Deploy OpenStack GA release failed due to anti-affinity rule

Question

When I deploy the OpenStack GA release, following error popped up at the last step on the deployment wizard page.
"Deploy OpenStack instance failed. Insufficient number of ESXi hosts to meet the anti-affinity rule"
I have 3 ESXi hosts in management cluster and 1 host in compute cluster, which I used to deploy the beta version without any problem.
Is there any workaround for this without adding more hosts?

Answer 1

I have the same/similar error:
"PlacementException: Cannot find datastore with enough space and meets the datastore anti-affinity rule for [VIO-Controller-0]. Nodes with the same role cannot be placed onto the same datastore."
According to the oms.log [VIO-DB-0], [VIO-DB-1] and [VIO-DB-2] can be properly placed but the problem is always with [VIO-Controller-0]. Whatever I have done (more disc space, more ESXi servers, more datastores) - the problem did not disappear.
Unfortunately I was not able to find any document which explains the requirements for datastore (anti) affinity - there is only 1 page in the documents which describe the storage requirements e.g. db server requires 80 GB (although actually [VIO-DB-0] could only be placed if the disc has 100GB).
But no word on affinity rules or number of required datastores or datastore size.
I am using a test environment with 3 to 5 nested ESXi v6.0 (!!) for the management cluster with a total of around 1.2 TB free disc space for the management cluster.
To me it appears that there exists "somewhere" an undocumented affinity/anti-affinity rule for the placement of the management VMs, especially for [VIO-Controller-x].
I would appreciate if someone could describe the disc/datastore configuration for a successful implementation of OpenStack GA release!
I do attach an older oms.log with above error message (not sure which configuration I tested at that time, probably 3 ESXi with 3 or 4 datastores with 100GB disc space each per ESXi).