ewoksjob

Utilities for job scheduling of ewoks workflows.

Ewoksjob provides an ewoks interface for asynchronous and distributed scheduling of ewoks from python.

Note that ewoksjob distributes the execution of workflows while ewoksdask distributes the execution of tasks in a workflow. So in the context of workflows, job scheduling exists on two levels.

The primary clients that need to schedule workflows are

• Ewoksserver: web backend for ewoks.
• Bliss: the ESRF beamline control system.
• Daiquiri: web backend for Bliss.

Installation

Install on the client side

pip install ewoksjob

Install on the worker side

pip install ewoksjob[worker]

Getting started

Start a worker pool that can execute ewoks graphs

ewoksjob worker

Submit a workflow on the client side

from ewoksjob.client import submit

workflow = {"graph": {"id": "mygraph"}}
future = submit(args=(workflow,))
result = future.result(timeout=None)

Note that both environments need to be able to import celeryconfig which contains celery configuration (mainly the message broker and result backend URL's).

Hello world example

Clone the git repository and start a worker pool

scripts/worker.sh --sql

Submit workflows

scripts/runjobs.sh --sql

Tests

pytest --pyargs ewoksjob

To run the redis tests you need redis-server (e.g. conda install redis-server).

Documentation

https://ewoksjob.readthedocs.io/