Docker and Singularity are container systems that allow users full control over their software environment. You can create your own container image or choose from a set of pre-defined images, and specify that your submitted jobs run within one of these.

For jobs on OSG, it does not matter whether you provide a Docker or Singularity image. Either is compatible with our system and can be used with little to no modification. Determining factors on when to use Singularity images over Docker images include if an image already exists, external to OSG distribution preferences, and if you have experience building images in one for format and not the other.

Because OSG is a distributed infrastructure and workloads consists of a large number jobs (and there container executions), it is important to consider how the container image is transferred to the execution nodes. The instructions below contain best practices when it comes to access both Singularity and Docker images.

When using a container for your jobs, the container image is automatically started up when HTCondor matches your job to a slot. The executable provided in the submit script will be run within the context of the container image, having access to software and libraries that were installed to the image, as if they were already on the server where the job is running. Job executables need not (and should not) run any commands to start the container. Nor should the container image contain any entrypoint/cmd - the job is the command to be run in the container.

Exploring Images on the Access Points

Just like it is important to test your codes and jobs at a small scale, you should make sure that your container is working correctly. One way to explore how OSG sees your container images, is to explore them on the OSG Connect access points. Start an interactive session with the Singularity "shell" mode. The recommended command line, similar to how containers are started for jobs, is:

singularity shell \
            --home $PWD:/srv \
            --pwd /srv \
            --bind /cvmfs \
            --scratch /var/tmp \
            --scratch /tmp \
            --contain --ipc --pid \
            /cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-ubuntu-20.04:latest/

This will give you an interactive shell in an Ubuntu 20.04 container, with your current working directory mounted under /srv. You can explore the container and test your code with for example your own inputs from your home directory. Once you are down exploring, exit the container by running exit or with CTRL+D

OSG-Provided Images

The OSG Team maintains a set of images that are already in the OSG Singularity repository. A list of available containers can be found on this page.

If the software you need isn't already supported in a listed container, you can use your own container or any container image in Docker Hub (see sections further below). Once the container you need is in the OSG Singularity repository, your can submit jobs that run within a particular container by listing the container image in the submit file.

For example, this is what a submit file might look like to run your job within our EL8 container:

+SingularityImage = "/cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-el8:latest"

<other usual submit file lines>
queue

Custom Singularity Images

If you already have software in the form of a .sif Singuilarity file, and that file is within the supported data sizes, you can stage the .sif file with your job. The image will be resused for each job, and thus the preferred transfer method is Stash. Store the .sif file under /public/$USERNAME/, and then use the stash url directly in the +SingularityImage attribute. Note that you can not use shell variable expansion in the submit file - be sure to replace the username with your actual OSG Connect username. Example:

+SingularityImage = "stash:///osgconnect/public/USERNAME/my-custom-image-v1.sif"

<other usual submit file lines>
queue

Be aware that Stash aggressively caches the image based on file naming. If you need to do quick changes, please use versioning of the .sif file so that the caches see a "new" name. In this example, replacing my-custom-image-v1.sif with new content will probably mean that some nodes get the old version and some nodes the new version. Prevent this by creating a new file named with v2.

More information on how to create Singularity images can be found in the Singularity Images Guide.

Custom Docker Images

If you would prefer to create or use an existing Docker Hub container, for example an authoritative container for your software which already exists in Docker Hub, OSG can distribute the image for you via CVMFS. The result is a synchronized copy of the image under /cvmfs/singularity.opensciencegrid.org/ which is cached and available to the execution nodes. Creating and/or registering a Docker image is described in the Docker Images Guide.

To run a job with a Docker image, use the +SingularityImage to specify the image the job should be using. Example:

+SingularityImage = "/cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-el8:latest"

<other usual submit file lines>
queue

Another example would be if your Docker Hub username is alice and you created a container called ncbi-blast added to the OSG Singularity repository, your submit file will include:

+SingularityImage = "/cvmfs/singularity.opensciencegrid.org/alice/ncbi-blast"

<other usual submit file lines>
queue

Using Containers from Non-OSG Connect Access Points

Users on non-OSG Connect access points can use all the container functionality described above, but will have to use slightly more complex job submit files. This is because the OSG Connect access points uses job transforms to update the jobs based on the +SingularityImage attribute, and OSG Connect users also have direct access to Stash.

To run a Singularity image from a non-OSG Connect access point, include a job requirements, and specify a method for image transfer. For example:

Requirements = HAS_SINGULARITY == TRUE && SINGULARITY_CAN_USE_SIF = TRUE
transfer_input_files = http://datastore.host/mycontainer.sif
+SingularityImage = "./mycontainer.sif"

<other usual submit file lines>
queue

For images available on CVMFS, just add job requirements:

Requirements = HAS_SINGULARITY == TRUE
+SingularityImage = "/cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-el8:latest"

<other usual submit file lines>
queue

Frequently Asked Questions / Common Issues

FATAL: kernel too old

If you get a FATAL: kernel too old error, it means that the glibc version in the image is too new for the kernel on the host. You can work around this problem by specifying the minimum host kernel. For example, if you want to run the Ubuntu 18.04 image, specfy a minimum host kernel of 3.10.0, formatted as 31000 (major * 10000 + minor * 100 + patch):

Requirements = HAS_SINGULARITY == True && OSG_HOST_KERNEL_VERSION >= 31000

Learning More

For more information about Docker, please see:

and Singularity, please see:

Singularity has become the preferred containerization method in scientific computing. The following talk describes Singularity for scientific computing:

 

This page was updated on Jul 20, 2022 at 21:25 from start/software/containers.md.