Thank you for investing your time in contributing to our project.
The operator uses unit tests and chainsaw for e2e tests to make sure that the operator is working as intended, we use make to generate a number of docs and scripts for us.
The operator embeds grafonnet-lib inside embeds/grafonnet-lib. It used to be a git submodule, however, as that folder was only filled when the git submodule was initiated, the grafana-operator couldn't be imported externally.
NOTE: please, run make all before opening a PR to make sure your changes are compliant with our standards and all automatically generated files (like CRDs) are up-to-date.
We use a number of code standards in the project that we apply using a number of different tools. As a part of the CI solution these settings will be validated, but all of them can be tested using the Makefile before pushing.
Before pushing any code we recommend that you run the following make commands.
make test
make code/golangci-lintDepending on what you have changed these commands will update a number of different files.
To speed up multi-arch image builds and avoid burden of maintaining a Dockerfile, we are using ko.
For more information on how to push the container image to your remote repository take a look at the official docs.
Some of us use kind some use crc, below you can find an example on how to integrate with a kind cluster.
When adding a grafanadashboard to our grafana instances through the operator and using make test to run the operator we need a way to send data in to the grafana instance.
There are multiple ways of doing so but this is one of them using kind.
There is a make target start-kind which will start a new instance of a
kind cluster
with dependencies (currently only ingress-nginx) and install the CRDs.
Two namespaces will be used, default and grafana-crds (name can be customized with CRD_NS environment variable).
Any resource placed in hack/kind/resources/[default|crd-ns]/ will be installed in the respective namespace.
Notice the spec.client.preferIngress: true in the hack/kind/resources/default/grafana.yaml,
which makes the grafana client in the operator to use the ingress address instead of the service name
and should only be used during development.
It uses the hostname created using nip.io which will steer traffic to your local deployment
through a DNS response (e.g. nslookup grafana.127.0.0.1.nip.io will respond with 127.0.0.1).
make start-kindYou should now be ready to develop the operator.
make runTo interact wit the cluster through kubectl you can ether run kind export kubeconfig -n kind-grafana
or follow the instructions in the output from the script.
As mentioned above we use chainsaw to run e2e tests for the operator, we normally run chainsaw on Kind
The make e2e command will
# Build the container
make ko-build-kind
# Create grafana-operator-system namespace
kubectl create ns grafana-operator-system
# Run the chainsaw tests
VERSION=latest make e2eWe support helm as a deployment solution and it can be found under deploy/helm.
The grafana-operator helm chart is currently manually created. When CRD:s is upgraded the helm chart will also get an update.
But if you generate new RBAC rules or create new deployment options for the operator you will need to add them manually.
Chart.yaml appVersion follows the grafana-operator version but the helm chart is versioned separately.
If you add update the chart don't forget to run make helm/docs, which will update the helm specific README file.
Kustomize, amongst other methods, can be used to install the operator.
To make sure its copy of CRDs is up-to-date, please, run make all.
NOTE: there's an option to call make kustomize-crd directly, though it assumes the CRDs have already been generated from Go code through make manifests. make all is just a simpler option to run since it doesn't require you to remember all previous steps.
All documentation is stored under docs, for our homepage we are using hugo and the docsy theme.
If you don't feel the needs to preview the changes that you have done to the docs all you need to do is to edit them. But if you want to see how your changes will look you need to install hugo locally, see Hugo