Using Singularity

Singularity is a containerization solution designed for high-performance computing cluster environments.  It allows a user on an HPC resource to run an application using a different operating system than the one provided by the cluster.  For example, the application may require Ubuntu but the cluster OS is CentOS.  Conceptually, it is similar to other container software such as Docker, but is designed with several important differences that make it more suited for HPC environments.  

  • Encapsulation of the environment
  • Containers are image based
  • No user contextual changes or root escalation allowed
  • No root owned daemon processes

To use Singularity on HCC machines, first load the singularity module. Singularity provides a few different ways to access the container. Most common is to use the exec command to run a specific command within the container; alternatively, the shell command is used to launch a bash shell and work interactively.  Both commands take the source of the image to run as the first argument.  The exec command takes an additional argument for the command within the container to run.  Singularity can run images from a variety of sources, including both a flat image file or a Docker image from Docker Hub.  For convenience, HCC provides a set of images on Docker Hub known to work on HCC resources.  Finally, pass any arguments for the program itself in the same manner as you would if running it directly.  For example, the Spades Assembler software is run using the Docker image unlhcc/spades and via the command To run the software using Singularity, the commands are:

Run Spades using Singularity
module load singularity
singularity exec docker://unlhcc/spades <spades arguments>

Using Singularity in a SLURM job is the same as any other software.

Example Singularity SLURM script
#SBATCH --time=03:15:00          # Run time in hh:mm:ss
#SBATCH --mem-per-cpu=4096       # Maximum memory required per CPU (in megabytes)
#SBATCH --job-name=singularity-test
#SBATCH --error=/work/[groupname]/[username]/job.%J.err
#SBATCH --output=/work/[groupname]/[username]/job.%J.out

module load singularity
singularity exec docker://unlhcc/spades <spades arguments>

Available Images

The following table lists the currently available images and the command to run the software.

If you would like to request an image to be added, please fill out the HCC Software Request Form and indicate you would like to use Singularity.

Software Version Command to Run Additional Notes
DREAM3D 6.3.29, 6.5.36 singularity exec docker://unlhcc/dream3d PipelineRunner
Spades 3.11.0 singularity exec docker://unlhcc/spades
Macaulay2 1.9.2 singularity exec docker://unlhcc/macaulay2 M2
CUDA (Ubuntu) 9.2 singularity exec docker://unlhcc/cuda-ubuntu <my CUDA program> Ubuntu 16.04.1 LTS w/CUDA Toolkit
TensorFlow GPU 1.4, 1.12 singularity exec docker://unlhcc/tensorflow-gpu python /path/to/ Use python3 for Python3 code
Keras w/Tensorflow GPU backend 2.0.4, 2.1.5, 2.2.4 singularity exec docker://unlhcc/keras-tensorflow-gpu python /path/to/ Use python3 for Python3 code
Octave 4.2.1 singularity exec docker://unlhcc/octave octave
Sonnet GPU 1.13, 1.27 singularity exec docker://unlhcc/sonnet-gpu python /path/to/ Use python3 for Python3 code
Neurodocker w/ANTs 2.2.0 singularity exec docker://unlhcc/neurodocker-ants <ants script> Replace <ants script> with the desired ANTs program
GNU Radio 3.7.11 singularity exec docker://unlhcc/gnuradio python /path/to/ Replace python /path/to/ with other GNU Radio commands to run
Neurodocker w/AFNI 17.3.00 singularity exec docker://unlhcc/neurodocker-afni <AFNI program> Replace <AFNI program> with the desired AFNI program
Neurodocker w/FreeSurfer 6.0.0 singularity run -B <path to your FS license>:/opt/freesurfer/license.txt docker://unlhcc/neurodocker-freesurfer recon-all Substitute <path to your FS license> with the full path to your particular FS license file. Replace recon-all with other FreeSurfer commands to run.
fMRIprep 1.0.7 singularity exec docker://unlhcc/fmriprep fmriprep
ndmg 0.0.50 singularity exec docker://unlhcc/ndmg ndmg_bids
NIPYPE (Python2) 1.0.0 singularity exec docker://unlhcc/nipype-py27 <NIPYPE program> Replace <NIPYPE program> with the desired NIPYPE program
NIPYPE (Python3) 1.0.0 singularity exec docker://unlhcc/nipype-py36 <NIPYPE program> Replace <NIPYPE program> with the desired NIPYPE program
DPARSF 4.3.12 singularity exec docker://unlhcc/dparsf <DPARSF program> Replace <DPARSF program> with the desired DPARSF program
Caffe GPU 1.0 singularity exec docker://unlhcc/caffe-gpu caffe Use python3 for Python interface
ENet Caffe GPU 427a014 singularity exec docker://unlhcc/enet-caffe-gpu <ENET program> Replace <ENET program> with the desired ENET program
ROS Kinetic 1.3.1 singularity exec docker://unlhcc/ros-kinetic <ROS program> Replace <ROS program> with the desired ROS program
Mitsuba 1.5.0 singularity exec docker://unlhcc/mitsuba mitsuba
FImpute 2.2 singularity exec docker://unlhcc/fimpute FImpute <control file> Replace <control file> with the control file you have prepared
Neurodocker w/FSL 5.0.11 singularity run docker://unlhcc/neurodocker-fsl <FSL program> Replace <FSL program> with the desired FSL program. This image includes GPU support.
gdc-client 1.4.0 singularity exec docker://unlhcc/gdc-client gdc-client <sub-command> Replace <sub-command> with the desired gdc-client sub-command
BLUPF90 1.0 singularity exec docker://unlhcc/blupf90 <command> Replace <command> with any command from the suite (blupf90, renumf90, etc.)
RMark 2.2.5 singularity exec docker://unlhcc/rmark Rscript my_r_script.r
SURPI 1.0.18 singularity exec docker://unlhcc/surpi -f </path/to/config> Replace </path/to/config> with the full path to your config file
PyTorch 1.0.1 singularity exec docker://unlhcc/pytorch python /path/to/ Use python3 for Python3 code. This image includes both CPU and GPU support.
bioBakery 1.1 singularity exec docker://unlhcc/biobakery <bioBakery program> Replace <bioBakery program> with the desired bioBakery program and its arguments

