To Be Or Not To Be?
I am doing it and it is working great for me. Part of the reason I did it this way was for DR and recovery
Both!
Generally....as a standby server is fine, with the database residing on a separate server.
OK so seperate HW still needs to be purchased...mmmm.... I like the idea of VM with the HA etc...any other thoughts?
Mainly would it be supported it totally encapsulated on a VM.
Yes you can run the VC server inside a VM, you can also run the corresponding database server in a VM. I can't see Vmware not supporting it just because it is a VM. There are some downfalls to running it as a VM (if your ESX host it is on fails, then you have to attach to ESX servers directly) but it should not be a problem. The documentation lists specific hardware/software requirements for VC Server as long as you meet those it should not matter if it's a VM.
Did it several times. Completely encapsulated with MSDE DB so that there are no dependencies to other DB servers in disaster case.
VC is only needed for configuration of HA and DRS, but HA and DRS do work without VC online or reachable.
If you run the VC as a VM and configure the cluster to use HA, you're sorted.
A failure will simply move the VC to another host.
VC does not need to be running for HA to function.
cheeko, I believe VC needs to be up and running in order for DRS to function, since DRS basically works off the VMotion function.
Your right HA works without it
I believe the official stance is that yes, VC as a VM is supported, however Vmotioning the Virtual Center VM is not, because VC is the one that takes care of orchestrating the VMotion operation, and it shouldn't be VMotioning itself!
In practice it does work, but you may run into problems; so make sure you set the VC VM to be ignored by DRS.
Also note MSDE isn't supported for production environments, you need to be running a real SQL server. In our case, we made VC a physical box to avoid the problem of VC being unavailable if the cluster was offline.
yeah that's true, that was crap with DRS ... sorry for mixing up things.
But as for HA it's certainly the case.
I would argue that it depends on the size of your environment. If you only have a hand full of ESX hosts then it makes little sense to maintain another physical box for VC, plus how does on justify that when you're trying to get many others to migrate from their physical servers and on to the magical virtual servers.
Of course the bigger the environment the more sense it make to go with a physical host for VC and even a separate clustered host for the DB. This is because VC & the DB have become more important.
We're still in the planning stages, still figuring out what we need and how big. But I hope to start with a VM for VC and then later re-use a freed up physical host, but we'll see what happens.