I didn't see the original posters question before I wrote up my own version of essentially the same question. I'll post it below, it may have some additional information.
I was doing some login timings this morning and I noticed something very odd, depending on my monitor arrangement the login times to a desktop can vary by as much as 500%. Really.
For background I'm testing with three monitors, one is a laptop display and the other two are 2560 x 1440 displays. We're on View 7.9 with UEM 9.8, the desktop is 1903 with nothing but the basics (Office). This is a test pool with a manually assigned machine, I'm sitting next to the datacenter.
My best time is using only the laptop display, I can go from disclaimer acceptance to desktop in 11 seconds (doing the same test without UEM brings it down to 9 seconds). When I change from the 1920 x 1080 laptop display to one of the other monitors the time jumps to 12.7 seconds. Again, for this test I'm only using a single display, the only thing I changed was which monitor the View desktop displayed on.
If that's as bad as it got, I could live with it…however it gets much worse. When I use the two standalone monitors the time jumps to 51.3 seconds. For a setup with all three displays the login time again jumps, this time to 56.7 seconds.
My thorough testing was all done with Blast, however spot testing with PCoIP shows that the problems exists there as well. The timings are similar between Blast and PCoIP with PCoIP being slightly slower on the single monitor logins, and PCoIP being slightly faster on the 3 monitor test.
The only thing I'm changing is the number of displays and that led to a dramatic, I would say unusable increase in login time. I ran each test three times and then averaged the timings.
Does anybody know what's going on and/or how to fix this issue?
As an aside, I utilized the logon monitor to try to determine what's up, however whatever the issue, it is not being captured by the logon monitor. A good portion of the delay seems to be happening prior to the logon monitor starting to collect data. The amount of time not captured by the logon monitor varies from 8 seconds to 25 seconds, the amount of time being lost is proportional to the overall logon delay.
I seem to be having a similar issue - if I have more than 1 monitor connected when logging in I get anywhere between an extra 40 - 90 seconds delay. You can tell its slow right away as the spinning circle animation during login looks slow and choppy. We're using Horizon 7.9, Windows 10 Enterprise 1903. Also doesn't seem to matter if its a zero client or the windows client. Any suggestions are welcome. We're still working the problem as this was just recently noticed.
Can anyone comment on what this could be? I've been testing - doesn't matter if its a zero client or if you are connecting with a windows client. I've tested it with the hardware config of the video card set to auto-detect and with it set to 2 monitors specifically. If I'm connecting with 2 monitors to start with, its a slow bootup. Typically 30+ extra seconds of delay. And its not as if only one spot of the boot up process is slow, it seems like each step is slower. Logon monitor doesn't really pick up on this much as it only reports about 10-20 extra seconds and the extra 20 or so seconds I am there waiting for the desktop to appear go unaccounted for.
I troubleshooted with VM for about 3 weeks. Couldn't locate the issue so I've created a new masterimage. This fixed the issue for us.. Not the ideal solution
Well that's good news that you have it fixed. Do you have any idea what fixed it? Even a suspicion? Did you start from the same ISO? Maybe something was different vm-hardware wise with the guest? More windows updates?
I appreciate your reply.
Did you ever try bumping up the video memory in the golden image? If its not high enough I've seen that cause problems.
Yeah i did. That didn't help for us.
No clue what it started. At the beginning we thought it had something to do with Nvidia drivers, but we completely removed all of the drivers but that also didn't help.
According to VM this can happen when you update the build version.
I also downloaded the latest ISO.
Thing we checked (and didn't worked for us but perhaps for someone else)
Reinstalled Horizon agent
Removed all video drivers
We tried a few different optimization tools for VM (one actually got the load time down to 40 seconds, but that was still 20+ seconds to slow)
Edit the pool settings in the Horizon Admin portal.
Increase the number of monitors and the resolution to the maximum value.
Shutdown and then power on all the virtual machines in the pool using the vSphere Client.
This sets the video memory to maximum setting, between 125 MB to 128 MB.
Perhaps things got worse when we updated the Horizon agent from 7.8 to 7.9? I uninstalled and rebooted the other agents in opposite order (so, remove appvol, reboot, UEM, reboot, Horizon Agent, reboot) and then installed 7.9. Figured that was the safest way. I wonder if something went sideways during that process. I'm rebuilding my image also from scratch, and ditching the use of sysprep copyprofile and mandatory profiles. Maybe we'll shake this dual monitor slowness in the process.
Our new image still has the same issue with the monitors. I've tried messing with many of the display settings to no success. Also I can't edit the pool and set the number of monitors - this is not available for instant clones. I've opened a support case with VMware on it.
What about upgrading to 7.10 ? Both Horizon client and Connection server ? Have anyone tried ?
Disphar, can you share a little more about exactly what you did? Right now I don't know if the problem is on the View side, or on the Windows 10 side...and if all you did to fix the issue is to recreate the master image, that points to something that is fixable on the Windows 10 side.
When the image was built, did you use the OSOT also? I have noticed that with the old verison of the OSOT it cause many problems, even with the VDILIKEAPRO one. I'm building an 1809 image now with the OSOT b1110 to see what this does to our logon times.
I'm sorry to say that 7.10 doesn't fix the issue, I'll post my configs at the bottom of this message. I've been testing on a brand new environment with 7.10 and a couple of app layers and my logins are about 5 minutes, 30 seconds across three monitors. It really does seem that like whatever is happening is just slowing down the entire login process by such and such percentage. Meaning that regardless of what I add into the mix (DEM or App Volumes) the timings are proportionally impacted by how many monitors are being used.
And actually, let me clarify the statement about the monitors, this problem actually seems to be related to how many pixels are being used. My fastest timings are always when using the desktop client and setting the display to "Window - Small". If I set the display to "Window - Large" the timings slow down...if I then go full screen the timings slow down more....and so on.
Disphar mentioned that they replaced the master image, I'm going to try this next...
7.10 Connections servers
App Volumes 18.104.22.168
ESXi 6.7 U2
Windows 10 1903, patched to about a month ago, Office 2016 baked in
Everything is running on flash only storage.