.
I wouldn't mix CPU models within the same cluster.
Regarding the movement of VMs, as Andre pointed, as long as the VM is powered on with a lower version of EVC you will be able to move it (this is because the CPU microcode and the CPU functions the VM is using at the moment might not be available)
If this vMotion is for a cluster migration/hardware refresh I would take the downtime, so I'd ensure that I don't loose the new CPU features on the Intel Gold 5200.
If this vMotion would be a regular operation you can lower the EVC level of the cluster, but you will loose CPU features.
Also you can enable per-vm overrides
I could be wrong here, but as far as I know, EVC will need to be enabled on both clusters, even when going from lower to higher family CPU. I have not done this in a while, so again, I could be wrong.
Hey Duncan, how is it going mate?
I'm not a 100% sure either, but i'm pretty sure that if you are trying to go from an older version of CPU let's say (and this is out of my mind) a Broadwell CPU on the source and you set the Destination to Broadwell both will have the same microcode and there should be no issues. You will loose some newer features of the CPU though.
All is well Nacho 🙂
Indeed, but the source doesn't have EVC enabled in this case. Unfortunately I don't have different families in the lab, so I can't test it.
<edit> just talking to Niels who is the Tech Marketing person for it, he says it should work</edit>
The CPU families I mentioned were merely examples to prove my point.
Enabling EVC will require you to reboot the VMs that are part of the cluster (since you are changing the CPU microcode)
If you mgirate the VM from a host with Broadwell to a host with cascade lake you will need to do it with a cold migration (VM powered off)
That's correct, the vCenter won't let you migrate because of the multiple failsafes that there are in place.
If that failsafes didn't exist if you could migrate from broadwell to skylake what would happen is that the VMs will fail with a BSOD.
Picture this, what happens if you hot swap a computer's CPU? Basically this is what you would be doing. If you would vMotion from Broadwell CPU to a Cascade Lake CPU
But though destination cluster has EVC of casecade lake but still we want to run VMs at broadwell during migration to destination cluster then reboot those VMs once migration completed at destination cluster.?
Is it feasible?