Oddness After Upgrading from 2006 to 2111

[kanid99]

After upgrading from 2006 to 2111, and converting our desktop pools from Floating Non-Persistent Linked Clones to Floating Instant Clones, we discovered we had an issue that we hadn't experienced in our initial pilot testing groups.

Setup :

Guest 
Windows 10 20H2
nVidia Tesla P40 GPU (1gb profile)
Host
ESXI 7.0u3d
Horizon 2111
Pool - Provision On Demand Floating Instant Clones
Client
Mix of 10zig thin clients and Windows clients

Problem :

We configured the pools to have enough minimum machines to cover initial logins, planning that the instant clones would be created fast enough to cover additional demand after that. This was working but what we were finding was that users who logged in right after the initial login storm were being logged into machines that were still saying 'Getting your devices ready' and then the user would get a 'Lock Screen' which if they CTRL ALT DEL on they would be allowed to proceed to login ... except, their display was all messed up. They had scroll bars on side and bottom as if the detected display was 2x the resolution of the actual display. Disconnecting and reconnecting resolve the problem but this is highly inconvenient to the users.

VMWare support suggested not using the BLAST protocol and using PCOIP instead and believe it or not that fixed the issue but this too is not ideal as we prefer the BLAST protocol. Has anyone experienced anything like this ? I suspect Ive missed something in the setup that has caused this but I dont know what!! The last thing Ive tried was to uninstall the nvidia driver, vmware agent and vmware tools and reinstall in the recommended order (tools-nvidia-agent) but that didnt fix it !

[gelliottaus]

I would have a play with using a postsync script. We had some issues with appsense not being prepared and ready for a user so we added delays in the script to allow the new guest a minute or two to settle after being created. Our script is similar to below... the timeout commands causes a delay in seconds. 

net start "AppSense Watchdog Service"
net start "AppSense Client Communications Agent"
timeout 20
net start "AppSense EmCoreService"
gpupdate
timeout 30

You will also need to update a registry key to tell the horizon agent to wait for your script (otherwise it will get killed off due to its long execution time). HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\vmware-viewcomposer-ga\ExecScriptTimeout

[kanid99]

The issue ended up being related to a group policy that is applied to the OU the instant clone desktops are in that sets the EncoderMaxFPS for Blast. For some reason, if this setting is applied during the login process when the user has grabbed a machine as its completing its prep process - it hoses up that connection. By disabling the EncoderMaxFPS setting from the GPO , the issue went away.