I was wondering what the best practice is when patching an esx host through update manager? Do you just put the host in maintenance mode and let all the guests move on their own? Or do you disable DRS and manually vmotion guests, then patch the host. I'm concerned about over saturating guests being vmotioned.Or is that not a concern, because it will only do x amount and queue the rest?
Please let me know what the best route is. I typically have about 15 - 25 guests on one server a time.
One great thing about VUM is it is fully automated. You just need to created your baseline, attach it to your cluster, remediate against that baseline. DRS will kick in, migrate all your VM's, patch the host, reboot, and take the host out of maintenance mode. The process will continue until all hosts have been patched.
DRS will always maintain management of the admission control, even when the process is initiated by Update Manager, so that it will ensure hosts don't become overloaded. In the case of VI3 with the calculation of slots and the single method of admission control selected it is actually overly conservative and halt the DRS process well before is necessary. If this takes place the automatic implementation of maintenance mode will not complete until manually vMotion the rest of the VMs after manually verifying that it is being overly conservative and there is actually plenty of remaining capacity to be able to take the host into maintenance mode. If going into maintenance mode actually does complete fully automatically then it indicates you have much more spare resource capacity that is necessary to complete the update process. vSphere HA has the other percentage and standby host admission control options that are more precise with how spare resource capacity is calculated and can help the automated entry into maintenance mode complete with tighter resource constraints.