Error is 104: connection reset by peer

[boxofrox]

Greetings. I’m having trouble connecting NoMachine to one of the desktops on my LAN.

The error reported on the client from nxplayer is:

The connection with the server was lost.
Error is 104: Connection reset by peer.

Just prior to the error, nxplayer briefly mentions an attempt to reconnect.

On the client, I’m running NoMachine 6.8.1 on an Arch Linux laptop. The remote side (nxserver) also runs NoMachine 6.8.1 on an ElementaryOS Juno desktop.

I upgraded NoMachine to 6.8.1 since I first experienced the issue with NoMachine 6.1 on both machines. I purged old settings from both machines: on the server side, I rm -rf /usr/Nx between uninstalling and reinstalling; on both machines, I rm -rf $HOME/.nx for the respective users. Error 104 still occurs 100%.

I attempted to use the SSH protocol instead of NX, but during the attempt to connect via password, nxplayer announced that SSH was not available and switched back to NX.

I dug through the logs and found little from the server side. There were some errors reported with pulseaudio, so I disabled AudioInterface in node.cfg.

On the client side, the logs mention a failure switching over to proxy. I’ve collected the client and server logs after reproducing this issue (it occurs 100%) per the instructions at https://www.nomachine.com/DT10O00163#1.1 and attached them to this post. Since the error occurs 100%, I cleared all the logs between restarting nxserver with systemctl stop nxserver && clear logs && systemctl start nxserver.

Oddly, I have a third machine (Arch Linux desktop) and my laptop connects to it just fine, and my Arch Linux desktop also reports Error 104 trying to connect to the ElementaryOS desktop, so I suspect the problem is between ElementaryOS and NoMachine, but I’m not sure what to look for.

Please advise.

Attachments:

1. nomachine-log-client.tar.gz
2. nomachine-log-server.tar.gz

[Mth]

Hello

From the logs you supplied it seems that the nxnode process that is accessing the physical desktop is segfaulting during the attach
procedure. This is causing the player to disconnect and trying to reconnect again, which with nxnode dead is causing a misleading
error “104: Connection reset by peer.”.

The nxnode process is dying with signal 11, so there should be a trace in the system logs or, if you have core dumps configured
a core file for investigation.

We are trying to reproduce this issue, but we would be grateful for a backtrace from this core file.

Please refer to the article https://www.nomachine.com/AR09L00810 on how to proceed with core file.

/Mth

[boxofrox]

I looked further, and found a crash file on the “server” side for nxnode.  I followed the instructions on https://www.nomachine.com/AR09L00810 to generate a backtrace after extracting the core dump from the apport utility.

The core dump is >300Mb, though I knocked that down to 4.3Mb with libxz.  I’ll attach the backtrace log for now.  If you also want the core dump, please instruct if you have a preferred method to attach large files.

[boxofrox]

Hmmm… let’s see if a tarball of the backtrace will attach.

Attachments:

1. nxnode-backtrace.tar.gz

[fra81]

Thank you for the backtrace. We are preparing a possible patch for you to test. Please write to forum[at]nomachine[dot]com so we can send it to you.

[boxofrox]

Britgirl emailed the patch to me and it works.  I can connect to the ElementaryOS desktop just fine again.

Thank you all for the help!

[Author]

Posts