Developing Lagoon

Development of Lagoon locally can now be performed on a local Kubernetes cluster, or via Docker Compose (as a fallback).

Docker

Docker must be installed to build and run Lagoon locally.

Install Docker and Docker Compose

Please check the official docs for how to install Docker.

Docker Compose is included in Docker for Mac installations. For Linux installations see the directions here.

Configure Docker

You will need to update your insecure registries in Docker. Read the instructions here on how to do that. We suggest adding the entire local IPv4 Private Address Spaces to avoid unnecessary reconfiguration between Kubernetes and Docker Compose. e.g. "insecure-registries" : ["172.16.0.0/12","192.168.0.0/16"],

Allocate Enough Docker Resources

Running a Lagoon, Kubernetes, or Docker cluster on your local machine consumes a lot of resources. We recommend that you give your Docker host a minimum of 8 CPU cores and 12GB RAM.

Build Lagoon Locally

Only consider building Lagoon this way if you intend to develop features or functionality for it, or want to debug internal processes. We will also be providing instruction to install Lagoon without building it (i.e. by using the published releases).

We're using make (see the Makefile) in order to build the needed Docker images, configure Kubernetes and run tests.

We have provided a number of routines in the Makefile to cover most local development scenarios. Here we will run through a complete process.

Build images

  1. Here -j8 tells make to run 8 tasks in parallel to speed the build up. Adjust as necessary.

  2. We have set SCAN_IAMGES=false as a default to not scan the built images for vulnerabilities. If set to true, a scan.txt file will be created in the project root with the scan output.

make -j8 build
  1. Start Lagoon test routine using the defaults in the Makefile (all tests).

make kind/test

There are a lot of tests configured to run by default - please consider only testing locally the minimum that you need to ensure functionality. This can be done by specifying or removing tests from the TESTS variable in the Makefile.

This process will:

  1. Download the correct versions of the local development tools if not installed - kind, kubectl, helm, jq.

  2. Update the necessary Helm repositories for Lagoon to function.

  3. Ensure all of the correct images have been built in the previous step.

  4. Create a local KinD cluster, which provisions an entire running Kubernetes cluster in a local Docker container. This cluster has been configured to talk to a provisioned image registry that we will be pushing the built Lagoon images to. It has also been configured to allow access to the host filesystem for local development.

  5. Clone Lagoon from https://github.com/uselagoon/lagoon-charts (use the CHARTS_TREEISH variable in the Makefile to control which branch if needed).

  6. Install the Harbor Image registry into the KinD cluster and configure its ingress and access properly.

  7. Docker will push the built images for Lagoon into the Harbor image registry.

  8. It then uses the Makefile from lagoon-charts to perform the rest of the setup steps.

  9. A suitable ingress controller is installed - we use the NGINX Ingress Controller.

  10. A local NFS server provisioner is installed to handle specific volume requests - we use one that handles Read-Write-Many operations (RWX).

  11. Lagoon Core is then installed, using the locally built images pushed to the cluster-local Image Registry, and using the default configuration, which may exclude some services not needed for local testing. The installation will wait for the API and Keycloak to come online.

  12. The DBaaS providers are installed - MariaDB, PostgreSQL and MongoDB. This step provisions standalone databases to be used by projects running locally, and emulates the managed services available via cloud providers (e.g. Cloud SQL, RDS or Azure Database).

  13. Lagoon Remote is then installed, and configured to talk to the Lagoon Core, databases and local storage. The installation will wait for this to complete before continuing.

  14. To provision the tests, the Lagoon Test chart is then installed, which provisions a local Git server to host the test repositories, and pre-configures the Lagoon API database with the default test users, accounts and configuration. It then performs readiness checks before starting tests.

  15. Lagoon will run all the tests specified in the TESTS variable in the Makefile. Each test creates its own project & environments, performs the tests, and then removes the environments & projects. The test runs are output to the console log in the lagoon-test-suite-* pod, and can be accessed one test per container.

Ideally, all of the tests pass and it's all done!

View the test progress and your local cluster

The test routine creates a local Kubeconfig file (called kubeconfig.kind.lagoon in the root of the project, that can be used with a Kubernetes dashboard, viewer or CLI tool to access the local cluster. We use tools like Lens, Octant, kubectl or Portainer in our workflows. Lagoon Core, Remote and Tests all build in the Lagoon namespace, and each environment creates its own namespace to run, so make sure to use the correct context when inspecting.

In order to use kubectl with the local cluster, you will need to use the correct Kubeconfig. This can be done for every command or it can be added to your preferred tool:

KUBECONFIG=./kubeconfig.kind.lagoon kubectl get pods -n lagoon

The Helm charts used to build the local Lagoon are cloned into a local folder and symlinked to lagoon-charts.kind.lagoon where you can see the configuration. We'll cover how to make easy modifications later in this documentation.

