Remote Desktop Session Host becomes increasingly unstable beyond 75-85GB RAM - WS2022

Question

Remote Desktop Session Host becomes increasingly unstable beyond 75-85GB RAM - WS2022

Masterchief07 106

Hey everyone! So, I have an RDS server set up with the following specs:

Proliant DL360 G10 w/ vSphere 7.03 Hypervisor
Win Server 2022 VM
48 CPUs Assigned (from 2 x XEON Platinum 6164s)
160GB RAM Assigned
1TB of Sata SSD Storage Assigned (from a RAID10 array of 6x2TB SSDs)
10GBit Ethernet between Exchange/DC/FMS/RDS/etc servers

We have about 70 users remoting into this server running Word, Excel, Outlook, Acrobat DC, Filemaker Pro Client, Firefox, and some JAVA applets. Supporting applications are Kaspersky Endpoint, Adguard, OpenShell. Nothing too crazy. In fact, this same deployment worked perfectly fine on Server 2012R2, and with less RAM and CPU allocation.

So what's baffling is that this deployment becomes incredibly unstable as RAM usage surpasses 75-85GB and things start to break around the 100GB mark. And by break, I mean the following:

-Right clicking stuff causes Explorer.exe to hang
-Kaspersky Endpoint's GUI crashes
-User profile service can get stuck leaving logins/logouts in limbo
-Windows Audio service will hang
-Constant icon refreshes
-General/intermittent application hangs

Page file is system managed, though I did try bumping it up artificially. No real difference. Most of the profile data is streaming from the DC, except for Documents and Appdata due to a limitation of the Filemaker program we use. So those are roaming. I have a bunch of the performance tweaks in play via GPO, like limited local redirections, display experiences, Office add-ins, etc etc. Time is perfectly synced across all servers as is DNS.

An interesting note too, if I log off a few folks to bring down RAM consumption, everything stabilizes even after shit hits the fan.

So yeah, maybe I am missing something? I know I can go the multi-session host route with a broker to evenly distribute users amongst smaller scale VMs, but I'm convinced that this can be rectified because it was working on older software with less resources.

Anyways, any insight would be appreciated. Thank you all in advance!

JodyD42 1 Reputation point

2022-09-27T17:43:36.227+00:00

HP ProLiant 380 G10
Win Server 2022 using Hyper-V
8 CPU Assigned from 2x Intel Xeon Gold 5218
Currently at 96GB RAM from 384 GB of Host system RAM
HP Smart Array P408i-a with 8x960GB SAS SSD (mixed use) R5 - 500 GB VHDX
10GB Aruba 3810 Switch between all servers

Just curious, did you ever find a solution for this?
I'm not all the way through the tuning, but am experiencing a high CPU usage already with only 7 users. My previous TS box on Server 2016 had no issues. For me it looks primarily like Windows Explorer is chewing up all the resources.
Masterchief07 106 Reputation points

2022-09-27T18:33:33.38+00:00

Actually, yes! Found the solution here: https://www.reddit.com/r/sysadmin/comments/un1i2z/windows_rds_server_2022_taskbar_and_start/

We implemented both solutions in that thread, but it was the TIP TestResults that was the culprit.

"Got familiar issue with freeze of the taskbar on user side. Clearing out HKLM\SYSTEM\Software\Microsoft\TIP\TestResults\ resolved that issue for us. I hope it will help you ;)"

https://learn.microsoft.com/en-us/answers/questions/752491/server-2022-and-teams-bloats-hkey-local-machinesys.html

1 answer

Your answer

JodyD42 1 Reputation point

2022-09-27T17:43:36.227+00:00

HP ProLiant 380 G10
Win Server 2022 using Hyper-V
8 CPU Assigned from 2x Intel Xeon Gold 5218
Currently at 96GB RAM from 384 GB of Host system RAM
HP Smart Array P408i-a with 8x960GB SAS SSD (mixed use) R5 - 500 GB VHDX
10GB Aruba 3810 Switch between all servers

Just curious, did you ever find a solution for this?
I'm not all the way through the tuning, but am experiencing a high CPU usage already with only 7 users. My previous TS box on Server 2016 had no issues. For me it looks primarily like Windows Explorer is chewing up all the resources.
Masterchief07 106 Reputation points

2022-09-27T18:33:33.38+00:00

Actually, yes! Found the solution here: https://www.reddit.com/r/sysadmin/comments/un1i2z/windows_rds_server_2022_taskbar_and_start/

We implemented both solutions in that thread, but it was the TIP TestResults that was the culprit.

"Got familiar issue with freeze of the taskbar on user side. Clearing out HKLM\SYSTEM\Software\Microsoft\TIP\TestResults\ resolved that issue for us. I hope it will help you ;)"

https://learn.microsoft.com/en-us/answers/questions/752491/server-2022-and-teams-bloats-hkey-local-machinesys.html

Answer 1

Hi there,

In context does older software means older versions of Windows and do you notice after any recent updates?

There were some reports of the latest updates breaking up the RDP and causing the issue, make sure you are not a victim of this by uninstalling any latest updates and checking the results.

You can also consider tuning the Remote Desktop Session Hosts. Most of the CPU usage on an RD Session Host server is driven by apps. Desktop apps are usually optimized toward responsiveness with the goal of minimizing how long it takes an application to respond to a user request.

Performance Tuning Remote Desktop Session Hosts https://learn.microsoft.com/en-us/windows-server/administration/performance-tuning/role/remote-desktop/session-hosts

I would also suggest using Process Monitor to identify the process behind the high RAM usage.
Process Monitor is an advanced monitoring tool for Windows that shows real-time file system, Registry, and process/thread activity.

You can get the tool from here https://learn.microsoft.com/en-us/sysinternals/downloads/procmon

--------------------------------------------------------------------------------------------------------------------------------

--If the reply is helpful, please Upvote and Accept it as an answer–

Masterchief07 106 Reputation points

2022-04-28T18:00:49.42+00:00

Hey, thanks for the reply! So when I said "older software" I was referencing the Server 2012 R2 instance that was the prior terminal server. This very same configuration was far more stable on that OS rather than the Server 2022 instance that we are on now.

As for the updates and performance tuning, I did see some folks discussing faulty updates. I've been able to rule that out. RDP is fine with current update set and relatively stable until we hit higher RAM utilization. I've tweaked as many things as I can regarding VM config, GPO, connection settings...at least according to those tuning articles.

And its not any one particular application that's causing the issue from what I can see. CPU and RAM usage per user is well within acceptable ranges (sub-3GB, averaging 2). Its the cumulative amount that is breaking things.

I did try running Process Monitor but like Process Explorer, it doesn't seem to handle so many concurrent things well. It was consuming 14GB of RAM itself to just render all of the activity and was barely responsive. Task Manager has been the only application so far to let me review performance details.

Share via

Remote Desktop Session Host becomes increasingly unstable beyond 75-85GB RAM - WS2022

1 answer

Your answer