I have 6 ESX 3.5.0 systems currently divided into three 2-node clusters.
Cluster 1 is HP DL380 G6 with Xeon E5540 CPU's and is not licensed for vMotion
Cluster 2 is HP DL380 G7 with Xeon X5650 CPU's and is licensed for vMotion
Cluster 3 if HP DL380 G6 with Xeon E5520 CPU's and is licensed for vMotion
All nodes are live and in production. Recently while enabling hyperthreading on nodes of cluster 2, the manual failovers didn't go well and we had a lot of downtime while systems fought for resources. HA has admission control enabled and 1 host failover capacity so I'm not sure why we had such extreme problems. Based on that issue, I'd like to create a larger cluster if possible but that brings a series of questions:
1) What nodes are hardware compatible to create a larger cluster? It looks like the nodes from 1 and 3 can be combined into one four node cluster but the CPU's are too different in cluster 2 to join them. Is that correct?
2) What implications are there of moving nodes to a new cluster? These are running VM's in a production environment, unplanned downtime is not tolerated (but planned is usually OK). It looks like I can just put a node into maintenance mode, remove it from a cluster, and then re-add it to a new cluster? If I license the nodes on cluster 1 for vMotion before this work, can that be done with no guest downtime?
Any gotcha's?
Everything is scheduled to be replaced later this year (new servers, VMware, and SAN) but with the failovers not working smoothly on cluster 2, I'd like to get this fixed as much as possible as soon as possible, with of course as little cash outlay as possible.
Thanks for any insight you can provide!
