Windows 10 1909 desktops unavailable

[AndrewPhebus]

I have SEV A incident 120072525000034  open this weekend for Windows 10 multi-user 1909 desktops which became unusable around 6pm EST Friday evening in EAST US 2 datacenter region.  

Users were disconnected and the session hosts fluctuate reporting status of unavailable, upgrading, needs assistance, upgrade failed and available.  I believe the issue stems from the agent upgrade failing and being stuck in an upgrade loop.  I have removed the agents and reinstalled them as well as deployed new machines.  Nothing I can do seems to help resolve this.

Azure support team is reporting no Global issue yet surrounding this behavior.  If you are having any issues similar please respond.

[#2]

@AndrewPhebus Removing session hosts and redeploying new ones does not work? Replace host pool with a new one built from scratch?

I have also seen issues with trying to "fix" broken agents with "unavailable" status on session hosts, have not been able to find a trustworthy way that always works, except totally remove and replace session hosts.

Very interessted to see what support comes up with in your case, good luck and please let us know.

[#3]

@Ola Holtberget 

On win 7 enterprise for one client we had an issue with a bad agent update push which resulted in desktops going unavailable.  Microsoft was able to resolve that by rolling back that code on the back end.

For the case over last weekend with Win 10 1909 clients, it turns out there was an issue with a virtual network with a NAT Gateway configured on the subnet in EastUS2 where the NAT Gateway was causing latency and timeouts to the internet.  They resolved that networking issue on Monday afternoon.  The internet connectivity issue was preventing it from downloading the latest update and subsequently cycling through the update cycle retrying as well as going unavailable as they couldn't talk to the broker.

I am still seeing increased latency in East US data-center when I turn on the NAT gateway on my subnet.  If you are having frequent issues with disconnects and user complaints I would suggest trying a test VM on a subnet without a NAT Gateway and see if performance is better or more stable.  I am continuing the support case.