Skip to content


Repository files navigation



Are you developing Docker images and you would like to run them on an HPC cluster supporting Singularity? Are you working on Mac or Windows with no easy access to a Linux machine? If the pull, build, and general commands to work with docker images provided by Singularity natively do not fit your needs, docker2singularity is an alternative way to generate Singularity images. The containers are available to you on, and older versions also available for you on Docker Hub.


$ docker run
USAGE: docker2singularity [-m "/mount_point1 /mount_point2"] [options] docker_image_name

          Image Format
              --folder   -f   build development sandbox (folder)
              --option   -o   add a custom option to build (-o --fakeroot or -option 'section post' )
              --writable -w   non-production writable image (ext3)         
                              Default is squashfs (recommended) (deprecated)
              --name     -n   provide basename for the container (default based on URI)
              --mount    -m   provide list of custom mount points (in quotes!)
              --help     -h   show this help and exit


Image Format

  • squashfs (no arguments specified) gives you a squashfs (*.simg) image. This is a compressed, reliable, and read only format that is recommended for production images. Squashfs support was added to Singularity proper in January of 2017 and thus available as early as the 2.2.1 release.
  • sandbox (-f) builds your image into a sandbox folder. This is ideal for development, as it will produce a working image in a folder on your system.
  • ext3 (-w) builds an older format (ext3) image (*.img). This format is not recommended for production images as we have observed degradation of the images over time, and they tend to be upwards of 1.5x to 2x the size of squashfs.

Note that you are able to convert easily from a folder or ext3 image using Singularity 2.4. If your choice is to develop, making changes, and then finalize, this approach is not recommended - your changes are not recorded and thus the image not reproducible.

Mount Points

  • -m specify one or more mount points to create in the image.


If you look at singularity build --help there are a variety of options available. You can specify some custom option to the command using the --option flag. Make sure that each option that you specify is captured as a single string. E.g.,:

--option --fakeroot 
--option '--section post'

Image Name

The last argument (without a letter) is the name of the docker image, as you would specify to run with Docker (e.g., docker run ubuntu:latest)


If you want a legacy version, see the repository branches and tag history on the registry.

Containers were previous built on Docker Hub and now are provided on A tag with prefix v corresponds to a release of the Singularity software, while the others are in reference to releases of Docker. Previously used scripts, including environment and action files, are provided in this repository for reference.


  • Docker (native Linux or Docker for Mac or Docker for Windows) - to create the Singularity image.
  • Singularity >= 2.1 - to run the Singularity image (versions 2.0 and older are not supported!). Note that if running a 2.4 image using earlier versions, not all (later developed) features may be available.


Build a Squashfs Image

Squashfs is the recommended image type, it is compressed and less prone to degradation over time. You don't need to specify anything special to create it:

This is a path on my host, the image will be written here

$ mkdir -p /tmp/test

And here is the command to run. Notice that I am mounting the path /tmp/test that I created above to /output in the container, where the container image will be written (and seen on my host).

$ docker run -v /var/run/docker.sock:/var/run/docker.sock \
-v /tmp/test:/output \
--privileged -t --rm \ \

Image Format: squashfs
Inspected Size: 188 MB

(1/10) Creating a build sandbox...
(2/10) Exporting filesystem...
(3/10) Creating labels...
(4/10) Adding run script...
(5/10) Setting ENV variables...
(6/10) Adding mount points...
(7/10) Fixing permissions...
(8/10) Stopping and removing the container...
(9/10) Building squashfs container...
Building image from sandbox: /tmp/
Building Singularity image...
Singularity container built: /tmp/ubuntu_14.04-2017-09-13-3e51deeadc7b.simg
Cleaning up...
(10/10) Moving the image to the output folder...
     62,591,007 100%  340.92MB/s    0:00:00 (xfr#1, to-chk=0/1)
Final Size: 60MB

We can now see the finished image!

$ ls /tmp/test

And use it!

$ singularity shell /tmp/test/ubuntu_14.04-2018-04-27-c7e04ea7fa32.simg
Singularity: Invoking an interactive shell within container...

Singularity ubuntu_14.04-2018-04-27-c7e04ea7fa32.simg:~/Documents/Dropbox/Code/singularity/docker2singularity> 

Take a look again at the generation code above, and notice how the image went from 188MB to 60MB? This is one of the great things about the squashfs filesystem! This reduction is even more impressive when we are dealing with very large images (e.g., ~3600 down to ~1800). A few notes on the inputs shown above that you should edit:

  • /tmp/test: the path you want to have the final image reside. If you are on windows this might look like D:\host\path\where\to\output\singularity\image. -ubuntu:14.04: the docker image name you wish to convert (it will be pulled from Docker Hub if it does not exist on your host system).

docker2singularity uses the Docker daemon located on the host system. It will access the Docker image cache from the host system avoiding having to redownload images that are already present locally.

