Disabling EVC Mode while the Two ESXi Hosts with Live VMs, is this safe?

[VM_JOf]

Hello Guys,

Would like to know if there's any implication disabling the EVC Mode in the cluster while the two Host and 20VMs in the said Cluster are Live?

I wanted to make sure if this is safe to disable the EVC mode as I need to add another Host (3rd host) in to the Cluster (Identical specs.)

Or Can I just simply leave it disabled this EVC Mode?

Thanks in advance,

VM_JOf

[a_p_]

Enabling, or disabling EVC will not affect powered on VMs, since the CPU features are only exposed to the VMs at power on.

Anyway, there's no need to disable EVC in order to add another host (it actually doesn't make sense). If the new host - as well as powered on VMs on this host - are compatible with the cluster's EVC level, you will be able to add the host to the cluster.

André

[VM_JOf]

Hello Andrei,

I have ask this question because, I got some issue in my 3rd Host. I have 3 Identical ESXi Host which for some reason the third Host was kicked-out from the cluster (disconnected) and I’m thinking if I can disable the EVC Mode in the original cluster first as I cannot add-back the 3rd host to its original cluster where the two hosts are connected.

I wanted to avoid downtime as 100s of VMs are my concern and I dont have other host resource to cater.

JOf

[a_p_]

I have 3 Identical ESXi Host ...

Can you confirm that they are completely identical, including BIOS, settings, and CPU-Microcode?

Which version/build of vCenter, and ESXi are you using?

You will certainly be able to add the 3rd host to the cluster after disabling EVC. However, if something doesn't match, you may not be able to enable EVC for the cluster again, which may not be what you want.

What you may try is to create a new, temporary cluster with only the 3rd host. Then check whether EVC can be enabled with the required level. If this works, disable EVC again, and try to move the host to the original cluster.

André

[VM_JOf]

Hello Andrei,

Thanks for the reply. I created a Temp Cluster adding the 3rd Host to it and enabled the EVC Mode which is same as the original cluster and its work well.

So I’m thinking if will work by Disabling the EVC in the Original Cluster for the two host and add back the 3rd Host. Then turn it back the EVC Mode.

But I saw in some article, after disabling the EVC Mode, VMs resides into the cluster have to restart to take effect the changes made? Is that true? If not then I will give it a try disabling EVC Mode and add back the 3rd Host.

-JOf

[a_p_]

Do you see an error/incompatibility warning when you try to add the 3rd host to the original cluster without disabling EVC?

André

[VM_JOf]

Yes, I do have the CPU Compatibility Check Message. It’s quite weird that its been working for years and just sudden that the 3rd Host was disconnected from the cluster. No recent patch applied on the hosts so I dont know yet what causing this sudden disconnection and I cannot put it back to the cluster it belongs.

JOf

[a_p_]

What exactly is it complaining about?

Again my questions:

Can you confirm that they are completely identical, including BIOS, settings, and CPU-Microcode?

Which version/build of vCenter, and ESXi are you using?

André

[VM_JOf]

Hello Andrei,

They are all Identical from CPU, BIOS and all Host are set to Westmere EVC Mode.

ESXi Hosts Hypervisor version, Server Model and CPU

BIOS wise is all identical into: 1.18

Is there any required ESXi Patch for this version I have? It might be OS issue and not a hardware issue.

Regards,

JOf

[VM_JOf]

vCenter version/build is:

Regards,

JOf

[a_p_]

Can you also share the exact error/incompatibility message, when you try to add the host to the cluster?

André

[VM_JOf]

This is the message always prompted every time I perform add-back to the original cluster.

Already check by the HW Vendor and all required features are set accordingly. Its just weird that this host was Disconnected unexpectedly from the Cluster. Is there any known Bug issue on my current ESXi Host Version/Build?

[a_p_]

On one side vCenter is aware of the additional Intel-Microcode features, which came for Spectre/Meltdown, but your BIOS/Microcode is not yet updated.

However, on the other side, all hosts are running the same BIOS/Microcode, so it's not really clear whether this is causing the issue.

Can you please try to create a new cluster, move the host to this cluster, and then try to enable EVC? Does this work?

André

[VM_JOf]

If you are pertaining to the disconnected Host. Yes, I created a new cluster add the disconnected Host into it and enabled the EVC, it work without any problem.

but if you are talking about create a new cluster add the disconnected host plus the two working hosts, this haven’t tried as I cannot afford to have a downtime right now.

Regards,

VM_JOf

[VM_JOf]

Seems this been a known issue before

Newer IBM servers not vMotion compatible despite identical CPUs

Regards,

JOf

[a_p_]

If you are pertaining to the disconnected Host. ...

Yes, that's what I was thinking of. Just to see whether EVC can be enabled if this host is the only one in the cluster.

Regarding the link you've posted. There's aslo something mentioned in https://kb.vmware.com/s/article/1003212​ that may be relevant. However, I'm not familiar with IBM hardware, so this is more of an information.

For Intel "Westmere" EVC Mode, some server OEMs might require you to enable the AES instructions in the BIOS.

André

[VM_JOf · Accepted solution]

Hello Andrei,

Wanted to share with you on the resolution.

1. From the original cluster, I have disabled the EVC Mode and HA then enabled it back.

2. Move the affected host back to original cluster and it was successful.

3. Test a vMotion and it works perfectly. fine.

After all, it was conclude it was not a Hardware Issue, but its a software issue (VMware).

Hope the others who had this same scenario could help them as well.

JOf