Quick Start

Thw Owl client is used to interact with the Owl Server and submit jobs to the Kubernetes cluster. It can also be used to perform some admin tasks.

Install

The Owl client requires Python 3.7 or higher. It is recommended to create a Python environment, e.g., using conda:

conda create -n pipelines python=3.7

In order to use the new environment:

conda activate pipelines

The latest stable version of the Owl client can be installed from PyPi:

pip install owl-pipeline-client

Authenticate to remote server

Using your credentials

owl auth login

This will ask for your username and password. Credentials will be saved to a file in your home directory $HOME/.owlrc.

Pipeline definition file

List all available pipelines in the remote server using:

owl pdef list

If the server has been configured as per the instructions the example pipeline will be available. Retrieve the pipeline definition file using:

owl pdef get example -o example.yml

This looks like:

# Version of the configuration file
version: 1

# Name of the pipeline
name: example

# Pipeline arguments
datalen: 100

# Resources requested
resources:
  threads: 10
  workers: 2
  memory: 10

The example pipeline runs a series of dummy computations using Dask. It has a unique input parameter datalen which basically controls how long the pipeline runs.

Submit pipeline

Adjust the resources and submit the pipeline using:

owl job submit example.yml