Windows 2019 based Xenapp VDA in “Maximized” mode freezes upon or shortly after login if an Nvidia grid vgpu is present

[Andy Vanderbeken]

Guys,

 

I thought I’d throw this into the group: After months of suffering from an occasional bug where people’s Citrix sessions seem to freeze during or very shortly after logon I have been able to deduce the cause down to something related to the gpu acceleration and maximized mode. If I for instance edit one of the Vmware virtual machines and remove the gpu this problem does not happen at all on that server. It also only happens when the user uses his Citrix Xenapp session in “Maximized” mode. However equally if someone does not set the registry value (screenpercent 99%) in order to start their daily Citrix sessions in "Maximized" mode the problem doesn't happen either despit an Nvidia grid gpu adapter being present. The combination of these 2 factors seem to trigger it somehow.

 

Also interesting to note is that whenever the screen at the end user laptop or pc seems to freeze (like a glass screen you see a frozen picture of the current/latest screenrefresh) there is in reality no problem on the backend VDA server. The session never froze there or showed any hickup or error. As such a simple confirmed workaround is disconnect+reconnect to the session.

So in summary this mystery feels like the Citrix is trying to switch codec or something similar in order to activate gpu acceleration and freezed at that point.

ps: of course I have opened a Citrix case as well but I thought I would work both fronts as people here might have this problem too or recognise it.
 

additional new information after comparing data from multiple test scenario's: I have a desktop pc with a discrete nvidia geforce adapter and 2 external monitors connected to it. There the issue never happens either despite starting my Citrix sessions every day in "maximized" mode. However on both a personal laptop of 1 brand a company laptop of another brand which both only have their internal screen and no external monitors the issue is 100% reproducible. They also have a different native screen resolution where 1 is Full HD and the other 1366 x 768

 

 

any ideas ?

[Krasimir Churov1709160152]

Hi Andy, 

 

We are affected by similar problem. Did you ever find resolution for this?

We have Win10 VDIs with nVidia GPU profiles Inside the VDI the user has Citrix workspace and makes connection to another 2019 VDA Citrix farm to run a published application there.

Inside the VDI the workspace freezes sporadically most often soon after the app is launched and the app is also launched always in Full screen mode. 

Very baffling also no events are logged anywhere, in order to resolve you have to forcefully disconnect the published app and reopen...

[Andy Vanderbeken]

17 minutes ago, Krasimir Churov1709160152 said:

Hi Andy, 

 

We are affected by similar problem. Did you ever find resolution for this?

We have Win10 VDIs with nVidia GPU profiles Inside the VDI the user has Citrix workspace and makes connection to another 2019 VDA Citrix farm to run a published application there.

Inside the VDI the workspace freezes sporadically most often soon after the app is launched and the app is also launched always in Full screen mode. 

Very baffling also no events are logged anywhere, in order to resolve you have to forcefully disconnect the published app and reopen...

 

Yes, I opened a case and after a long chain of troubleshooting and escalating it became clear that this is a shortcoming in the legacy thinwire graphics engine which we triggered by defining the 99% value (for maximized) in the registry. The solution to decide to move forward and abandon that all together by switching over to the Citrix Desktop Viewer (Citrix Workspace app full app) which handles graphics alot better. Ever since we have no issues. 

 

The issue you are having is clearly of a different nature but I can see that you very concept induces this problem already. In order to handle graphics (especially with vGPU) properly you are supposed to run the CWA clients on the physical machines (because it leverages the local machine's hardware even in combination with vGPU) rather than on a  remote Windows 10 VDI and thus connect directly to your VDA. I would not be surprised if it turns out that the extra step you introduced causes these problems in fact. Try it out and let me know.