Singularity is good friends with Docker. The reason is because the developers use and really like using Docker, and scientists have already put much resources into creating Docker images. Thus, one of our early goals was to support Docker. What can you do?
- You don’t need Docker installed
- You can shell into a Singularity-ized Docker image
- You can run a Docker image instantly as a Singularity image
- You can import Docker images, including environment, guts, and labels, into your Singularity image (without sudo!)
TLDR (Too Long Didn’t Read)
You can shell, import, run, and exec.
singularity shell docker://ubuntu:latest singularity run docker://ubuntu:latest singularity exec docker://ubuntu:latest echo "Hello Dinosaur!" singularity create ubuntu.img singularity import ubuntu.img docker://ubuntu:latest printf "Bootstrap:docker\nFrom:ubuntu:latest" > Singularity singularity create ubuntu.img sudo singularity bootstrap ubuntu.img Singularity
Import a Docker image into a Singularity Image
The core of a Docker image is basically a compressed set of files, a set of
.tar.gz that (if you look in your Docker image folder on your host machine, you will see. The Docker Registry, which you probably interact with via Docker Hub, serves these layers. These are the layers that you see downloading when you interact with the docker daemon. We are going to use these same layers for Singularity!
Quick Start: The Docker Registry
The Docker engine communicates with the Docker Hub via the Docker Remote API, and guess what, we can too! The easiest thing to do is create an image, and then pipe a Docker image directly into it from the Docker Registry. You don’t need Docker installed on your machine, but you will need a working internet connection. Let’s create an ubuntu operating system, from Docker:
singularity create ubuntu.img Initializing Singularity image subsystem Opening image file: ubuntu.img Creating 768MiB image Binding image to loop Creating file system within image Image is done: ubuntu.img
Note that the default size is 768MB, you can modify this by adding the
-s argument like:
singularity create --size 2000 ubuntu.img
If you aren’t sure about the size? Try building into a folder first.
mkdir fatty singularity import fatty docker://ubuntu:latest du -sh fatty/
Next, let’s import a Docker image into it!
singularity import ubuntu.img docker://ubuntu Cache folder set to /home/vanessa/.singularity/docker Importing: base Singularity environment Importing: /home/vanessa/.singularity/docker/sha256:6d9ef359eaaa311860550b478790123c4b22a2eaede8f8f46691b0b4433c08cf.tar.gz Importing: /home/vanessa/.singularity/docker/sha256:9654c40e9079e3d5b271ec71f6d83f8ce80cfa6f09d9737fc6bfd4d2456fed3f.tar.gz Importing: /home/vanessa/.singularity/docker/sha256:e8db7bf7c39fab6fec91b1b61e3914f21e60233c9823dd57c60bc360191aaf0d.tar.gz Importing: /home/vanessa/.singularity/docker/sha256:f8b845f45a87dc7c095b15f3d9661e640ebc86f42cd8e8ab36674846472027f7.tar.gz Importing: /home/vanessa/.singularity/docker/sha256:d54efb8db41d4ac23d29469940ec92da94c9a6c2d9e26ec060bebad1d1b0e48d.tar.gz Importing: /home/vanessa/.singularity/docker/sha256:fe44851d529f465f9aa107b32351c8a0a722fc0619a2a7c22b058084fac068a4.tar.gz singularity shell ubuntu.img Singularity: Invoking an interactive shell within container... Singularity ubuntu.img>
The Build Specification file, Singularity
Just like Docker has the Dockerfile, Singularity has a file called Singularity that (currently) applications like Singularity Hub know to sniff for. For reproducibility of your containers, our strong recommendation is that you build from these files. Any command that you issue to change a container with
--writable is by default not recorded, and your container loses its reproducibility. So let’s talk about how to make these files! First, let’s look at the absolute minimum requirement:
Bootstrap: docker From: tensorflow/tensorflow:latest
We would save this content to a file called
Singularity and then issue the following commands to bootstrap the image from the file
singularity create --size 4000 tensorflow.img sudo singularity bootstrap tensorflow.img Singularity
but just those two lines and doing bootstrap is silly, because we would achieve the same thing by doing: environment
singularity create --size 4000 tensorflow.img singularity import tensorflow.img docker://tensorflow/tensorflow:latest
The power of bootstrap comes with the other stuff that you can do! This means running specific install commands, specifying your containers runscript (what it does when you execute it), adding files, labels, and customizing the environment. Here is a full Singularity file:
Bootstrap: docker From: tensorflow/tensorflow:latest %runscript exec /usr/bin/python "$@" %post echo "Post install stuffs!" %files /home/vanessa/Desktop/analysis.py /tmp/analysis.py relative_path.py /tmp/analysis2.py %environment TOPSECRET=pancakes HELLO=WORLD export HELLO TOPSECRET %labels AUTHOR Vanessasaur
In the example above, I am overriding any Dockerfile
CMD because I have defined a
%runscript. If I want the Dockerfile
ENTRYPOINT to take preference, I would remove the
%runscript section. If I want to use
CMD instead of
ENTRYPOINT, I would again remove the runscript, and add IncludeCmd to the header:
Bootstrap: docker From: tensorflow/tensorflow:latest IncludeCmd: yes %post echo "Post install stuffs!"
