Running a Singularity container

A container can be run with

singularity run mycontainer.img

which executes a pre-specified command in the container.

To get more flexibility, running

singularity shell mycontainer.img

will spawn a shell inside the containerised environment.

The exec Singularity sub-command allows you to spawn an arbitrary command within your container image as if it were running directly on the host system.

Sharing containers

Sharing a container with a collegue is as simple as giving them access to the container. This could mean copying the image to them or in a cluster environment, setting the permisions on the image file to let that user read the file.

Changing the contents of a container

By default, the container will be mounted as read-only. If you need to make changes to the container (e.g. install new software) then this can be acheived with

singularity shell --writeable mycontainer.img

which, as long as your user has permission to edit the mycontainer.img will let you make changes to the container.

In general, it will make most sense to construct your workflows so that the container image is not changed frequently.

A containerised application can read and write from files outside the container.

Accessing data outside the container

Data from outside the container can be accessed by mounting it inside the container with the --bind option. Running

singularity shell --bind ~/mydata:/scratch mycontainer.img

will make the directory ~/mydata outside the container appear inside the container at /scratch.