MetalLB Operator

This is the official MetalLB Operator, implementing the operator pattern for deploying MetalLB on a kubernetes cluster, as described in the related design proposal.

Prerequisites

Need to install the following packages:

• operator-sdk 1.8.0+
• controller-gen v0.11.1+

To install controller-gen, run the following:

go get sigs.k8s.io/controller-tools/cmd/[email protected]

Quick Setup

To install the MetalLB Operator using the prebuilt manifests, run the following:

kubectl apply -f bin/metallb-operator.yaml

Installation

To install the MetalLB Operator using a prebuilt image, run the following:

make deploy

Installation on OpenShift

To install the MetalLB Operator using a prebuilt image on OpenShift, run the following:

make deploy-openshift

Note: This requires kustomize 4.5.6 or above.

Note: To undeploy on OpenShift, use the undeploy-openshift target.

Usage

Once the MetalLB Operator is installed, you have to create a MetalLB custom resource to deploy a MetalLB instance. The operator will consume this resource and create all required MetalLB resources based on it. The MetalLB custom resource needs to be created inside the metallb-system namespace and be named metallb. Only one MetalLB resource can exist in a cluster.

Following is a sample MetalLB resource:

apiVersion: metallb.io/v1beta1
kind: MetalLB
metadata:
  name: metallb
  namespace: metallb-system

Setting up a development environment

Quick local installation

A quick, local installation can be done using a kind cluster and a local registry. Follow the steps below to run a locally-built metallb-operator on kind.

Install and run kind

Install kind using the instructions here.

Once kind is installed, run the following to start a kind cluster:

kind create cluster
kind get kubeconfig > kubeconfig
export KUBECONFIG=$(pwd)/kubeconfig

Build and deploy the operator

To build and deploy the operator, run the following:

export IMAGE_NAME=metallb-operator

make docker-build IMG=$IMAGE_NAME
kind load docker-image $IMAGE_NAME
IMG=$IMAGE_NAME KUSTOMIZE_DEPLOY_DIR="config/kind-ci/" make deploy

Alternatively, the image can be pushed to a Docker registry.

Building and deploying using a remote repo

To build and push an image, run the following commands, specifying the preferred image repository and image:

make docker-build IMG=quay.io/example/metallboperator
make docker-push IMG=quay.io/example/metallboperator

Once the images are pushed to the repo, you can deploy MetalLB using your custom images by running the following:

make deploy IMG=quay.io/example/metallboperator

Create a MetalLB deployment

To create a MetalLB deployment, a MetalLB Operator configuration resource needs to be created. Run the following command to create it:

cat << EOF | kubectl apply -f -
apiVersion: metallb.io/v1beta1
kind: MetalLB
metadata:
  name: metallb
  namespace: metallb-system
EOF

Running tests

To run metallb-operator unit tests (no cluster required), execute the following:

make test

To run metallb-operator e2e tests, execute the following:

make test-e2e

The e2e test need a running cluster with a MetalLB Operator deployed.

Make

Most tasks in the project are automated using a Makefile. Run make help to see the details.

Releasing

The Operator assumes the same branching structure as MetalLB. Each minor version must have a corresponding branch where we tag releases out. Versioned branches must be pinned to specific MetalLB / Operator images.

The current version of the images is bumped under hack/metallb_version.txt and under hack/operator_version.txt.

A convenience make bump_versions makefile target aligns the versions in the manifests to the content of those files.

Another Makefile target make fetch_metallb_version updates hack/metallb_version.txt with the latest tag of metallb.

In order to make a release, a tag must be made out of a release branch, pinning the relevant images.