I have two esx servers on which I want to house my vcenter. These will be separate from the development clusters as it will only be a mgmt cluster and will only house infrastucture type VMs like the vcenter, the DB server the SSO server.
What is the best way to set this up?
Do I put esx on both make them a cluster and house the vcenter VM on that two node cluster along with the vcenter for the development clusters? Using HA to protect the two vcenters?
Or Do I use heartbeat and there and put a copy of the VC and DB on each esx server?
Or a combination of HB and HA?
Since I'm not such a huge fan of HB I would prefere the solution
"I put esx on both make them a cluster and house the vcenter VM on that two node cluster along with the vcenter for the development clusters? Using HA to protect the two vcenters?"
-> you have your own infra-mgmt cluster which is protected with HA including all VMs you need for running your environment.
Of course it depends on the availability level you want to reach, but in most scenarios a short outage of the vCenter (until HA kicks in) can be tolerated
If you want to have 2 vcenter servers then install the 2nd one in a linked mode with the first one. You can either choose to have a single remote DB for both of them (as a downside you have single point of failure ) or have their separate remote or local DB (hence you will have redundancy against the failure of a DB of any vcenter ). If you have both the vms in the same cluster then u will have a single point of failure (what if the cluster fails, what if the separate network that you have for this cluster itself fails?) and to avoid that you might want to have both the vcenter servers on a different cluster with their different local/remote DB in linked mode with each other to offer the best redundancy possible with the least cost and worry.
IF you are talking about just 1 VC then install it on ur new setup/cluster and enable HA and if u have enough resources then u can even try enabling FT but do u need that level of redundancy if u r taking regular backups of the VC?
u r the judge since all solutions r only specific to a specific config
award points if u find it useful
Seems to me that you are decreasing, rather than increasing availability in creating a separate cluster for vCenter and infrastructure VMs. I have never thought 2 host clusters were a "best practice." Furthermore, in any environment, given the choice between 2 smaller clusters and one larger cluster, I would always go with the larger cluster; up to fairly substantial deployments. Unless you are reaching some of the supported vSphere maximums (32 hosts/cluster), consider guaranteeing/constraining resources for infrastructure management with a resource pool. Also, I agree with lenzker, HB adds complexity not avaiability.