Skip to content

Latest commit

 

History

History
269 lines (185 loc) · 8.44 KB

contributing.md

File metadata and controls

269 lines (185 loc) · 8.44 KB

Contributing to MongoDB Kubernetes Operator

First you need to get familiar with the Architecture guide, which explains from a high perspective how everything works together.

After our experience building the Enterprise MongoDB Kubernetes Operator, we have realized that it is very important to have a clean environment to work, and as such we have adopted a strategy that makes it easier for everyone to contribute.

This strategy is based on using envtest for setting up the tests and go test for running the tests, and making the test-runner itself run as a Kubernetes Pod. This makes it easier to run the tests in environments with access to a Kubernetes cluster with no go toolchain installed locally, making it easier to reproduce our local working environments in CI/CD systems.

High-Perspective Architecture

The operator itself consists of 1 image, that has all the operational logic to deploy and maintain the MongoDBCommunity resources in your cluster.

The operator deploys MongoDB in Pods (via a higher-level resource, a StatefulSet), on each Pod there will be multiple images coexisting during the lifetime of the MongoDB server.

  • Agent image: This image includes a binary provided by MongoDB that handles the local operation of a MongoDB server given a series of configurations provided by the operator. The configuration exists as a ConfigMap that's created by the operator and mounted in the Agent's Pod.

  • MongoDB image: Docker image that includes the MongoDB server.

  • Version upgrade post-start hook image: This image includes a binary that helps orchestrate the restarts of the MongoDB Replica Set members, in particular, when dealing with version upgrades, which requires a very precise set of operations to allow for seamless upgrades and downgrades, with no downtime.

Each Pod holds a member of a Replica Set, and each Pod has different components, each one of them in charge of some part of the lifecycle of the MongoDB database.

Getting Started

PR Prerequisites

  • Please ensure you have signed our Contributor Agreement. You can find it here.

  • Please ensure that all commits are signed.

Developer Configuration

The operator is built using golang. We use a simple json file that describe some local options that you need to set for the testing environment to be able to run properly. Create a json file with the following content:

{
  "namespace": "mongodb",
  "repo_url": "localhost:5000",
  "operator_image": "mongodb-kubernetes-operator",
  "e2e_image": "community-operator-e2e",
  "version_upgrade_hook_image": "community-operator-version-upgrade-post-start-hook",
  "agent_image_ubuntu": "mongodb-agent-dev",
  "agent_image_ubi": "mongodb-agent-ubi-dev",
  "readiness_probe_image": "mongodb-kubernetes-readiness",
  "s3_bucket": ""
}

Config Options

  1. namespace is the namespace that will be used by scripts/tooling. All the resources will be deployed here.
  2. repo_url the repository that should be used to push/pull all images.
  3. operator_image will be used as the name of the operator deployment, and the name of the operator image when build.
  4. e2e_image the name of e2e test image that will be built.
  5. version_upgrade_hook_image the name of the version upgrade post start hook image.
  6. image_type this can be either ubi or ubuntu and determines the distro of the images built. (currently only the agent image has multiple distros)
  7. agent_image_ubuntu the name of the ubuntu agent image.
  8. agent_image_ubi the name of the ubi agent image.
  9. s3_bucket the S3 bucket that Dockerfiles will be pushed to as part of the release process. Note: this is only required when running the release tasks locally.

You can set the MONGODB_COMMUNITY_CONFIG environment variable to be the absolute path of this file. It will default to ~/.community-operator-dev/config.json

Please see here to see how to build and deploy the operator locally.

Configure Docker registry

The build process consists of multiple Docker images being built. You need to specify where you want the locally built images to be pushed. The Docker registry needs to be accessible from your Kubernetes cluster.

Local kind cluster

For local testing you can use a local Kind cluster.

Test Namespace

You can change the namespace used for tests, if you are using Kind, for instance, you can leave this as mongodb.

Python Environment

The test runner is a Python script, in order to use it a virtualenv needs to be created.

Python 3.9 is not supported yet. Please use Python 3.8.

Pip

python -m venv venv
source venv/bin/activate
python -m pip install -r requirements.txt

Pipenv

  • create a python environment and install dependencies.
pipenv install -r requirements.txt
  • activate the python environment.
pipenv shell

Deploying the Operator

In order to deploy the Operator from source, you can run the following command.

make operator-image deploy

This will build and deploy the operator to namespace specified in your configuration file.

See the operator deployment

kubectl get pods

(Optional) Create a MongoDBCommunity Resource

Follow the steps outlined here to deploy some resource.

Cleanup

To remove the operator and any created resources you can run

make undeploy

Alternatively, you can run the operator locally with

make run

Running Tests

Unit tests

Unit tests should be run from the root of the project with:

make test

E2E Tests

If this is the first time running E2E tests, you will need to ensure that you have built and pushed all images required by the E2E tests. You can do this by running.

make all-images

For subsequent tests you can use

make e2e-k8s test=<test-name>

This will only re-build the e2e test image.

We have built a simple mechanism to run E2E tests on your cluster using a runner that deploys a series of Kubernetes objects, runs them, and awaits for their completion. If the objects complete with a Success status, it means that the tests were run successfully.

The available tests can be found in the tests/e2e directory, at the time of this writing we have:

$ ls -l test/e2e
replica_set
replica_set_change_version
replica_set_readiness_probe
replica_set_scale
...

The tests should run individually using the runner like this:

make e2e-k8s test=replica_set

This will run the replica_set E2E test which is a simple test which installs a MongoDB Replica Set and asserts that the deployed server can be connected to.

Run the test locally with go test & Telepresence

make e2e-telepresence test=<test-name>

This method uses telepresence to allow connectivity as if your local machine is in the kubernetes cluster, there will be full MongoDB connectivity using go test locally.

Note: you must install telepresence before using this method.

If on MacOS, you can run make install-prerequisites-macos which will perform the installation.

Running with Github Actions

Run a single test

make e2e-gh test=<test-name>

Run all tests.

  • Navigate to the Actions tab on the github repo
  • Run E2E > Run Workflow > Your Branch

Note: the code must be pushed to a remote branch before this will work.

Troubleshooting

When you run a test locally, if the e2e-test pod is present, you will have to first manually delete it; failing to do so will cause the e2e-test pod to fail.

Writing new E2E tests

You can start with the replica_set test as a starting point to write a new test. The tests are written using envtest and they are run using go test.

Adding a new test is as easy as creating a new directory in test/e2e with the new E2E test, and to run them:

make e2e test=<test-name>

Before Committing your code

Set up pre-commit hooks

To set up the pre-commit hooks, please create symbolic links from the provided hooks:

  • Navigate to your .git/hooks directory:

    cd .git/hooks

  • Create a symlink for every file in the scripts/git-hooks directory:

    ln -s -f ../../scripts/git-hooks/* .