What if I need other Python packages not in the image?

An alternative to the steps below is to create your own custom image as described farther down. Start with an HCC-provided image as the base for your Dockerfile (i.e. FROM unlhcc/spades) and add any additional packages you desire.

Unfortunately it’s not possible to create one image that has every available Python package installed for logistical reasons.  Images are created with a small set of the most commonly-used scientific packages, but you may need others.  If so, you can install them in a location in your $WORK directory and set the PYTHONPATH variable to that location in your submit script.  The extra packages will then be “seen” by the Python interpreter within the image.  To ensure the packages will work, the install must be done from within the container via the singularity shell command.  For example, suppose you are using the tensorflow-gpu image and need the packages nibabel and tables.  First, run an interactive SLURM job to get a shell on a worker node.

Run an interactive SLURM job
srun --pty --mem=4gb --qos=short $SHELL

After the job starts, the prompt will change to indicate you’re on a worker node.  Next, start an interactive session in the container.

Start a shell in the container
module load singularity
singularity shell docker://unlhcc/tensorflow-gpu

This may take a few minutes to start.  Again, the prompt will change and begin with Singularity to indicate you’re within the container.

Next, install the needed packages via pip to a location somewhere in your work directory.  For example, $WORK/tf-gpu-pkgs.  (If you are using Python 3, use pip3 instead of pip).

Install needed Python packages with pip
export LC_ALL=C
pip install --system --target=$WORK/tf-gpu-pkgs --install-option="--install-scripts=$WORK/tf-gpu-pkgs/bin" nibabel tables

You should see some progress indicators, and a “Successfully installed..." message at the end.  Exit both the container and the interactive SLURM job by typing exit twice.  The above steps only need to be done once per each image you need additional packages for.   Be sure to use a separate location for each image’s extra packages.

To make the packages visible within the container, you’ll need to add a line to the submit script used for your Singularity job.  Before the lines to load the singularity module and run the script, add a line setting the PYTHONPATH variable to the $WORK/tf-gpu-pkgs directory. For example,

Example SLURM script
#SBATCH --time=03:15:00          # Run time in hh:mm:ss
#SBATCH --mem-per-cpu=4096       # Maximum memory required per CPU (in megabytes)
#SBATCH --job-name=singularity-test
#SBATCH --partition=gpu
#SBATCH --gres=gpu
#SBATCH --error=/work/[groupname]/[username]/job.%J.err
#SBATCH --output=/work/[groupname]/[username]/job.%J.out
export PYTHONPATH=$WORK/tf-gpu-pkgs
module load singularity
singularity exec docker://unlhcc/tensorflow-gpu python /path/to/

The additional packages should then be available for use by your Python code running within the container.

What if I need a specific software version of the Singularity image?

You can see all the available versions of the software built with Singularity in the table above. If you don’t specify a specific sofware version, Singulariy will use the latest one. If you want to use a specific version instead, you can append the version number from the table to the image. For example, if you want to use the Singularity image for Spades version 3.11.0, run:

singularity exec docker://unlhcc/spades:3.11.0

What if I want to build a custom image to use on the HCC clusters?

You can create custom Docker image and use it with Singularity on our clusters. Singularity can run images directly from Docker Hub, so you don’t need to upload anything to HCC. For this purpose, you just need to have a Docker Hub account and upload your image there. Then, if you want to run the command “mycommand“ from the image “myimage”, type:

module load singularity
singularity exec docker://myaccount/myimage mycommand

where “myaccount” is your Docker Hub account.

In case you see the error ERROR MANIFEST_INVALID: manifest invalid when running the command above, try:

module load singularity
singularity exec docker://myaccount/myimage mycommand

If you get the error FATAL: kernel too old when using your Singularity image on the HCC clusters, that means the glibc version in your image is too new for the kernel on the cluster. One way to solve this is to use lower version of your base image (for example, if you have used Ubuntu:18.04 please use Ubuntu:16.04 instead).

All the Dockerfiles of the images we host on HCC are publicly available here You can use them as an example when creating your own image. The only thing you need to note when creating custom Docker images you want to use on HCC is to add the line:

RUN mkdir -p /work
at the end of your Dockerfile. This creates a /work directory inside your image so your $WORK directory on Crane/Tusker is available.