dss

dss command [options]

dss is the command-line front end to Canonical's Data Science Stack, an opinionated bundle of Jupyter Notebook images, MLflow model tracking, and Kubernetes plumbing distributed as a snap. It targets local Canonical Kubernetes (k8s snap) clusters but works against any Kubernetes that can host the underlying components.After dss initialize, the cluster runs a shared MLflow server and is ready to host notebook pods. dss create spawns a notebook backed by a configurable container image, persistent storage, and a service account that can talk to MLflow. The remaining lifecycle commands (list, status, logs, start, stop, remove, purge) operate on those notebooks without requiring direct kubectl use.DSS aims to be the easiest way to get a reproducible local data-science environment with GPU support and experiment tracking, while still leaving the underlying Kubernetes resources accessible to power users.

initialize [--kubeconfig file]

Store cluster credentials, allocate persistent storage for notebooks, and deploy the MLflow model registry.

Create a new Jupyter notebook of the given name wired to the shared MLflow instance. image may be a shorthand (pytorch, tensorflow) or a fully qualified image like kubeflownotebookswg/jupyter-scipy:v1.8.0.

List every notebook tracked by DSS in the current cluster.

Display deployment status, MLflow URL, and detected GPU availability.

Print logs for the named notebook, for --mlflow (the MLflow pod), or for everything with --all.

Start a stopped notebook.

Stop a running notebook.

Delete a single notebook and its persistent volume claim.

Tear down every DSS component (notebooks, MLflow, supporting resources). Irreversible - all stored data is lost.

Path to a kubeconfig file. Used by initialize and logs when the user is not already authenticated to a cluster.

Show usage for dss or for a specific subcommand.

DSS state (storage class, MLflow endpoint, kubeconfig reference) is persisted under the snap's data directory, typically /var/snap/data-science-stack/current/. Notebook resources, including the MLflow deployment, live in a dedicated namespace inside the connected Kubernetes cluster.

dss purge removes every notebook and the MLflow registry; backups must be made beforehand. The default Kubernetes target is the k8s snap; on other distributions pass --kubeconfig explicitly. Image shorthands like pytorch and tensorflow track moving upstream tags, which can change between dss releases.The name dss is overloaded - Dataiku Data Science Studio ships separate dsscli and dssadmin binaries, and Open-E's dss_cli targets the unrelated DSS storage server. This page documents the Canonical snap.

Data Science Stack is published by Canonical as the data-science-stack snap. It launched as a streamlined alternative to the full Charmed Kubeflow distribution, focusing on a single-machine workflow that pairs Jupyter with MLflow on top of the k8s snap.

kubectl(1), microk8s(1), snap(1), mlflow(1), jupyter(1)