Here is my set up (fresh installation):
Windows 10 1809 LTSC (Fully patched)
Citrix VDA 1912
VMWare ESXi 6.7U3
VMware Tools v11.0.5-15389592 (SVGA Driver not installed)
Microsoft Basic Display Driver Adapter (disabled)
NVIDIA-GRID-vSphere-6.7-440.53-440.56-442.06 - v10.1
GPU: NVIDIA Tesla M10
vGPU profile: M10-1B
When the Citrix VDI session is initiated a grey screen is all the user can see. Please refer to the screen shot below:
Surprisingly, if a master image, same setup specified above, but running Citrix VDA 7.15 LTSR CU5 is upgraded to Citrix VDA 1912 LTSR, the issue goes away and the VDIs work wonders.
can you try to resize the window? Does the grey screen goes away? I have this issue on some clients for example my home pc aswell. It seems that the video stream freezes and when i resize the window it gets refreshed, i only have this issue on the start and not while connected.
The reason we are using Windows 10 1809 LTSC, which is supported by MS for 10 years, is because it does not come with all the crapware and modern apps, what consequently makes it lightweight and ideal for a VDI environment.
If removing the default rubbish is the sole reason you’re running an older OS, then a better option may be Server VDI.
You could also use the Citrix Optimiser or VMware OS Optimization Tool to remove that default rubbish, and of course there are plenty of custom Powershell options to remove it as well if that helps give you more options.
Something else you can quickly try … Can you remove / disable all of the Citrix Policies applied to that specific VM (or Delivery Group) and try again. More of a sanity check really.
Just out of interest, are your Delivery Controllers upgraded to 1912, or are they an earlier version?
As for what I think … I think I’m going to bring up my Lab and have a test for you to see if I can replicate it … :-)
We use Citrix Cloud Service, so the Delivery controllers should be running the latest version available.
Let me rephrase this comment from my original post above:
If a master image, same setup specified above, but running Citrix VDA 7.15 LTSR CU5, and then upgraded to Citrix VDA 1912 LTSR, the issue goes away…but there is a nasty memory leak with the CtxGfx driver.
"Something else you can quickly try … Can you remove / disable all of the Citrix Policies applied to that specific VM (or Delivery Group) and try again. More of a sanity check really."
Removed the Citrix Policies mentioned above:
…from the VDIs’ OU and the user I am using for testing, but it is still reproducible.
Install Windows > Install VMTools (Without vSGA Driver) > Install vGPU Driver > Install Citrix VDA
Yes, I did it in that order without without vSGA Driver.
When you upgrade the VDA from 7.15 to 1912, do you just run the upgrade or do you do anything with the vGPU driver?
I just run the upgrade, nothing else.
Unfortunately I can’t replicate it. 1912 works without any issues for me out the box.
Win 10 1909 clean build from .iso and fully patched as of this morning > VMTools 11.0.1 > vGPU 10.1 > VDA 1912
It’s a clean build Citrix environment (reinstalled this morning) consolidated onto a single VM (DDC, Storefront, Director, SQL Express). No Citrix Policies have been created and only the default ones are applied.
Apply your registry settings via GPO, it’s easier to review and keep track of what’s being applied when you upgrade your VDA or Operating System.
Regarding the memory leaks, the only advice I can offer is to make sure you’re running the latest software throughout your stack. The 2003 VDA is now available, upgrade to that in combination with vGPU 10.1 and see if that helps.
As I mentioned earlier, we are assigning M10-1B vGPU profiles to these Windows 10 VDIs.
These users are using Windows-based thin clients with 2 x monitor @1080p. They are office power users that pretty much use enterprise apps, but they rarely run multimedia stuff such as YouTube, video conferences, and video training. On top of that, we use Pi-Hole to block ads and push ad-blockers extensions/add-ons to their browsers.
I used the following guide as a reference: http://images.nvidia.com/conten t/pdf/grid/guides/vgpu-profile-sizing-guidance-for-windows-10.pdf
Based on your experience, would it make sense to upgrade their vGPU profiles to M10-2B?
Unfortunately this isn’t an answer anyone else can give you, as your environment specifications, applications, user working habits and therefore overall utilisation will vary to anyone else’s. You’ll need to monitor the utilisation on a subset of your VMs and decide what’s appropriate. Here’s my favourite utility for this kind of thing: Releases · JeremyMain/GPUProfiler · GitHub It’s nice and easy to understand, just capture the metrics for a group of users (the more the better) and take appropriate action based on the results. Other tools are available if you wanted to dive a bit deeper.
If you increase the vGPU Profile size, at best, you halve the density of the GPU potentially doubling your overall Server hardware requirement, unless you have some headroom built-in.
This is why it’s absolutely critical to run well defined POCs before the hardware specifications that will be used are finalised. If unsure, it’s far better to over-spec than under-spec the environment. I’ve seen it happen before in an environment where the customer used the the wrong vGPU Profile and it worked out very, very expensive for them to put right.