I haven't been able to find an answer to this one. Hopefully someone might point me in the right direction. I have six ESXi 4.1 U2+ hosts in a cluster with HA and DRS enabled. HA's ACP is set to tolerate one host failure.
I made some NFS/Net changes recently and needed a reboot to put them into effect. I put one host into Maintenance Mode and it evacuates all the VMs without issue. I reboot the host, it comes back, and I remove it from Maintenance Mode.
When I move onto the other host after five or ten minutes and start the process again...VMs will NOT migrate to the host that I had just rebooted. It's pushing them all to the hosts that have been running longest.
I'm not seeing any rule or advanced options that would prevent guest VMs from going to a newly rebooted host that's ready to go. my das.minUptime value in the HA advanced options is set to 120 seconds...which is default. I thought it might have been that option, but don't believe so. I've let a rebooted host sit for 15 minutes and still no VMs will migrate to it when I "flush" the next host.
DRS:s purpouse in life is to make the VM:s to run as optimal as possible and not to evenly distribute the load throughout the hosts.
Maybe you have not enough contention on the hosts for the VM:s to benefit from being moved.
You could try to set the DRS Level to "Aggressive", then you should see more action.
I've seen it wake up within a few minutes on a cluster that needs the load-balancing to me watching wondering if VMs can't move. Because I can be impatient, when the host is up and I think is ready for VMs, I just kick off a manual DRS run (DRS tab, Run DRS in the top right corner).
If nothing needs to be done, nothing will be done and setting it aggressive can work against you by forcing VMs to move where there' s little chance of them getting a performance benefit.
Thanks for the suggestion...i'll move it towards the aggresive side. It's currently in the middle.
Or manually move VMs around if you feel the need. In the end vSphere DRS is all about providing the resources required. If there is no big imbalance and everyone gets what it is requested... it won't do anything. You can check the DRS tab on your cluster to see if there is an imbalance.