Development

The operator is built upon the quarkus-operator-sdk library.

Requirements

Creates the operator docker image

mvn package -DskipTests -pl kaap -am -Pskip-crds

If you want to use a public k8s cluster, you would need to deploy the operator image to DockerHub or another public registry.

$dockerhub_repo=<your dockerhub repo>
docker tag datastax/kaap:latest-dev
$dockerhub_repo/kaap:latest
docker push $dockerhub_repo/kaap:latest

Local deployment

In order to quickly test your changes, it’s possible to create a local k8s cluster. This repository contains several scripts to work with a K3s cluster. The only requirements is Docker.

Startup K3s

./tests/src/test/scripts/local-k3s.sh

It creates a namespace ns for convenience. Now the ~/.kube/config has been modified to point to the local k3s cluster.

kubectl config get-contexts

The cluster is deployed in a docker container. You can kill and recreate it whenever you want. No mounts are configured.

docker ps

After testing, do remember to remove unused volumes from docker.

docker volume prune -f

If you did some changes to the operator code and wants to redeploy it, you can use the following script.

./tests/src/test/scripts/local-k3s-update-operator-no-crds.sh

This will build and push the new docker image to the container. You’ll need to manually restart the operator pod.

Use the Quarkus dev mode

In this mode, the kaap code is hot reloaded at every change in the source code.

mvn quarkus:dev -pl kaap -am -Pskip-crds

Deploy the operator and the Pulsar cluster

Check at helm/examples to get some ideas about how to deploy a Pulsar cluster. Note that the k3s registry DOES NOT use your host docker registry, so every image will be downloaded from upstream. The k3s cluster already contains the dev version of the operator, named datastax/kaap:latest-dev. The k3s cluster also contains a Pulsar/LunaStreaming image to use.

Integration tests

Integration tests under tests spin up a K3s cluster in docker. In order to troubleshoot a failure test, sometimes it’s useful to use your own cluster to run the tests. All the tests create a new temporary namespace but some cluster-wide resources are created and deleted, so be careful.

In order to run a test targeting the current kube context:

mvn -pl tests test -Dtest='CRDsTest' -Prun-tests-current-context

If you’re using a public cluster (e.g. GCP, EKS) you have to deploy the operator image to a public Docker registry. You can also use a specific StorageClass for the test by setting ´kaap.tests.existingenv.storageclass`.

To run a test in GCP:

docker push <operator-image>
gcloud container clusters get-credentials gcp-cluster # or set it as current context with kubectl

mvn -pl tests test -Dtest='CRDsTest' -Prun-tests-current-context \
    -Dkaap.tests.operator.image=<operator-image>