Ubuntu / gdm3 / headless crashes

[behinger]

– I am running NoMachine Terminal Server (6.12.3) on a headless server (at least no monitor connected) with an dedicated nvidia 1050 graphics card, Ubuntu 20, gnome with gdm3 (but lightdm doesnt change the error)

– I initially got semi-random crashes in different applications (rstudio, matlab) when running them individually under virtualgl.  I think I simplified the problem and can get the same crashes only with glxgears and no virtualgl:

– I have to maximize/minimize glxgears several times. At one point, I will shortly (less than a sec to a second) see the purple ubuntu-login screen color, the program continues for some seconds and then I get kicked out of the NoMachine session.

– The identical problem happens with other software, e.g. matlab, or with virtualgl. Under virtualgl it tends to happen faster.

– If I log-in again, my zsh shell reports the following:
> 140 frames in 5.7 seconds = 24.529 FPS
> X connection to :1002 broken (explicit kill or server shutdown).

– I don’t have an entry in .xsession-errors

– I have a debug file for the virtualgl case ready

– If I activate virtualgl for all applications, then I cannot login through the nomachine interface – it immediately crashes.
– nvidia-smi reports:  NVIDIA-SMI 455.32.00    Driver Version: 455.32.00    CUDA Version: 11.1

I hope you can help me fix this! It makes working on the server impossible.

[Mth]

Hello.

First some questions.

1. You say it’s headless machine (without monitor) but there is gdm3/gnome installed. Does the desktop manager run and NoMachine connects to it, or does NoMachine create virtual session? If during connection you have a question asking to create desktop or if in server.cfg you have

CreateDisplay key uncommented and set to 1 – it is virtual one.

2. Is there any signs of nxnode/nxserver processes crashing in the system? By default there could
be a report in /var/crash directory. If that’s the case please provide us with backtrace, eg.

apport-retrace --output=nxnode.retraced /var/crash/_usr_NX_bin_nxnode.bin.1000.crash

Also please provide us with the content of /usr/NX/var/log directory from server machine.

Please send the logs to forum[at]nomachine[dot]com using the title of this forum’s thread as the mail’s subject.

/Mth

[behinger]

Hi!
1. CreateDisplay is set to 0 and commented. I was writing headless, because there is no physical monitor/keyboard etc. connected (as written I do have a dedicated graphics card).

2. No crash reports of nxnode/nxserver. If matlab/rstudio crashes I sometimes get a crash report here, if the whole session drops, I dont get one.
Today I was looking at journalctl and found these errors:
——————————————————————————————————————————————
Nov 16 12:39:02 ccs-srv-001 gnome-shell[743676]: NOTE: Not using GLX TFP!
Nov 16 12:39:02 ccs-srv-001 org.gnome.Shell.desktop[743676]: GNOME Shell crashed with signal 11
Nov 16 12:39:02 ccs-srv-001 org.gnome.Shell.desktop[743676]: == Stack trace for context 0x557ab880c5d0 ==
Nov 16 12:39:02 ccs-srv-001 gnome-session[727116]: gnome-session-binary[727116]: WARNING: Application ‘org.gnome.Shell.desktop’ killed by signal 11
Nov 16 12:39:02 ccs-srv-001 gnome-session[727116]: gnome-session-binary[727116]: WARNING: App ‘org.gnome.Shell.desktop’ respawning too quickly
Nov 16 12:39:02 ccs-srv-001 gnome-session[727116]: gnome-session-binary[727116]: CRITICAL: We failed, but the fail whale is dead. Sorry….
Nov 16 12:39:02 ccs-srv-001 gnome-session-binary[727116]: Unrecoverable failure in required component org.gnome.Shell.desktop
Nov 16 12:39:02 ccs-srv-001 gnome-session-binary[727116]: WARNING: Application ‘org.gnome.Shell.desktop’ killed by signal 11
Nov 16 12:39:02 ccs-srv-001 gnome-session-binary[727116]: WARNING: App ‘org.gnome.Shell.desktop’ respawning too quickly
Nov 16 12:39:02 ccs-srv-001 gnome-session-binary[727116]: CRITICAL: We failed, but the fail whale is dead. Sorry….
——————————————————————————————————————————————

I get the “”NOTE: Not using GLX TFP!”” error more regularly. Google told me to reinstall gnome-shell which I did (did not reboot because calculations are running though).

I will send the log-files.

[behinger]

I resolved my issues. a) I installed XFCE instead of Gnome, that resolved the issues. b) I then had troubles with virtualGL (graphics card was not found). After purging & reinstalling the nvidia drivers it works now, also in Gnome-shell.

 

Thanks for your help!

[Author]

Posts