Working with containers

Singularity tutorial 1

💬 In this tutorial, we will get familiar with the basic usage of Singularity containers.

Getting started

Download a test container image:

wget https://a3s.fi/saren-2001659-pub/tutorial.sif
ls -lh tutorial.sif

The file we downloaded is a container image. It contains all the software and data of the container in a single file. In this case, the container is very bare-bones and thus quite small, about 40 MB.
- Actual application containers are typically larger since they also contain the software installation and may in some cases include reference data, etc.

Basic usage

💬 There are three basic ways to run software in a Singularity container:

`singularity exec`

To execute a command inside the container, use singularity exec:
```
singularity exec tutorial.sif hello_world
```
Compare the outputs of the following commands:
```
grep "^NAME" /etc/os-release
singularity exec tutorial.sif grep "^NAME" /etc/os-release
```
- The first command is run on the host, the second command is run inside the container.
💭 The tutorial container is based on Ubuntu 18.04. The host and the container use the same kernel, but the rest of the system can vary.
- This means that a container can be based on a different Linux distribution than the host (as long as they are kernel-compatible), but it can’t run a totally different OS, such as Windows or macOS.

In batch jobs

💡 singularity exec is the run method you would typically use in batch job scripts.

Create a file called test.sh:
```
nano test.sh
```

Copy the following contents into the file and replace <project> with your actual LUMI project, e.g. project_465001234:

#!/bin/bash
#SBATCH --job-name=test           # Name of the job visible in the queue.
#SBATCH --account=<project>       # Choose the billing project. Has to be defined!
#SBATCH --partition=debug         # Job queues: https://docs.lumi-supercomputer.eu/runjobs/scheduled-jobs/partitions/
#SBATCH --time=00:01:00           # Maximum duration of the job. Max: depends of the partition. 
#SBATCH --mem=1G                  # How much RAM is reserved for job per node.
#SBATCH --ntasks=1                # Number of tasks. Max: depends on partition.
#SBATCH --cpus-per-task=1         # How many processors work on one task. Max: Number of CPUs per node.

singularity exec tutorial.sif hello_world

Submit the job to the queue with:
```
sbatch test.sh
```

💡 For more information about batch jobs, see the batch jobs section.

`singularity run`

💬 When containers are created, a standard action called the runscript is defined. Depending on the container, it may simply print out a message, or it may launch a program or service inside the container.

💭 If you are using a container created by someone else, you will need to check the documentation provided by the creator for details.

In our test container the runscript prints out a simple message:
```
singularity run tutorial.sif
```
Give the container image execution rights so that you can run it directly:
```
chmod u+x tutorial.sif
./tutorial.sif
```

You can see the actual script with the command:

singularity inspect --runscript tutorial.sif

`singularity shell`

Open a shell inside the container:
```
singularity shell tutorial.sif
```
Notice that the command prompt changed. You can now run any software inside the container interactively:
```
hello_world
```
Exit the container with:
```
exit
```

More information

💬 This tutorial is meant as a brief introduction to get you started.

💡 For more detailed instructions, see the official Singularity documentation.