Using containers (Docker/Singularity) in science

Running containers on clusters and HPCs

Overview

Teaching: 5 min
Exercises: 0 min
Questions
  • How can I run my containers on clusters/HPCs if my admin refuse to install Docker?

Objectives
  • Learn how to use Singularity

Running containers on clusters and HPCs

Docker works great on modern operating systems in a single tenant scenarios. However, due to kerl feature requirements and the need for elevated security it is hard sell for administrators of academic clusters. All of the reasons for the deal breaker Docker requirements are for functionality (networking, resource sandboxing etc.) that are not crucial in context of academic use. A different container framework Singularity provides ability to run containers without the need for special kernel versions or security workarounds. It still needs to be installed on the cluster, but it’s much more popular among HPC admins (ask your admin if they could install it!).

To run your image on the cluster: singularity shell --cleanenv docker://filo/fsl_with_conda

Key Points