Introduction

This repository collects interactive applications that run on compute nodes using Open OnDemand. All applications are executed using Singularity, a Linux container runtime. The Singularity image OS is Rocky Linux 9.7, and the supported architectures are x86_64 and aarch64.

In this repository, interactive applications are categorized into "Visualization Applications" and "Development Applications". The applications currently provided are listed below.

Visualization Applications

Name	x86_64	aarch64	Description
Gnuplot	5.4.3	5.4.3	Command-line driven graphing program
ParaView	6.0.1	5.11.1	Scientific and technical data visualization program
XCrySDen	1.6.3	1.6.3	Visualization program for crystal and molecular structures
PyMOL	2.5.0	2.5.0	Visualization and analysis program for 3D structures of biomacromolecules
GrADS	2.2.3	-	Visualization and analysis program for gridded data in meteorology and climate fields
VisIt	3.4.2	-	Visualization and analysis program supporting various scientific data formats
VESTA	3.5.8	-	Visualization program for crystal structures and 3D data such as electron/nuclear densities
Smokeview	6.10.6	-	Visualization program for displaying results from FDS and CFAST
OVITO	3.14.1	-	Visualization and analysis program for large datasets such as particle simulations
ImageJ	1.54	-	Image processing software running on the Java Virtual Machine

Development Applications

Name	x86_64	aarch64	Description
Desktop (Xfce)	4.18	4.18	Lightweight desktop environment running on the X Window System
VSCode	4.108.2	4.108.2	Code editor developed by Microsoft
JupyterLab	4.5.3	4.5.3	Interactive execution environment for programs running in a web browser
ttyd	1.7.7	1.7.7	Command-line tool that enables terminal access through a web browser

Usage

Download the Repository

Save this repository under /var/www/ood/apps/sys/.

$ cd /var/www/ood/apps/sys
$ sudo git clone https://github.com/OpenOnDemandJP/oodjp_apps.git

Build Singularity Images

Use containers/rocky97_x86_64.def if the compute node architecture is x86_64, or containers/rocky97_aarch64.def if it is aarch64. Commands are shown below. Build time is about 20 minutes, depending on server and network performance.

For x86_64

$ singularity build --fakeroot rocky97_x86_64.sif rocky97_x86_64.def

For aarch64

$ singularity build --fakeroot rocky97_aarch64.sif rocky97_aarch64.def

Edit the Configuration Files

Edit /var/www/ood/apps/sys/oodjp_apps/utils/config.yml to match your environment. Settings for unused architectures are ignored, so you do not need to modify them.

• xdg_data_home: Directory for storing application data
• container_image: Path to the Singularity image

xdg_data_home:
  x86_64:  "<%= ENV['HOME'] %>/ondemand/x86_64"
  aarch64: "<%= ENV['HOME'] %>/ondemand/aarch64"

container_image:
  x86_64:  /cloud_opt/ondemand/rocky97_x86_64.sif
  aarch64: /cloud_opt/ondemand/rocky97_aarch64.sif

Register Applications in Open OnDemand

Create a symbolic link under /var/www/ood/apps/sys/ for each interactive application you want to use. The example below is for Desktop.

$ cd /var/www/ood/apps/sys/
$ sudo ln -s oodjp_apps/apps/Desktop .

Next, edit the application configuration files. For syntax details, refer to the Open OnDemand manuals linked below.

• Desktop/manifest.yml: Manual
• Desktop/form.yml.erb: Manual
• Desktop/submit.yml.erb: Manual

Verification

Use the data in sample_images/. For details, see sample_images/README.md.

Note

• Keep in mind that the development environments inside the Singularity container and on the host are different. This is especially important when using development applications. It is more convenient if the development environments in the container and on the host are aligned. To do that, in addition to proper bind settings for directories, you need to pass host environment variables, but even then it is difficult to make them completely identical. When possible, we recommend installing development applications on the host and using them from Open OnDemand without Singularity.
• The aarch64 Singularity container uses SBSA (Server Base System Architecture) nvidia-driver-libs. If you want to use this container on a non-SBSA server, create a container that does not use nvidia-driver-libs.
• The original ttyd sets one-time passwords from the command line, so depending on your environment, the password may be visible using the ps command. To avoid this, we forked and used the new ttyd, which allows you to set one-time passwords via environment variables.