Horizon 2206 Session Disconnects and Clone VM High CPU Usage

[mbrundage]

I have a small set of users in separate desktops pools running into an issue where their instant clone VM for their session is always at 100% CPU usage in VSphere and at about the 6 hour mark of usage each day they get disconnected from the VM. We've already tried a entirely new golden image and new thin clients, but the issue still persists. There are other users in the same pool that don't have this issue at all. Any suggestions where to look next?

[Digory34]

Hi,

do you used instant clone or linked clone? i understand that the issue is affecting only some user randomly.

What is the protocol used ( RDP, blast or pcopip)?

You can check from Horizon Agent log and Horizon client log what can be the cause of the issue.

You should see the disconnection from this log (especially agent log).

Are you able to reconnect after that? what is the message when you get disconnected? 

What is your VM guest os?

[mbrundage]

Hi Digory34, 

We use instant clones with blast. I reviewed the logs and added them below. Users are able to re-connect to their sessions after they are disconnected and we are using server 2019 standard for our VM guest os. 

16:06:08 0x00000bd4 WARNING client protocol is not available, set session 1 as not connected
16:06:09 0x00000bd4 WARNING client protocol is not available, set session 1 as not connected
16:06:09 0x00000bd4 INFO Added current session id 1 to env.
16:06:09 0x00000bd4 INFO Session 1 has disconnected.
16:06:09 0x00000bd4 INFO Add message to queue, message type: 5
16:06:09 0x00000bd0 INFO Data with type eUserSessionDisconnect is sent with response - 0x00000000 -- The operation completed successfully.

16:08:24 0x00000bd4 INFO Added current session id 1 to env.
16:08:25 0x00000bd4 INFO Session 1 has reconnected.
16:08:25 0x00000bd4 INFO Add message to queue, message type: 4
16:08:26 0x00000bd0 INFO Data with type eUserSessionReconnect is sent with response - 0x00000000 -- The operation completed successfully.

[Digory34]

Hi Mbrundage,

Seems that you may have an issue with blast protocol.

16:06:08 0x00000bd4 WARNING client protocol is not available, set session 1 as not connected
16:06:09 0x00000bd4 WARNING client protocol is not available, set session 1 as not connected

We should have a log for that.

VMware Knowledge Base

Blast logs
The virtual desktop Blast logs are Blast-Service.log and Blast-Worker-SessionIdX.log and they are located under <Drive>:\ProgramData\VMware\VMware Blast.
The Client logs are called vmware-mks-XXXXX.log and are located C:\Users\%username%\AppData\Local\Temp\vmware-username-XXXXXX.
Log File Locations for Client and Agent Components
<Drive>:\ProgramData\VMware\VMware Blast\ This directory contains logs for the three Blast Extreme components: Blast-Service. log. 
Techzone article - Log File Locations for Client and Agent Components

[mbrundage]

So I found the follow section in the blast log that looks a bit off compared to the rest of the log file: 

2022-08-30 16:06:07.064-0400 [WARN ] 0x093c bora:⚠️ VvcQueueChannelMessagesToSendTree: AsockXBe is down, not queueing messages on channel 15
2022-08-30 16:06:07.117-0400 [INFO ] 0x2124 bora::Log: VNCSERVER-WIN32: Destroying capture backend (WDDM) for \\.\DISPLAY1 completed
2022-08-30 16:06:07.351-0400 [INFO ] 0x5d28 bora::Log: VNCSERVER-WIN32: No capture finished posted as captured is not requested by screen 1. Current mask is 0
2022-08-30 16:06:07.351-0400 [INFO ] 0x2124 bora::Log: VNCSERVER-WIN32: Destroying capture backend (WDDM) for \\.\DISPLAY2 completed
2022-08-30 16:06:07.351-0400 [INFO ] 0x2124 bora::Log: VNCConnectionMgr: Unsubscribing capture type 0
2022-08-30 16:06:08.357-0400 [INFO ] 0x0ecc bora::Log: VNCSERVER-WIN32: Received WM_POWERBROADCAST message.
2022-08-30 16:06:08.357-0400 [INFO ] 0x0ecc bora::Log: VNCSERVER-WIN32: Powering off monitor
2022-08-30 16:06:08.395-0400 [WARN ] 0x2124 bora:⚠️ VNCSERVER-WIN32: SetThreadDesktop failed: The requested resource is in use (0xaa)
2022-08-30 16:06:08.670-0400 [WARN ] 0x6190 bora:⚠️ VvcQueueChannelMessagesToSendTree: AsockXBe is down, not queueing messages on channel 41

Let me know if it would be beneficial to upload the entire log file. I will just have to redact some information from it if so. 

[Digory34]

Yes you can upload Agent log and blast log.

Very interesting is that VNCSERVER-WIN32: Received WM_POWERBROADCAST message seems to be a windows event related :

WM_POWERBROADCAST Messages - Win32 apps | Microsoft Docs

System broadcasts a message to all applications and installable drivers whenever a power management event occurs.

If you check the Windows Event log, did you see something relevant?

Do you have an UAG? Do you encounter any error during the logoff like "Logout requested by system" ?