Kubernetes projects require that you sign a Contributor License Agreement (CLA) before we can accept your pull requests. Please see https://git.k8s.io/community/CLA.md for more info
If you are new to CSI, please go through the CSI Spec and General CSI driver development guideline to get some basic understanding of CSI driver before you start.
- (Optional, but recommended for large changes) Submit an issue describing your proposed change to the repo in question.
- If your proposed change is accepted, and you haven't already done so, sign a Contributor License Agreement (see details above).
- Fork the desired repo, develop and test your code changes (see "Development Quickstart Guide" below).
- When your changes are complete - including tests and documentation if necessary, submit a pull request.
- Reviewer(s) from the repo owners will review your PR. Communicate with your reviewers and address their feedback as required.
- After merge, your change will typically be included in the next minor release of the EBS CSI Driver and associated Helm chart, which is performed monthly.
This guide demonstrates the basic workflow for developing for the EBS CSI Driver. More detailed documentation of the available make
targets is available in the Makefile documentation.
If your changes are Helm-only, skip to section 2 (manual testing) below after making your changes.
Run make
to build a driver binary, and thus discover compiler/syntax errors.
When your change is ready for testing, run make test
to execute the unit test suite for the driver. If you are making a significant change to the driver (such as a bugfix or new feature), add new unit tests or update existing tests where applicable.
When your change is ready to test in a real Kubernetes cluster, create a kops
cluster to test the driver against:
make cluster/create
If your change affects the Windows implementation of the driver, instead create an eksctl
cluster with Windows nodes:
export WINDOWS="true"
# Note: CLUSTER_TYPE must be set for all e2e tests and cluster deletion
# Re-export it if necessary (for example, if running tests in a separate terminal tab)
export CLUSTER_TYPE="eksctl"
make cluster/create
Export the KUBECONFIG
environment variable to access the cluster for manual testing:
eval "$(make cluster/kubeconfig)"
To test a change, build the driver image and then install it:
make cluster/image
# If testing a Helm-based change, you can use HELM_EXTRA_FLAGS
# to set your new paremeters for testing, for example:
# export HELM_EXTRA_FLAGS="--set=controller.newParameter=true"
make cluster/install
When you are finished manually testing, uninstall the driver:
make cluster/uninstall
Before running any E2E tests, you must create a cluster (see step 2) and build an image if you have not already done so:
make cluster/image
The driver should not be preinstalled when running E2E tests (the tests will automatically install the driver). If it is, uninstall it first:
make cluster/uninstall
If you are making a change to the driver, the recommended test suite to run is the external tests:
# Normal external tests (excluding Windows tests)
make e2e/external
# Additionally, run the windows tests if changing Windows behavior
make e2e/external-windows
If you are making a change to the Helm chart, the recommended test suite to run is the Helm ct
tests:
make e2e/helm-ct
After finishing testing, cleanup your cluster :
make cluster/delete
Run make update
to automatically format go source files and re-generate automatically generated files. If make update
produces any changes, commit them before submitting a PR.
Run make verify
to run linters and other similar code checking tools. Fix any issues make verify
discovers before submitting a PR.
Your PR should contain a full set of unit/E2E tests when applicable. If introducing a change that is not exercised fully by automated testing (such as a new Helm parameter), please provide evidence of the feature working as intended in the PR description.