Upgrade Advisory

This documentation is for Flux (v1) and Helm Operator (v1). Both projects are in maintenance mode and will soon reach end-of-life. We strongly recommend you familiarise yourself with the newest Flux and start looking at your migration path.

For documentation regarding the latest Flux, please refer to this section.

Get started developing

This guide shows a workflow for making a small (actually, tiny) change to Flux, building and testing that change locally.


From a very high level, there are at least 3 ways you can develop on Flux once you have your environment set up:

  1. The “minimalist” approach (only requires and kubectl):

    1. make
    2. copy the specific image tag (e.g. docker.io/fluxcd/flux:master-a86167e4) for what you just built and paste it into /deploy/flux-deployment.yaml as the image you’re targeting to deploy
    3. deploy the resources in /develop/*.yaml manually with kubectl apply
    4. make a change to the code
    5. see your code changes have been deployed
    6. repeat
  2. Use freshpod to deploy changes to the /deploy directory resources:

    1. make
    2. make a change to the code
    3. see your changes have been deployed
    4. repeat
  3. Remote cluster development approach:

    1. ensure local kubectl access to a remote Kubernetes cluster
    2. have an available local memcached instance
    3. make a change to the code
    4. go run cmd/fluxd/main.go \
          --memcached-hostname localhost  \
          --memcached-port 11211 \
          --memcached-service "" \
          --git-url git@github.com:fluxcd/flux-get-started \

This guide covers approaches 1 and 2 using minikube. freshpod is superseded by Skaffold and is generally the future. That said, freshpod is very simple to use and reason about (and is still well supported by minikube) which is why it’s used in this guide.

Run fluxcd/flux-getting-started

We’re going to make some changes soon enough, but just to get a good baseline please follow the “Get started with Flux” tutorial and run the fluxcd/flux-getting-started repo through its normal paces.

Now that we know everything is working with flux-getting-started, we’re going to try and do nearly the same thing as flux-getting-started, except instead of using official releases of flux, we’re going to build and run what we have locally.

Prepare your environment

  1. Install the prerequisites. This guide is written from running Linux, but the same instructions will generally apply to OSX. Although everything you need has been known to work independently in Windows from time to time, results may vary.

  2. Configure your environment so you can run tests. Run:

    make test
  3. We want to make sure we’re starting fresh. Tell minikube to clear any previously running clusters:

    minikube delete
  4. The minikube addon called freshpod that will be very useful to us later. You’ll see. It’s gonna be cool.

    minikube addons enable freshpod
  5. This part is really important. You’re going to set some environment variables which will intercept any images pulled by docker. Run minikube docker-env to see what we’re talking about. You’ll get an output that shows you what the script is doing. Thankfully, it’s not terribly complicated - it just sets some environment variables which will allow minikube to man-in-the-middle the requests Kubernetes makes to pull images. It will look something like this:

    export DOCKER_TLS_VERIFY="1"
    export DOCKER_HOST="tcp://"
    export DOCKER_CERT_PATH="/home/fluxrulez/.minikube/certs"
    export DOCKER_API_VERSION="1.35"
    # Run this command to configure your shell:
    # eval $(minikube docker-env)

    So, as the script suggests, run the following command:

    eval $(minikube docker-env)

    Now, be warned. These are local variables. This means that if you run this eval in one terminal and then switch to another for later when we build the Flux project, you’re gonna hit some issues. For one, you’ll know it isn’t working because Kubernetes will tell you that it can’t pull the image when you run kubectl get pods:

    NAME                        READY   STATUS         RESTARTS   AGE
    flux-7f6bd57699-shx9v       0/1     ErrImagePull   0          35s

Prepare the repository

  1. Fork the repo on GitHub.

  2. Clone git@github.com:<YOUR-GITHUB-USERNAME>/flux.git replacing <YOUR-GITHUB-USERNAME> with your GitHub username.

    In the same terminal you ran eval $(minikube docker-env), run GO111MODULE=on go mod download followed by make from the root directory of the Flux repo. You’ll see docker’s usual output as it builds the image layers. Once it’s done, you should see something like this in the middle of the output:

    Successfully built 606610e0f4ef
    Successfully tagged docker.io/fluxcd/flux:latest
    Successfully tagged docker.io/fluxcd/flux:master-a86167e4

    This confirms that a new docker image was tagged for your image.

