VMware Cloud Community
D4veB
Contributor
Contributor
Jump to solution

EVC Level after Migration from ESX4.0 to ESXi5.0

Hello!

I have a problem (niggle) with the EVC level of my new ESXi 5.0 Cluster under vCentre 5. The EVC level is set to "Sandy Bridge" and the four hosts in the cluster are ProLiant DL380p Gen8 servers running E5-2650 16 core cpus. I think its key to point out that the ESXi 5 platform including vCentre server is all new hardware.

Our old cluster was ESX4.0 running on "two"  DL380 G5 servers with a vCentre 4 Server (Hardware)

Using the "Add Host" option the OLD 4.0 hosts were imported into the vCentre 5 enviroment without problem and vmotione worked like a dream moving all servers from old (4) to new (5).

Its key to note that both VMware Infrastructures are using the same backend datastores (SANs).

At this stage, once vmotioned, the server still had their old EVC levels. Default behaviour and all looking well.

So we power cycled the VM's. And then powered them back on.

The EVC level changed ... to "Westmere"?

I have tried everything but I cannot get any of the vmotioned servers to hit the "Sandy Bridge" EVC level?

If I build a new server either from template or from scratch and start it up its starts with an EVC level of "Sandy Bridge" but the vmotioned servers from the now decomissioned ESX4 infrastructure only go as high as "Westmere".

I had a good look around on the forums etc to no avail?

I appreciate that the EVC level is used only for "Enhanced Vmotion" and performance doesnt differ a great deal unless you use encryption but we do use encryption.

EVCLevel.jpg

Any help or ideas much appreciated Smiley Happy

Reply
0 Kudos
1 Solution

Accepted Solutions
jrmunday
Commander
Commander
Jump to solution

The EVC mode is set on power up, so I suspect there may be some legacy entries in the VMX files which are affecting this. Can you upgrade the virtual hardware to the latest version supported by your hosts and test this again (including resetting the CPUID mask).

If this also fails, or you are already on the correct hardware version can you please post the VMX file of an affected VM and the VMX file of a newly created VM (with the same virtual hardware version).

Another option to test if the VMX file is the cause, is to unregister the VM, create a new VM but attach the existing disks - power this up and check the EVC mode. But it might be simpler to upgrade the virtual hardware, reset the CPUID masks, remove the offending entries in the VMX and power them up.

EDIT : Obviously upgrade the VM tools before the virtual hardware.

Cheers,

Jon

Message was edited by: jrmunday

vExpert 2014 - 2022 | VCP6-DCV | http://www.jonmunday.net | @JonMunday77

View solution in original post

Reply
0 Kudos
5 Replies
jrmunday
Commander
Commander
Jump to solution

Have you tried powering off the affected VMs' and resetting the CPUID mask?

CPUID-Mask.png

vExpert 2014 - 2022 | VCP6-DCV | http://www.jonmunday.net | @JonMunday77
Reply
0 Kudos
D4veB
Contributor
Contributor
Jump to solution

I hadnt so I just tried it against a test server I have and It didnt work. Thanks for the great suggestion! Smiley Happy

Reply
0 Kudos
jrmunday
Commander
Commander
Jump to solution

The EVC mode is set on power up, so I suspect there may be some legacy entries in the VMX files which are affecting this. Can you upgrade the virtual hardware to the latest version supported by your hosts and test this again (including resetting the CPUID mask).

If this also fails, or you are already on the correct hardware version can you please post the VMX file of an affected VM and the VMX file of a newly created VM (with the same virtual hardware version).

Another option to test if the VMX file is the cause, is to unregister the VM, create a new VM but attach the existing disks - power this up and check the EVC mode. But it might be simpler to upgrade the virtual hardware, reset the CPUID masks, remove the offending entries in the VMX and power them up.

EDIT : Obviously upgrade the VM tools before the virtual hardware.

Cheers,

Jon

Message was edited by: jrmunday

vExpert 2014 - 2022 | VCP6-DCV | http://www.jonmunday.net | @JonMunday77
Reply
0 Kudos
D4veB
Contributor
Contributor
Jump to solution

"Another option to test if the VMX file is the cause, is to unregister the VM, create a new VM but attach the existing disks"

I have tested this with my test server and the EVC level is now reported as "Sandy Bridge" which means it was something buried within the VMX file from the old ESX4 infrastructure. However the Vm in this case a Windows 2008 R2 Enterprise server would boot but blue screened before completing windows start.

Im going to try "But it might be simpler to upgrade the virtual hardware, reset the CPUID masks, remove the offending entries in the VMX and power them up." next and will report back..

Thanks for your Support!

Reply
0 Kudos
jrmunday
Commander
Commander
Jump to solution

No problem - glad you know where the issues is now Smiley Happy

vExpert 2014 - 2022 | VCP6-DCV | http://www.jonmunday.net | @JonMunday77
Reply
0 Kudos