In this lesson, you will learn how to package your Operator controller and metadata into distributable container images to ensure consistent deployment across various Kubernetes clusters.
The Operator Packaging Paradigm
When you develop an Operator, you aren't just writing code; you are building a system that lives inside the cluster. To distribute this system, you must bundle two distinct components: the Operator image (the binary controller) and the Bundle manifests (CRDs, RBAC rules, and OLM metadata).
Building the Operator Image
First, you containerize your controller logic. Using the Operator SDK, you generate a Dockerfile that packages your controller binary into a minimal base image (like distroless). You build and push this image to a registry just like any other application:
# Build the controller image
docker build -t my-registry.com/my-operator:v1.0.0 .
# Push to the container registry
docker push my-registry.com/my-operator:v1.0.0
Creating the Operator Bundle
Distribution via the Operator Lifecycle Manager (OLM) requires a "Bundle." A bundle is a container image containing your CRD YAMLs, CSV (ClusterServiceVersion) files, and RBAC manifests. The operator-sdk command automates this by packaging your local config files into an image format that registries can host:
# Generate the bundle manifests
make bundle VERSION=1.0.0
# Build and push the bundle image
operator-sdk bundle build my-registry.com/my-operator-bundle:v1.0.0
docker push my-registry.com/my-operator-bundle:v1.0.0
Why This Workflow Matters
By treating your Operator as a containerized bundle, you create a "single source of truth." When you push your bundle image to a registry, you are providing a versioned, immutable artifact. Tools like OLM can then point to this registry, pull the image, and perform a controlled installation or upgrade of your Operator. This keeps your CI/CD pipeline clean: you build your code, push the controller, bundle the metadata, and publish the bundle. This ensures that every cluster—whether it is a developer sandbox or a production environment—is running the exact same version of your CRDs and controller logic.
Key Takeaway: Distributing Operators requires packaging both the controller binary as a container image and the deployment manifests as a versioned bundle image, providing a reliable, standardized way for OLM to install and update your software across any Kubernetes environment.