Strategies for managing long lived multi-session AVD VMs
Title
We are using AVD to make some vendor software and its central data store available to remote users (the vendor's traditional approach is to put the exe and data on SMB shares on prem and have users launch over the network). The app is designed for all concurrent users to launch the same exe and be pointed at the same shared datafiles.
The license is tied to the device (well hard drive serial number) of the machine (or VM) its installed on. It can be moved through an import/export process in the UI, but not in an automatable fashion - this means traditional solutions like appv, app attach or golden image installs won't work for us & we want the session host to persist as long as possible.
AVD with a single pooled session host with the app installed by IT with a 'desktop session' app access and the app made available to users as a remote app is working. To manage costs (app is used 'as required' by the team who owns it depending on current projects requirements) we have made a scaling plan that scales down to 0 hosts, and starts the host on connect. I'd expect it to be online a couple of dozen hours per month at most.
I'm being cognizant that we are 'doing it wrong' from a traditional VDI perspective, where the hosts would be ephermal/spawned from golden images, which is making me worried about day 2 operations.
The first that springs to mind is updates, as the built in update process would normally involve redeploying the VMs (which will then kill the app install and product license). If it was an always on dedicated (instead of pooled) host, I'd just let WUfB handle it, but I'm not sure how well that would work for a pooled host. I've also considered changing the scaling plan schedule to keep it online for a few hours on a weekend to handle updates in a time window - but I'm curious if anyone else has a better solution.
I suspect Intune compliance screaming at us if the VM wasn't used for a few weeks to a month would also get annoying.
Does anyone else have similar deployments that can share other challenges and work arounds?