Singularity is a container solution created for scientific and application driven workloads. Singularity containers can be used to package entire scientific workflows, software and libraries, and even data. This means that you don’t have to ask your cluster admin to install anything for you - you can put it in a Singularity container and run.
Podman is a daemonless container engine for developing, managing, and running OCI Containers on your Linux System.
Both container engines support the reusage of docker images.
When using singularity as runtime, be sure to have created and set predefined folders for building and pulling:
[create these folders]
export SINGULARITY_TMPDIR=${HOME}/.singularity/tmp
export SINGULARITY_LOCALCACHEDIR=${HOME}/.singularity/local_cache
- Move to the definiton file (
definiton.def) folder:
cd ml_keras_tf
- Build the image:
singularity build --fakeroot ml_keras_tf.sif definition.def
- Once it's ready, run the container:
Example: running container with default runscript (in this case jupyter notebook)
singularity run --nv ml_keras_tf.sif
Example: running container with a custom command
singularity exec --nv ml_keras_tf.sif jupyter notebook --no-browser
NOTE: By default singularity binds the directories
$HOME,/tmp,/proc,/sys, and/dev. So that you can modify the files wher your permissions (as user) apply.
Example: running a command from inside the container --in shell mode--
singularity shell --nv --bind data:/data ml_keras_tf.sif
jupyter notebook --no-browser
This will open a shell inside the container. Additionally, we can add extra bindings using --bind. In this example, the folder data in our current working directory is bind to the folder /data in the containers file system.
For more information refer to singularity user's guide https://sylabs.io/guides/3.4/user-guide/quick_start.html
You can do so by: mounting the container in sandbox mode and rebuilding once it is ready; using the writable mode; or by modifying the definition.def file and rebuilding the image.
For more information refer to singularity user's guide https://sylabs.io/guides/3.4/user-guide/quick_start.html
Currently, there is a basic definition file to defined an image based on ubuntu that includes python 3.
Note: podman share a quasi identical CLI as docker. So if you have docker installed you can replace
podmanfordocker. Only build requires an extra.at the end of this instruction call.
-
Move to
podman/ubuntu-python -
Build the image
podman build -f Definitionfile -t ubuntu18_10-python3
- Run the image (with podman (or docker) in an interactive shell)
podman run -it -rm localhost/ubuntu18_10-python3
NOTE: podman attaches localhost to local images
Test it
cat /etc/lsb-release
exit
Run the image (with a specific command an remove the container when it is done)
podman run -rm localhost/ubuntu18_10-python3 bash -c "cat /etc/lsb-release"
Run the image (in an interactive shell binding an external folder) the user parameter is to properly add permissions when modifying files
podman run -rm -it -v "$(pwd)"/some_folder:/development/ws:z localhost/ubuntu18_10-python3
:zoption is only for selinux issue