When using interactive GUI applications via SSH with X11 forwarding, input delays resulting from network latency are often an issue. For the initial connection from your machine to the HPC frontends, X2Go - everywhere@home is a possible alternative to remedy this situation, since it uses the NoMachine NX technology.
We first describe how to set up X2Go on your system and afterwards how to prepare your login session for using the batch system.
The first step is to visit Installing the Qt-based X2Go Client and download the appropriate client for your platform (Win/Mac/Linux)
When setting up a session to one of the frontends gwdu101
, gwdu102
or gwdu103
, be sure to to put the complete domain name into the field Host
and your user name into Login
. Just like for raw SSH connections, key-based authentication is required. This can be done by specifying the location of your private key. An alternative is to just activate the checkbox “Try auto login” in order to acquire a key from the SSH agent or to use the default one.
Because the default use case for X2Go is to start a desktop environment, which is not installed on the HPC frontends, change the Session type
to Single application and type xterm
as the run command.
In order to ensure the correct shell environment for the batch system, enter the following in the xterm session
source /etc/profile.d/slurm.sh
To make the setting permanent, just add this line it to the file ~/.bashrc
.
You can now continue by submitting an interactive job with X-Forwarding from the frontend to the compute nodes as usual.
The X2Go server is installed on the HPC frontends but not on the login server login.gwdg.de
. If you're not connecting from within GöNET, please establish a VPN connection first.
— ckoehle2 2022/06/17 19:20