I am having an issue on a two node ESXi cluster (4.1 Update 1) and VCenter (4.1 Update 1)
Two ESXi servers are in a HA and DRS enabled cluster. 10VM's in total, no memory or CPU reservations. One resource pool with all VM's inside of the Pool.
When I attempt to put one of the hosts into maintenance mode I get the old DRS error.... "Insufficient resources to satisfy the configured failover level for HA"
HA admission control policy is set to "Cluster tolerates one host failure" If I turn off HA admission control I have no problem, so clearly HA admission control is not happy with something, but I cannot determine what exactly the cause might be.
This is really expected behavior. In a 2-host cluster if you put 1 host into maintenance mode you're basically giving up HA protection since you're effectively making it a 1-host cluster which obviously cannot protect against a host failure. If HA admission control is enabled (specifically with the "host failures cluster tolerates" policy) HA won't allow DRS to migrate vms off the host entering maintenance mode since there's not enough failover resources to protect them. You can temporarily disable HA admission control if you need to put one of the hosts in maintenance mode.
Hi there Elisha
Thank you much for your responses. Your comment "checks that there are at least 2 HA-enabledhosts" is this documented anywhere as I cannot recall ever coming across it ?
Removing a host from a 2 host cluster (or putting a host into maintenance mode) - reduces you to a 1 host cluster.
Of course, with 1 host in the cluster, you can not satisfy ANY level of HA - so the only way to perform your maintenance is to tell HA to allow policy violations, or disable HA while performing your maintenance. - not optimal, but unfortunately this is the position you'll be in wheneevr you have a cluster with just 2 nodes.
This is expected behavior indeed and I would be shocked if it wouldn't work like this to be honest. Placing the host in maintenance mode would violate admission control and this is something that needs to be avoided at all times
Available now on Amazon: vSphere 4.1 HA and DRS technical deepdive
I do not disagree with what the expected behaviour should be in 2 two host cluster with admission control enabled and attempting to put 1 host in maintenance mode. However in the same scenario and with isolation response for VMs set to shutdown, meaning they will restart on another host in cluster. Why if I pull the power on 1 host do the VMs shutdown and restart on the other host? It seems if the cluster is configured to tolerate 1 host failover and one host has lost power the VMs would exhibit the same behaviour as they would by putting a host in maintenance mode, that is, generate an error that there is insuffciant resources... That is not the case.
I don't really understand your comment Mark. It is fairly simple:
Admission Control ensures enough resources are available to restart you virtual machines and this is handled by vCenter(!).
Host01 is placed in maintenance mode, admission control is disabled, Host02 fails. No host left to restart VMs
Admission control is enabled, as such vCenter and HA will disallow placing the host in maintenance mode as that would violate availability constraints!
Admission control is enabled, host01 fails, virtual machines are restarted by HA on host02.
** the reason the virtual machines are restarted is because Admission Control is done by vCenter and NOT by the host or the HA agent.
Most of this is explained in my book about HA by the way: http://www.amazon.com/dp/B004V49JGW/ref=as_li_tf_til?tag=yellowbricks-20&camp=0&creative=0&linkCode=...