NxFrameBuffer error while starting

[JustAnotherLinuxGuy]

I did look through the previous posts, but I wasn’t able to find anything relevant to our setup. I sent over an email with some more info

[fisherman]

Upon reviewing our system configurations, I’ve noticed an issue with the permissions in the /rhome directory. Specifically, the home folder of DV9G49 appears to be owned by root with only root having write access.

Here’s a snapshot of the permissions:

root@workstation-analyst3:~# ls -la /rhome/
total 736
drwxr-xr-x 15 root root 4096 May 31 14:36 .
drwxr-xr-x 1 root root 4096 May 31 13:26 ..
drwxr-xr-x 2 root root 4096 May 23 16:03 DV9G49

However, upon further inspection, it seems that the correct permissions are applied in your home directory /home/DV9G49:

root@workstation-analyst3:~# ls -la /home/DV9G49
total 28
drwxr-xr-x 2 DV9G49 DV9G49 4096 Feb 27 14:52 .

To resolve this discrepancy, I recommend checking if your NFS mounting, if used, correctly applies permissions. If not, please ensure that the folder /rhome/DV9G49 is owned by the user DV9G49 and has write permissions for the same user.

Thank you for your attention to this matter. If you require any assistance or further clarification, please don’t hesitate to reach out.

[JustAnotherLinuxGuy]

Perhaps I haven’t been very clear…

We do not care to set a default home directory. The frame buffer issue was appearing before we had any sort of permission errors.

Furthermore, I have attempted to disable the configuration for setting a default home directory. Despite numerous attempts, the permission error still remains. Is there a setting I can adjust to prevent the system from making these changes? Please refer to my previous post where I noted the server.cfg change I made in regards to the default directory.

[fisherman]

I have never tried to use and setup gnome to be started using docker containers. I feel that your issue is related to the docker configuration and not at all to the NoMachine. I have tried installing gnome in my docker container and could not start a gnome-session unrelated to the NoMachine. I used xvfb-run, and the session was failing with something related to dbus, I did not explore and debug more.

While I can confirm that I have tested xfce and Mate and they work very well.

sudo apt install xfce4 xfce4-goodies

and then edit /usr/NX/etc/server.cfg and set that NoMachine will start xfce4
DefaultDesktopCommand /usr/bin/startxfce4

[JustAnotherLinuxGuy]

You need to set the default desktop to gnome-classic in the config file in order for it to launch properly. Or at least that’s the case for GNOME 42.

I cleaned up that VM and started over with a brand new install. I am still seeing a permission error, I think this is being generated because we have set NoMachine to create its own X11 display. Which requires CreateDisplay 1 and DisplayOwner "DV9G49". If I understand correctly, this causes a directory to be set up in /rhome/ but due to the permissions of this directory, NoMachine can’t and therefore we see an error.

However, this time it is not a frame buffer error as before. Can you take another look at the logs to see if there is something different occuring this time around?

[Britgirl]

Can you forward those logs to forum[at]nomachine[dot]com?

[fisherman]

Removing the NoMachine software from the picture entirely, we were NOT able to launch Debian Gnome desktop from a docker file. So, this is something that we cannot investigate further. Whilst, as we mentioned, XFCE runs fine.

Putting NoMachine back into the picture with XFCE, again everything works as expected. So you could possibly try with XFCE as well, as we recommended earlier.

However, the issue you are having with the /rhome/ folder is not clear to us. For NoMachine to work properly, the user which is connecting must have write permissions to the user’s specific home folder. We suspect that there is a system misconfiguration, but we cannot know where. So it might be that even with XFCE, you will not be successful because of the misconfiguration of user permissions.

We will wait for the logs in any case and check them.

[JustAnotherLinuxGuy]

This time I was able to resolve the permissions error. But I think we are running into the issue you were referring to with GNOME 46.

Using the gnome-classic command, enabled us to “launch” the VM but it errored out stating falling back to default session and then force closes. I changed the command to gnome-session and that gives us the option to “select a desktop” but none were listed. I have forwarded the logs to your email. Would you be so kind as to take a look and confirm if this is in fact due to using GNOME versus XFCE?

[fisherman]

It looks as if the application stderr is redirected so would be good if you could provide a .xsession-errors file from the user home /DV9G49/. In the logs, we can see that the application closes unexpectedly during startup.

Please also add localhost to the /etc/hosts file of the docker container as:
127.0.0.1 localhost

Check if starting the container with --cap-add=SYS_PTRACE can make any difference.

Then, please enable logs by executing sudo /etc/NX/nxserver --debug --enable all, reproduce problem and collect logs as well as including .xsession-errors file.

[JustAnotherLinuxGuy]

We don’t user docker run to start the VM; it’s managed by Cloud Workstations

So we use a gcloud command to start it, another gcloud command to start a tcp tunnel, then NoMachine to log into the VM.

We switched to XFCE as you suggested and we still run into an error. I have sent over the logs for review.

[Britgirl]

Logs show:
1945 2016 2024-06-06 21:05:11 234.255 TcpConnector/TcpConnector: ERROR! Error is 99 ‘Cannot assign requested address’.
Error: Connection to ‘localhost:25001’ protocol ‘TCP’ failed.

Did you add localhost to the /etc/hosts file as suggested by Fisherman? We will be checking installing NoMachine on GCP/Docker, also to make sure our current documentation is kept up-to-date (https://www.nomachine.com/accessing-your-remote-desktop-on-google-cloud-platform-via-nomachine) and extended.

[JustAnotherLinuxGuy]

Yes, localhost is in the /etc/hosts profile by default. I verified, and it is there.

[Britgirl]

Our initial tests show that it’s necessary to uncomment the “DefaultDesktopCommand” key in the node.cfg.

#For GNOME session:
DefaultDesktopCommand "env XDG_SESSION_TYPE=wayland dbus-run-session /etc/X11/Xsession gnome-session"

#For XFCE:
DefaultDesktopCommand "dbus-run-session startxfce4"

Maybe you could try these.

[JustAnotherLinuxGuy]

Thanks for the info, we have decided to pursue a different route but I will keep this in mind if we decide to come back to this option.

[CRCinAU]

Was there ever any kind of followup for this? In my current testing with the KDE Spin of Fedora, I’ve found the following:

1) F39 clean install from the Live ISO -> NoMachine works fine

2) After running updates on F39, NoMachine no longer works with a similar error

3) F40 doesn’t work from a fresh install, or after updates have been installed.

[Author]

Posts