If you ever need to make changes, you can easily export the squashfs image into either a sandbox folder or ext3 (legacy) image, both of which have writable.

sudo singularity build --sandbox sandbox/ production.simg
sudo singularity build --writable ext3.img production.simg

Custom Naming

Added for version 2.5.1, you can specify the name of your container with the -n/--name argument, as follows:

docker run -v /var/run/docker.sock:/var/run/docker.sock \
-v /tmp/test:/output \
--privileged -t --rm \ \
--name meatballs ubuntu:14.04


$ ls /tmp/test/

Inspect Your Image

New with docker2singularity 2.4, the labels for the container are available with inspect:

 singularity inspect ubuntu_14.04-2017-09-13-3e51deeadc7b.simg 
    "": "squashfs",
    "org.label-schema.docker.version": "17.06.2-ce",
    "org.label-schema.schema-version": "1.0",
    "": "docker2singularity",
    "": "sha256:dea1945146b96542e6e20642830c78df702d524a113605a906397db1db022703",
    "": "2017-10-28-17:19:18",
    "org.label-schema.singularity.version": "2.4-dist",
    "org.label-schema.docker.created": "2017-09-13"

as is the runscript and environment

singularity inspect --json -e -r ubuntu_14.04-2017-09-13-3e51deeadc7b.simg 
    "data": {
        "attributes": {
            "environment": "# Custom environment shell code should follow\n\n",
            "runscript": "#!/bin/sh\n/bin/bash $@\n"
        "type": "container"

Build a Sandbox Image

A sandbox image is a folder that is ideal for development. You can view it on your desktop, cd inside and browse, and it works like a Singularity image. To create a sandbox, specify the -f flag:

docker run -v /var/run/docker.sock:/var/run/docker.sock \
-v /host/path/change/me:/output \
--privileged -t --rm \ \
-f \

Importantly, you can use --writable, and if needed, you can convert a sandbox folder into a production image:

sudo singularity build sandbox/ production.simg

Build a Legacy (ext3) Image

You can build a legacy ext3 image (with --writable) with the -w flag. This is an older image format that is more prone to degradation over time, and (building) may not be supported for future versions of the software.

docker run -v /var/run/docker.sock:/var/run/docker.sock \
-v /host/path/change/me:/output \
--privileged -t --rm \ \
-w \

You can also use --writable and convert an ext3 image into a production image:

sudo singularity build ext3.img production.simg

Contributed Examples

The following are a list of brief examples and tutorials generated by the Singularity community for using docker2singularity. If you have an example of your own, please let us know!

  • docker2singularity-demo: an example of using docker2singularity on MacOS and using Vagrant to test the output Singularity image, complete with notes and a nice Makefile.

Tips for making Docker images compatible with Singularity

  • Define all environmental variables using the ENV instruction set. Do not rely on .bashrc, .profile, etc.
  • Define an ENTRYPOINT instruction set pointing to the command line interface to your pipeline
  • Do not define CMD - rely only on ENTRYPOINT
  • You can interactively test the software inside the container by overriding the ENTRYPOINT docker run -i -t --entrypoint /bin/bash bids/example
  • Do not rely on being able to write anywhere other than the home folder and /scratch. Make sure your container runs with the --read-only --tmpfs /run --tmpfs /tmp parameters (this emulates the read-only behavior of Singularity)
  • Don’t rely on having elevated user permissions
  • Don’t use the USER instruction set


Here are some frequently asked questions if you run into trouble!

"client is newer than server" error

If you are getting the following error: docker: Error response from daemon: client is newer than server

You need to use the docker info command to check your docker version and use it to grab the correct corresponding version of docker2singularity. For example:

     docker run \        
     -v /var/run/docker.sock:/var/run/docker.sock \
     -v D:\host\path\where\to\output\singularity\image:/output \
     --privileged -t --rm \
     singularityware/docker2singularity:1.11 \            

Currently only the 1.10, 1.11, 1.12, and 1.13 versions are supported. If you are using an older version of Docker you will need to upgrade.

My cluster/HPC requires Singularity images to include specific mount points

If you are getting WARNING: Non existant bind point (directory) in container: '/shared_fs' or a similar error when running your Singularity image that means that your Singularity images require custom mount points. To make the error go away you can specify the mount points required by your system when creating the Singularity image:

     docker run \        
     -v /var/run/docker.sock:/var/run/docker.sock \
     -v D:\host\path\where\to\output\singularity\image:/output \
     --privileged -t --rm \ \            
     -m "/shared_fs /custom_mountpoint2" \


1. Build the container

You can build a development container as follows. First, update the VERSION to be correct.

docker build -t ${image} .

2. Test the container

We have a Circle CI builder that tests generation of the final image, and basic running to ensure the entrypoint is functioning. Since we cannot run the priviledged Docker daemon on Circle, a script is provided for local testing.

chmod u+x

If there are missing tests or you have added new features, please add the test here!

3. Documentation

If you have added new features, please describe usage in the here. Don't forget to read the along with the code of conduct and add yourself to the authors file.


This work is heavily based on the docker2singularity work done by vsoch and gmkurtzer. The original record of the work can be read about in this commit. Thank you kindly to all the contributors, and please open an issue if you need help.