I have a MUDP that has Base 6 host, Min 1, Burst 2 with AVD agent Available. Auto-scaling is set to scale out 1 host if CPU is over 65% for 5 minutes. Also pre-staging for the morning influx. There are about 50 users over two shifts. I have Depth first load balancing of 6 users, but got a call from users saying they could not log in. I saw the first host was full, but no other hosts were powered on. I added Scaling Logic Avg active sessions: start 1 host of avg active sessions across all host exceeds or equals 6. Stop 1 host if avg active sessions across all host below 1. This caused one host to power on. I am worried auto-scaling like this may cause power cycling of hosts or if two hosts are full that the next one will not power on.
Is this the best config for auto-scaling or am I missing something?
Issue with Auto-Scaling Multi user desktop pooled
I'd be interested to see what other partners are doing with this or what Nerdio recommends.
We've seen that same problem with the avg active sessions feature.
The best thing we've seen to do is to combine the User Driven (Scaling Logic), Start VM on Connect, Set all hosts to running OS disk type, and Pre-Stage Hosts (Multiple Schedules to accommodate different shift schedules) features together. That gives us a good balance of host availability and cost.
We also are using E-Series Hosts with a 15 user threshold and haven't had users complain about performance.
Depended on the workload, you may be able to put more users per host to help with your problems as well.
DStephenson Interesting, though I do not see User Drive in Multi-User Pooled, just in the single-user pooled and single-user personal. We are trying to stay with the Multi-user environment as it appears to be the most cost effective for what the client does.
We're only using Multi-User for our customers too.
However, the "User Driven" can only be used alone (with no other scaling plans). That's one of the "features" right now.
Please sign in to leave a comment.
Comments (3 comments)