Aloha - Yesterday we tested HA on our cluster by yanking power on two of our four hosts. Test succeeded and VMs restarted elsewhere. However when we restarted the hosts, VMs did not migrate back to them. Generate Reccomendations did nothing. I am however able to manually VMotion to the servers. Any ideas on wha the problem may be?? Servers at times are very unbalanced.
Even though your servers are unbalanced in numbers of VM's pr. host does not mean that your hosts are overloaded. I have seen similar behaviour many times
in our enviroment. How Aggresive is your DRS? You can always test DRS by setting a host in maintenance mode.
Frank Brix Pedersen
I'm bugged mostly by no VMs migrating back to the previously powered off hosts. I have DRS set to automatic and most agressive. I have also turned DRS off and then back on. In the past we averaged 2 or 3 migrations a day. My definition of unbalanced is based on CPU and RAM utilization and not VM quantity.
It may not really be a problem.
I have 2 hosts, Dell R900 4 Socket Quad Core 96 GB RAM. I have 14 VM's running on 1 server. Unless I set the DRS to extremely aggressive the VM's will not move. They only take up 20% CPU, and 30% RAM.
So internally there must be some threshold that isn't met which would force a migration.
What does your host performance look like BEFORE you attempt to migrate them? Maybe the machines that are idle don't warrant a move.. or maybe it will happen later, but if it migrated before, it will migrate again, I wouldn't worry about it. If you can manully migrate the machines, everything is working fine.
also, check your VMware DRS Resource Distribution table. If your "Percent of Entitled Resources Delivered" is high (90+) no matter what you do, you may not be able to generate a DRS event. If the graph is closer to the middle or left, you may be able to generate an event by changing the Migration Threshold