Can someone point me to an article or explain what the best practices are for a esxi host cluster dedicated to VDI when floating instant clones are in use? If the user logs off or the host crashes and the VMs along with them, is there any type of HA worth having on? This shouldn't affect maintenance mode. Here are the failure scenarios available where HA can be on or off:
Datastore with Permanent Data Loss
Datastore with All Paths Down
Guest not heartbeating
Seems silly to restart VMs on another host since the horizon software should see that there's not enough VMs and spin up new ones. Or maybe not since it won't be a graceful shutdown of the VMs?
Thankfully we do. The servers are in one cluster with it on and the desktops in another cluster. Currently it is off.
Since the pools have some VMs in an unused state at all times to protect against large logins at the same time. Our two main pools have 5 each. In an HA scenario those would restart automatically if they were on a host that crashed, and since not in use Horizon should theoretically keep them available to log in so that users can log in right away. If they were not restarted then that buffer would have to be created before users could start logging in.
Also not sure if the ADAM db cleans up the database correctly or not. and how would this change if we had the need for a couple persistent desktops. would the instant clones be OK with it on?
Horizon should catch the persistent desktops too as long as the power policy is set to always on, so you don't really need them there either, just try and turn one off and they always come back on.
Thanks. So if HA is turned on and user sessions are logged in and some disconnected and some Available and a host goes down, do the desktops just restart on the other host and when users re-log-in, does it reconnect them or are the desktops fresh again (instant clones, floating)? And does the ADAm database take care of itself usually or do we need to clean up old ones manually?
I mean you weant ha turned off, there is no need for that feature for anything managed my horizon. If a host goes down and any persisten desktops crash horizon should see it and start them. For non persistent horizon will see it and delete the vms and recreate them. The adam database should be fine as long as none of the vms are recreated by anything outside of horizon. Mine has been setup this way for awhile and I've had host failures and never had any issues .
Do you know any reason to use HA for anything else in Horizon? The documentation seems to make a point that Horizon supports it. Is there a case where desktops would want it to be on? I can see why you'd want it for the server side pieces.