Did you know that you can commit this Singularity file to a Github repo and it will automatically build for you when you push to Singularity Hub?. This will ensure maximum reproducibility of your work.
How does the runscript work?
Docker has two commands in the
Dockerfile that have something to do with execution,
ENTRYPOINT. The differences are subtle, but the best description I’ve found is the following:
CMDis to provide defaults for an executing container.
ENTRYPOINThelps you to configure a container that you can run as an executable.
Given the definition, the
ENTRYPOINT is most appropriate for the Singularity
%runscript, and so using the default bootstrap (whether from a
docker:// endpoint or a
Singularity spec file) will set the
ENTRYPOINT variable as the runscript. You can change this behavior by specifying
IncludeCmd: yes in the Spec file (see below). If you provide any sort of
%runscript in your Spec file, this overrides anything provided in Docker. In summary, the order of operations is as follows:
- If a
%runscriptis specified in the
Singularityspec file, this takes prevalence over all
- If no
%runscriptis specified, or if the
importcommand is used as in the example above, the
ENTRYPOINTis used as runscript.
- If no
%runscriptis specified, but the user has a
IncludeCmd, then the Docker
- If no
%runscriptis specified, and there is no
ENTRYPOINT, the image’s default execution action is to run the bash shell.
How do I specify my Docker image?
In the example above, you probably saw that we referened the docker image first with the uri
docker:// and that is important to tell Singularity that it will be pulling Docker layers. To ask for ubuntu, we asked for
docker://ubuntu. This uri that we give to Singularity is going to be very important to choose the following Docker metadata items:
- registry (e.g., “index.docker.io”)
- namespace (e.g., “library”)
- repository (e.g., “ubuntu”)
- tag (e.g., “latest”) OR version (e.g., “@sha256:1234…)
When we put those things together, it looks like this:
By default, the minimum requirement is that you specify a repository name (eg, ubuntu) and it will default to the following:
If you provide a version instead of a tag, that will be used instead:
You can have one or the other, both are considered a “digest” in Docker speak.
If you want to change any of those fields, then just specify what you want in the URI.
For both import and bootstrap using a build spec file, by default we use the Docker Registry
index.docker.io. Singularity first tries the call without a token, and then asks for one with pull permissions if the request is defined. However, it may be the case that you want to provide a custom token for a private registry. You have two options. You can either provide a
Password in the build specification file (if stored locally and there is no need to share), or (in the case of doing an import or needing to secure the credentials) you can export these variables to environmental variables. We provide instructions for each of these cases:
Authentication in the Singularity Build File
You can simply specify your additional authentication parameters in the header with the labels
Username: vanessa Password: [password]
Again, this can be in addition to specification of a custom registry with the
Authentication in the Environment
You can export your username, and password for Singularity as follows:
export SINGULARITY_DOCKER_USERNAME=vanessasaur export SINGULARITY_DOCKER_PASSWORD=rawwwwwr
If you are having trouble, you can test your token by obtaining it on the command line and putting it into an environmental variable,
CREDENTIAL=$(echo -n vanessa:[password] | base64) TOKEN=$(http 'https://auth.docker.io/token?service=registry.docker.io&scope=repository:vanessa/code-samples:pull' Authorization:"Basic $CREDENTIAL" | jq -r '.token')
This should place the token in the environmental variable
TOKEN. To test that your token is valid, you can do the following
http https://index.docker.io/v2/vanessa/code-samples/tags/list Authorization:"Bearer $TOKEN"
The above call should return the tags list as expected. And of course you should change the repo name to be one that actually exists that you have credentials for.
While most docker images can import and run without a hitch, there are some special cases for which things can go wrong. Here is a general list of suggested practices, and if you discover a new one in your building ventures please let us know.
1. Installation to Root
When using Docker, you typically run as root, meaning that root’s home at
/root is where things will install given a specification of home. This is fine when you stay in Docker, or if the content at
/root doesn’t need any kind of write access, but generally can lead to a lot of bugs because it is, after all, root’s home. This leads us to best practice #1.
Don’t install anything to root’s home,
2. Library Configurations
The command ldconfig is used to update the shared library cache. If you have software that requires symbolic linking of libraries and you do the installation without updating the cache, then the Singularity image (in read only) will likely give you an error that the library is not found. If you look in the image, the library will exist but the symbolic link will not. This leads us to best practice #2:
Update the library cache at the end of your Dockerfile with a call to ldconfig.
3. Don’t install to $HOME or $TMP
We can assume that the most common Singularity use case has the $USER home being automatically mounted to
$TMP also mounted. Thus, given the potential for some kind of conflict or missing files, for best practice #3 we suggest the following:
Don’t put container valuables in
Have any more best practices? Please let us know!
Why won’t my image bootstrap work? If you can’t find an answer on this site, please ping us an issue. If you’ve found an answer and you’d like to see it on the site for others to benefit from, then post to us here.
This entire process will hopefully change in two ways. First, we hope to collapse the image creation and bootstrapping, so you have the option to do them both in one full swing. Second, we hope to eventually figure out some kind of solution to import Docker containers without needing sudo.