runtimes.md

runtimes

runtime is a container for data science and machine learning applications to be used in KDL and KRE projects. A runtime refers to a pre-configured containerized environment, equipped with the tools, libraries, and dependencies which data scientists need to develop, test and deploy models.

Tip

Read the konstellation-runtimes repository for more information.

How-to create a new runtime

Important

If you are creating a new runtime, make sure that the name is not already in use, because it will be used as a unique identifier.

This is the JSON schema to add new runtime to the runtimes collection in the kdl MongoDB database:

{
  "name": "<name>",
  "desc": "<description>",
  "labels": ["label1", "labelN"],
  "docker_image": "<repository/image>",
  "docker_tag": "<tag-image>"
}

Tip

All images are in this konstellation/kdl-py registry.

How-to add runtimes to the MongoDB

Add port-forward on MongoDB pod:

kubectl -n <namespace> port-forward svc/<mongo-svc> 27017:27017

Connect MongoDB instance:

mongosh --port 27017 --username <username> --authenticationDatabase admin --password <password>

Use the kdl database:

use kdl

Insert runtimes:

Note

Example of a runtime for Python 3.10 CPU workloads.

db.runtimes.insertOne({
  "name" : "Python 3.10 CPU - 1.4.3",
  "desc" : "Rumtime with Python 3.10 for CPU workloads",
  "labels" : [
      "Python 3.10",
      "Ubuntu 22.04",
      "CPU"
  ],
  "docker_image" : "konstellation/kdl-py",
  "docker_tag" : "3.10-cuda11.8.0-runtime-ubuntu22.04-1.2.2"
})