  3. Open up deploy/flux-deployment.yaml and update the image at spec.template.spec.containers[0].image to be simply docker.io/fluxcd/flux. While we’re here, also change the --git-url to point towards your fork. It will look something like this in the YAML:

          - name: flux
            image: docker.io/fluxcd/flux
            imagePullPolicy: IfNotPresent
              - --git-url=git@github.com:<YOUR-GITHUB-USERNAME>/flux-getting-started
              - --git-branch=master
  4. We’re ready to apply your newly-customized deployment! Since kubectl will apply all the Kubernetes manifests it finds (recursively) in a folder, we simply need to pass the directory to kubectl apply:

    kubectl apply --filename ./deploy

    You should see an output similar to:

    serviceaccount/flux created
    clusterrole.rbac.authorization.k8s.io/flux created
    clusterrolebinding.rbac.authorization.k8s.io/flux created
    deployment.apps/flux created
    secret/flux-git-deploy created
    deployment.apps/memcached created
    service/memcached created
    secret/flux-git-deploy configured

    Congrats you just deployed your local Flux to your default namespace. Check that everything is running:

    kubectl get pods --selector=name=flux

    You should get an output that looks like:

    NAME                   READY   STATUS    RESTARTS   AGE
    flux-6f7fd5bbc-hpq85   1/1     Running   0          38s

    If (instead) you see that Ready is showing 0/1 and/or the status is ErrImagePull double back on the instructions and make sure you did everything correctly and in order.

  5. Pull the logs for your “fresh off of master” copy of Flux that you just deployed locally to minikube:

    kubectl logs --selector=name=flux

    You should see an output that looks something like this:

    ts=2019-02-28T18:58:45.091531939Z caller=warming.go:268 component=warmer info="refreshing image" image=docker.io/fluxcd/flux tag_count=60 to_update=60 of_which_refresh=0 of_which_missing=60
    ts=2019-02-28T18:58:46.233723421Z caller=warming.go:364 component=warmer updated=docker.io/fluxcd/flux    successful=60 attempted=60
    ts=2019-02-28T18:58:46.234086642Z caller=images.go:17 component=sync-loop msg="polling images"
    ts=2019-02-28T18:58:46.234125646Z caller=images.go:27 component=sync-loop msg="no automated services"
    ts=2019-02-28T18:58:46.749598558Z caller=warming.go:268 component=warmer info="refreshing image" image=memcached    tag_count=66 to_update=66 of_which_refresh=0 of_which_missing=66
    ts=2019-02-28T18:58:51.017452675Z caller=warming.go:364 component=warmer updated=memcached successful=66 attempted=66
    ts=2019-02-28T18:58:51.020061586Z caller=images.go:17 component=sync-loop msg="polling images"
    ts=2019-02-28T18:58:51.020113243Z caller=images.go:27 component=sync-loop msg="no automated services"

Make some changes

  1. Now for the part you’ve been waiting for! We’re going to make a cosmetic change to our local copy of Flux. Navigate to git/operations.go. In it, you will find a private function to this package that goes by the name execGitCmd. Paste the following as the (new) first line of the function:

  2. Run make again. Once this finishes you can check on your running pods with:

    kubectl get pods --selector=name=flux

    Keep your eye on the AGE column. It should be just a few seconds old if you check out the AGE column:

    NAME                   READY   STATUS    RESTARTS   AGE
    flux-6f7fd5bbc-6j9d5   1/1     Running   0          10s

    This pod was deployed even though we didn’t run any kubectl commands or interact with Kubernetes directly because of the freshpod minikube addon that we enabled earlier. Freshpod saw that a new Docker image was tagged for docker.io/fluxcd/flux:latest and it went ahead and redeployed that pod for us.

    Consider that simply applying the flux-deployment.yaml file again wouldn’t do anything since the actual image we’re targeting (which is actually docker.io/fluxcd/flux with no :latest tag, but it’s the same difference) hasn’t changed. The Kubernetes api server will get that JSON request from kubectl and go: “right… so nothing has changed in the file so I have nothing to do… IGNORE!”.

    There is another way to do this, of course. Remember that before when we ran make that we did also get an image tagged with the :<branch>-<commit hash> syntax (in our specific example above it was :master-a86167e4). We could, in theory, grab that tag every time we make, and then paste it into spec.template.spec.containers[0].image of our deployment. That’s tedious and error prone. Instead, freshpod cuts this step out for us and accomplishes the same end goal.

  3. Check the logs again (with kubectl logs --selector=name=flux) to find that your obnoxious chain of Zs is present.


You have now modified Flux and deployed that change locally. From here on out, you simply need to run make after you save your changes and wait a few seconds for your new pod to be deployed to minikube. Keep in mind, that (as in the situation where you run make without saving any changes) if the Docker image you pointed to in the Kubernetes deployment for Flux is not Successfully tagged, freshpod won’t have anything new to deploy. Other than that, you should be good to go!