UVNC server High CPU / Non responsive

[newmann]

Hi, I have an issue with uvnc server on windows 2016 on an aws ec2 instance
Currently using v1.3.4.1 x64

At a random point Uvnc will stop responding and become unuseable and the system process will shoot up to 50% usage. Stopping uvnc server service kills the high system process usage but if I restart the service the system usage spikes again and uvnc is unusable until I reboot the server. It will then be ok again for a random number of days.

Has anyone else seen this behaviour?

If someone could advise on what could be causing this and how to debug / generate log files to capture what is happening it would be appreciated.

It is not happening on all my ec2 instances that have uvnc installed. They all have the encryption plugin enabled.

Thanks in advance

Neil

[Rudi De Vos]

when the service is stopped
DebugMode=2
path=C:\temp\Debug
DebugLevel=10
If the debug files exist clear

Then start the service back and when at 50 %, and check what's in in the log file.

In the taskmgr you see 2x winvnc.exe
1x service ( session0) and running as system
1x desktop part running in the desktop session
Is the service or the desktop part using the cpu

[newmann]

Hi Rudi,

How do I identify which session is which ? using process explorer they both appear to have been spawned with the same details ?

Also it is the system process ( normally pid 4 ) that escalates to 50% not the vnc processes

Regards Neil

[Rudi De Vos]

Is it only when you connect the viewer again ? Or will starting the server also trigger it.

Try to play with options ( server properties)
[] system hookdll
[] DDengine
possible one of the 2 can trigger it.
This are the only two that heavy use the hardware/software to detect the screen changes.

Try when you have the 50%, to disable one and to restart the service.

[newmann]

Is it only when you connect the viewer again ? Or will starting the server also trigger it

I believe it is when I restart the service but I am not 100% sure - I will be a bit more methodical on the next occurance

I will also try the options you mentioned to try to narrow it down. I have enabled the debugging already.

Regards Neil

[newmann]

Hi Rudi, the original server hasn't failed since I turned the debug on. But I had another server exhibit the same behaviour.
On trying to uVNC to server, connection established and either grey screen no update or the locked screen and unable to issue Ctrl Alt Del and get logon screen

While testing was connected to server via RDP session, fully responsive
testing steps
1. While uvnc server stopped - system pid 4 normal levels of activity, less than 10% occasional spikes of activity. - Enabled the debug logging.
2. Started uvnc service, no noticeable performance issue, system pid 4 still low activity occasional spikes.
3. attempted to uvnc to server, connected but unresponsive screen. Seconds after uvnc connection system pid 4 process goes to 50% solid cpu utilisation, using procexp from sysinterals tools, one thread is consuming all the 50% utilisation of the system process, this is the dxgkrnl.sys - a directx driver.
4. stopped uvnc service, system pid 4 takes a couple of minutes before this thread drops to 0% utilisation and system process returns to normal usage (low ocassional spike)
5. Unticked the use systemhook dll as advised and restarted uvnc service. No performance issue as per step 2
6. repeated step 3 and 4 with the same results.
7. ticked the use systemhook dll and unticked the duplicate_desktop option as advised and restarted uvnc service. No performance issue as per step 2
8. repeated step 3 and 4 with the same results.
9. rebooted server
10. Uvnc now working again ok

Please let me know how to send you the debug logs and a process explorer screenshot of the offending thread