The Operator Framework

This Operator Framework simplifies Kubernetes operator development for model-driven application management.

A Kubernetes operator is a container that drives lifecycle management, configuration, integration and daily actions for an application. Operators simplify software management and operations. They capture reusable app domain knowledge from experts in a software component that can be shared.

This project extends the operator pattern to enable universal operators, not just for Kubernetes but also operators for traditional Linux or Windows application management.

Operators use an Operator Lifecycle Manager (OLM) to coordinate their work in a cluster. The system uses Golang for concurrent event processing under the hood, but enables the operators to be written in Python.

Simple, composable operators

Operators should 'do one thing and do it well'. Each operator drives a single microservice and can be composed with other operators to deliver a complex application.

It is better to have small, reusable operators that each drive a single microservice very well. The operator handles instantiation, scaling, configuration, optimisation, networking, service mesh, observability, and day-2 operations specific to that microservice.

Operator composition takes place through declarative integration in the OLM. Operators declare integration endpoints, and discover lines of integration between those endpoints dynamically at runtime.

Pure Python operators

The framework provides a standard Python library and object model that represents the application graph, and an event distribution mechanism for distributed system coordination and communication.

The OLM is written in Golang for efficient concurrency in event handling and distribution. Operators can be written in any language. We recommend this Python framework for ease of design, development and collaboration.

Better collaboration

Operator developers publish Python libraries that make it easy to integrate your operator with their operator. The framework includes standard tools to distribute these integration libraries and keep them up to date.

Development collaboration happens at Charmhub.io where operators are published along with integration libraries. Design and code review discussions are hosted in the Charmhub forum. We recommend the Open Operator Manifesto as a guideline for high quality operator engineering.

Event serialization and operator services

Distributed systems can be hard! So this framework exists to make it much simpler to reason about operator behaviour, especially in complex deployments. The OLM provides operator services such as provisioning, event delivery, leader election and model management.

Coordination between operators is provided by a cluster-wide event distribution system. Events are serialized to avoid race conditions in any given container or machine. This greatly simplifies the development of operators for high availability, scale-out and integrated applications.

Model-driven Operator Lifecycle Manager

A key goal of the project is to improve the user experience for admins working with multiple different operators.

We embrace model-driven operations in the Operator Lifecycle Manager. The model encompasses capacity, storage, networking, the application graph and administrative access.

Admins describe the application graph of integrated microservices, and the OLM then drives instantiation. A change in the model is propagated to all affected operators, reducing the duplication of effort and repetition normally found in operating a complex topology of services.

Administrative actions, updates, configuration and integration are all driven through the OLM.

Getting started

A package of operator code is called a charm. You will use charmcraft to register your operator name, and publish it when you are ready.

$ sudo snap install charmcraft --beta
charmcraft (beta) 0.6.0 from John Lenton (chipaca) installed

Charms written using the operator framework are just Python code. The goal is to feel natural for somebody used to coding in Python, and reasonably easy to learn for somebody who is not a pythonista.

The dependencies of the operator framework are kept as minimal as possible; currently that's Python 3.5 or greater, and PyYAML (both are included by default in Ubuntu's cloud images from 16.04 on).

A quick introduction

Make an empty directory my-charm and cd into it. Then start a new charm with:

$ charmcraft init
All done.
There are some notes about things we think you should do.
These are marked with ‘TODO:’, as is customary. Namely:
      README.md: fill out the description
      README.md: explain how to use the charm
  metadata.yaml: fill out the charm's description
  metadata.yaml: fill out the charm's summary

Charmed operators are just Python code. The entry point to your charm can be any filename, by default this is src/charm.py which must be executable (and probably have #!/usr/bin/env python3 on the first line).

You need a metadata.yaml to describe your charm, and if you will support configuration of your charm then config.yaml files is required too. The requirements.txt specifies any Python dependencies.

$ tree my-charm/
my-charm/
├── actions.yaml
├── config.yaml
├── LICENSE
├── metadata.yaml
├── README.md
├── requirements-dev.txt
├── requirements.txt
├── run_tests
├── src
│   └── charm.py
├── tests
│   ├── __init__.py
│   └── my_charm.py

src/charm.py here is the entry point to your charm code. At a minimum, it needs to define a subclass of CharmBase and pass that into the framework main function:

from ops.charm import CharmBase
from ops.main import main

class MyCharm(CharmBase):
    def __init__(self, *args):
        super().__init__(*args)
        self.framework.observe(self.on.start, self.on_start)

    def on_start(self, event):
        # Handle the start event here.

if __name__ == "__main__":
    main(MyCharm)

That should be enough for you to be able to run

$ charmcraft build
Done, charm left in 'my-charm.charm'
$ juju deploy ./my-charm.charm

🛈 More information on charmcraft can also be found on its github page.

Happy charming!

Testing your charms

The operator framework provides a testing harness, so you can check your charm does the right thing in different scenarios, without having to create a full deployment. pydoc3 ops.testing has the details, including this example:

harness = Harness(MyCharm)
# Do initial setup here
relation_id = harness.add_relation('db', 'postgresql')
# Now instantiate the charm to see events as the model changes
harness.begin()
harness.add_relation_unit(relation_id, 'postgresql/0')
harness.update_relation_data(relation_id, 'postgresql/0', {'key': 'val'})
# Check that charm has properly handled the relation_joined event for postgresql/0
self.assertEqual(harness.charm. ...)

Talk to us

If you need help, have ideas, or would just like to chat with us, reach out on IRC: we're in #smooth-operator on freenode (or try the webchat).

We also pay attention to Charmhub discourse

You can also deep dive into the API docs if that's your thing.

Operator Framework development

To work in the framework itself you will need Python >= 3.5 and the dependencies in requirements-dev.txt installed in your system, or a virtualenv:

virtualenv --python=python3 env
source env/bin/activate
pip install -r requirements-dev.txt

Then you can try ./run_tests, it should all go green.

For improved performance on the tests, ensure that you have PyYAML installed with the correct extensions:

apt-get install libyaml-dev
pip install --force-reinstall --no-cache-dir pyyaml

If you want to build the documentation you'll need the requirements from docs/requirements.txt, or in your virtualenv

pip install -r docs/requirements.txt

and then you can run ./build_docs.