Some of our VDI users intermittently see one monitor displaying black screen and then the second monitor displaying black screen. Then it goes back to normal and displays expected content. This happens every once in a while and it can't be replicated on demand. Has anyone seen this? Thoughts?
Windows 10 1607
Horizon 7.3.1 server + agent
We had something similar - the resolution would drop to 800x600 and the screen would flash black. Do you have anything in the WIndows event logs when the issue occurs? For us, we would get dwm.exe errors and dwm.exe would terminate because the process could not allocate additional memory. It turned out we had incorrectly configured the virtual memory settings on the parent VM.
When you create a View Composer linked-clone desktop pool, you can redirect the clones' paging and temp files to a separate disk. You must configure this disk to be larger than the size of the paging file on the parent virtual machine.
Set the paging file size to a larger value than the size of the memory that is assigned to the virtual machine.
If the Maximum size (MB) setting is smaller than the virtual-machine memory size, type a larger value and save the new value.
When you configure a linked-clone pool from this parent virtual machine, configure a disposable-file disk that is larger than the paging-file size.
After setting the virtual memory to be slightly larger than the RAM and the disposable disk to be slightly larger than the virtual memory setting, the problem went away.
We had this exact issue, it occurred roughly every 90 minutes or so-- at the same time Group Policy updated in the background. This occurred only over PCoIP for us, but try to replicate it by doing a gpupdate /force, your screens will flash if it's the same issue. For us this was caused by the "Configure PCoIP Image Quality Levels" ADMX setting, the issue was resolved after disabling that setting and allowing it to automatically dictate quality based on the bandwidth restrictions we set. In either case, if forcing a group policy update replicates the issue for you, it's likely one of your policies causing that like it did for us.
I've seen this many times and it's always a resource issue (We once resolved this by upgrading VMware Tools due to a memory leak it had). First make sure VMware Tools/Horizon Agent are current. Then address any resource constraints in the VDI.