Containers and Workflows in Bioinformatics

Converting Docker image to Singularity

We converted a Docker image to singularity one using singularity build subcommand in previous episodes. And also, we can use singularity pull subcommand to achieve the same outcome. As an example for the image conversion, we will use a biconda package named, trimmomatic software which is available as docker image in Quay container registry. Trimmomatic is a flexible read trimming tool for Illumina NGS data.

Expected outcome of this tutorial:

Upon completion of this tutorial, you will be able to:

Launch a singularity container from a bioconda package which available as a docker image (e.g., trimmomatic image)
Manage Singularity TMPDIR and CACHEDIR directories in HPC environment

Launching trimmomatic software as a singularity container

Launch interactive session on Puhti as below:

# start interactive node as below and choose your project name on prompt
sinteractive -c 2 -m 4G -d 100

Visit trimmomatic software webpage as available on bioconda channel. Look for the URI (Looks something like the following: quay.io/biocontainers/trimmomatic:tag) of trimmomatic image and incorporate the latest tag information for trimmomatic software. Once you have URI path with tag information, you can build a singularity image as below:
```
 singularity pull docker://quay.io/biocontainers/trimmomatic:0.32--hdfd78af_4
```
Upon successful completion of above command, you will be able to see trimmomatic singulairty image (name: trimmomatic_0.32–hdfd78af_4.sif) in the current directory.
when you convert multiple smaller images or a bigger image, the singularity cache can take up lot of space. Please note that the singularity cache directory is by default your HOME directory which is limted by disk space quota in CSC HPC environment. As a best practice tip, do not pull a Singularity image cache to your Puhti Home Directory. In order to avoid such cache issues which would result in disk space errors, we recommend resetting Singularity TMPDIR and CACHEDIR directories either to Lustre scratch or $LOCAL_SCRATCH before performing image conversion as below:
```
 export SINGULARITY_TMPDIR=/scratch/project_xxxx/$USER  (or $LOCAL_SCRATCH)
 export SINGULARITY_CACHEDIR=/scratch/project_xxxx/$USER (or $LOCAL_SCRATCH)
 unset XDG_RUNTIME_DIR
 singularity pull docker://quay.io/biocontainers/trimmomatic:0.32--hdfd78af_4
```
It is better to use LOCAL_SCRATCH directories as Singularity TMPDIR and CACHEDIR directories especially if the image sizes are quite big (in the order of GBs). Usually larger docker images are composed of several layers and importing and unpacking the layers into one single image file for singularity can be I/O intensive.

Launch trimmomatic container and check comamndline help

 singularity exec trimmomatic_0.32--hdfd78af_4.sif trimmomatic --help   # or simply ./trimmomatic_0.32--hdfd78af_4.sif

Few useful tips for managing singularity cache:

 singularity cache list  # You can check the disk space taken up by the image in cache folder
 singularity cache clean  # clean singularity cache

This site is open source. Improve this page.