We just replace an existing ESX Host (1 out of 3 in a cluster). Unfortunately, after setting up ESX 3.5 Update 3, we discover that the CPU of the new ESX Host is not compatible with the existing 2 ESX Hosts. We are aware that we can change the CPU mask for VMotion but my supervisor is reluctant as it is not supported by VMWare.
Currently, the new ESX Host is still part of the cluster with HA and DRS enabled.
We would like to ask
1) If the new ESX Host goes to maintenance mode, what will be the behaviour of the VMs running on it ? Will they be shutdown ? From our unstanding, they will not be able to migrate to other ESX Hosts as CPU is not compatible.
2) Should the new ESX Host be part of the cluster ? What the pros & cons ?
Your advice is sought.
It says that "EVC cannot be enabled because one or more hosts in the cluster does not support EVC". Does it mean that the hardware doesn't support or I have to shut down all VMs in the cluster and try again ?
Its a little more complicated. here's what worked for me (assuming your new host DOES support EVC)
1) Create new, empty cluster. setup HA, DRS, EVC as you like
2) Move an OLD host with no running VMs into the cluster (vmotion off its VMs if needed)
3) Add NEW host into new cluster (again, with no running VMs) - this should force its CPU masking down transparently
4) Vmotion VMs from OLD cluster into NEW cluster (yes, this can be done across clusters)
5) Move last host over form old cluster.
The new host does support EVC. I would like to add more information about the CPU in the old hosts and new one.
For 2 existing ESX hosts, they are running Xeon Dual Core P7140M while the new one is running Quad Core X5460.
I follow your advice and enable EVC in the new cluster. When I move the OLD server to it, it says "A general system error occured: The host's CPU hardware does not support EVC, which is enabled for this cluster".
What type of CPU are you using?
- Just saw your other post, new processor should be good but can be confirmed here:
Also, please check in your server BIOS that for the options for "Virtualisation" (Can't remember the exact name, sorry) are "enabled" and in addition if your using Intel the "Execute Disable" feature is ENABLED.
Both of these must be enabled for EVC to function, just so long as your CPU is supported by the current version of EVC - you could also run the CPU Compatibility utility to see if EVC is supported on your hardware.
Unfortunatley, the OLD servers are using Xeon 7140 that is not on the list. The NEW server CPU X5460 does in the list.
When I create an EVC cluster, the NEW Server joins it without any problem.
What will be your advice in my situation ?
Hi Tony - from what you've said the older processor is unsupported for EVC so I guess you could do one of two things, depending on the number of hosts you have -
1/ have 2 clusters - one EVC enabled with your newer and EVC support equipment/processors
2/ Have one cluster with all your hosts but you will not enjoy the full benefits of EVC - time to upgrade the older processors maybe?
Since there is only 1 ESX Host supports EVC, we don't know which approach is better
2 Cluster - 1 with the new server only & the other cluster with 2 old servers OR
1 Cluster - It contains all ESX Hosts though we are not able to vMotion VMs between NEW and OLD servers. However, it still gives us benefits like DRS / HA.
Your advice is sought.
Hi, tbh neither situation is great but I'd go with case 2. At least you will have some resiliency, especially baring in mind you have alread paid for the DRS/HA features.
I do think optioin 2 is more suitable for my case.
I would like to confirm my understanding
1) If the NEW Server is down, all VMs on it will be shut down and start up on the OLD Servers (HA)
2) If there is a new VM has to be started, DRS will determine which ESX Host to host that VM ?