As I discovered last week, while playing with my Vblock lab, there is a potential problem in the way, vCloud Director works with Virtual Center.
As vCD uses resource pools to separate the compute resources in the virtual data center, these resource pools are vital for vCD to work.
What I did was:
I disabled the DRS feature of the ESXi Cluster, just to check some settings.
This led to the loss of all resource pools, which wouldn’t be such a big problem, as you can easily recreate them, if you use them in a normal environment.
But with vCloud Director, things are a little different:
If you loose the resource pools, vCD can’t create any new machines, as the resource pools are missing.
I tried to recreate the resource pools by hand, but this didn’t work.
Next thing was a discussion on the VMware forums with Chris Colotti.
No solution over there, except the GSS support call, but as I am still in the phase of testing, I didn’t want to bother them and the whole recreation process of the vCloud Director organizations etc. doesn’t take that long and is a nice training for the productive phase when new customers will be placed on our Vblocks.
Today I red this article on Chris Colotti’s blog and thought to myself:
Let’s give the Inventory Snapshot scriptthe from VMware Labs a try.
I recreated some customers, VMs, networks, etc. and let the script do it’s snapshot work.
After creating my inventory snapshot, I disabled/enabled the DRS/HA features of the cluster and all resource pools were deleted.
Next thing to do is let the recovery script, which was created by the inventory snapshot script, run against the vCenter to let it recreate the resource pools and folders.
As promising as it looked, the script did not help, as, according to Chris’s article, the MoREF ID’s don’t seem to match (this is just an educated guess, but seems legit) so vCloud Director doesn’t take the recreated resource pools.
Maybe someone from VMware could take a look at this and adjust the script, so it will cover the missing parameters.
Until that day, DRS and HA have to stay untouched, to keep the vCloud Director working.
