After upgrading 3 nodes of a 12 node cluster from 6.5 to 7.0, we noticed that DRS was forcing VMs to stay on the remaining 9 non-upgraded hosts on 6.5.
Manual vMotion is successful but soon after DRS moves the VM back to a 6.5 host. vCenter as well as 7.0 ESXI host reboot did not help.
Any ideas?
The vCenter is already upgrade to 7.0 and we have checked the compatibility matrix and all the versions are compatible with each other.
Do you have any rules on the cluster?
What happens if you set one of the hosts into maintenance mode?
Could be the "risk/benefit" analysis that is skewed as a result of the upgrade and the associated reboot. (New DRS algorithm was also introduced in 7) I have not seen this before.