Interact with your local Lagoon cluster

The Makefile includes a few simple routines that will make interacting with the installed Lagoon simpler:

make kind/port-forwards

This will create local ports to expose the UI (6060), API (7070) and Keycloak (8080). Note that this logs to stdout, so it should be performed in a secondary terminal/window.

make kind/get-admin-creds

This will retrieve the necessary credentials to interact with the Lagoon.

  • The JWT is an admin-scoped token for use as a bearer token with your local GraphQL client. See more in our GraphQL documentation.

  • There is a token for use with the "admin" user in Keycloak, who can access all users, groups, roles, etc.

  • There is also a token for use with the "lagoonadmin" user in Lagoon, which can be allocated default groups, permissions, etc.

make kind/dev

This will re-push the images listed in KIND_SERVICES with the correct tag, and redeploy the lagoon-core chart. This is useful for testing small changes to Lagoon services, but does not support "live" development. You will need to rebuild these images locally first, e.g rm build/api && make build/api.

make kind/local-dev-patch

This will build the typescript services, using your locally installed Node.js (it should be >16.0). It will then:

  • Mount the "dist" folders from the Lagoon services into the correct lagoon-core pods in Kubernetes

  • Redeploy the lagoon-core chart with the services running with nodemonwatching the code for changes

  • This will facilitate "live" development on Lagoon.

  • Note that occasionally the pod in Kubernetes may require redeployment for a change to show. Clean any build artifacts from those services if you're rebuilding different branches with git clean -dfx as the dist folders are ignored by Git.

make kind/local-dev-logging

This will create a standalone OpenDistro for Elasticsearch cluster in your local Docker, and configure Lagoon to dispatch all logs (Lagoon and project) to it, using the configuration in lagoon-logging.

make kind/retest
# OR
make kind/retest TESTS='[features-kubernetes]'

This will re-run a suite of tests (defined in the TESTS variable) against the existing cluster. It will re-push the images needed for tests (tests, local-git, and the data-watcher-pusher). You can specify tests to run by passing the TESTS variable inline.

If updating a test configuration, the tests image will need to be rebuilt and pushed, e.g rm build/tests && make build/tests && make kind/push-images IMAGES='tests' && make kind/retest TESTS='[api]'

make kind/push-images
# OR
make kind/push-images IMAGES='tests local-git'

This will push all the images up to the image registry. Specifying IMAGES will tag and push specific images.

make kind/clean

This will remove the KinD Lagoon cluster from your local Docker.

Ansible

The Lagoon test uses Ansible to run the test suite. Each range of tests for a specific function has been split into its own routine. If you are performing development work locally, select which tests to run, and update the $TESTS variable in the Makefile to reduce the concurrent tests running.

The configuration for these tests is held in three services:

  • tests is the Ansible test services themselves. The local testing routine runs each individual test as a separate container within a test-suite pod. These are listed below.

  • local-git is a Git server hosted in the cluster that holds the source files for the tests. Ansible pulls and pushes to this repository throughout the tests

  • api-data-watcher-pusher is a set of GraphQL mutations that pre-populates local Lagoon with the necessary Kubernetes configuration, test user accounts and SSH keys, and the necessary groups and notifications. Note that this will wipe local projects and environments on each run.

The individual routines relevant to Kubernetes are:

  • active-standby-kubernetes runs tests to check active/standby in Kubernetes.

  • api runs tests for the API - branch/PR deployment, promotion.

  • bitbucket, gitlab and github run tests for the specific SCM providers.

  • drupal-php74 runs a single-pod MariaDB, MariaDB DBaaS and a Drush-specific test for a Drupal 8/9 project (drupal-php73 doesn't do the Drush test).

  • drupal-postgres runs a single-pod PostgreSQL and a PostgreSQL DBaaS test for a Drupal 8 project.

  • elasticsearch runs a simple NGINX proxy to an Elasticsearch single-pod.

  • features-api-variables runs tests that utilize variables in Lagoon.

  • features-kubernetes runs a range of standard Lagoon tests, specific to Kubernetes.

  • features-kubernetes-2 runs more advanced kubernetes-specific tests - covering multi-project and subfolder configurations.

  • nginx, node and python run basic tests against those project types.

  • node-mongodb runs a single-pod MongoDB test and a MongoDB DBaaS test against a Node.js app.

There are a few other legacy Openshift-specific tests in there that may or may not work with Openshift-based clients.

Local Development

Most services are written in Node.js. As many of these services share similar Node.js code and Node.js packages, we're using a feature of Yarn, called Yarn workspaces. Yarn workspaces need a package.json in the project's root directory that defines the workspaces.

The development of the services can happen directly within Docker. Each container for each service is set up in a way that its source code is mounted into the running container (see docker-compose.yml). Node.js itself is watching the code via nodemon , and restarts the Node.js process automatically on a change.

lagoon-commons

The services not only share many Node.js packages, but also share actual custom code. This code is within node-packages/lagoon-commons. It will be automatically symlinked by Yarn workspaces. Additionally, the nodemon of the services is set up in a way that it checks for changes in node-packages and will restart the node process automatically.

Troubleshooting

I can't build a docker image for any Node.js based service

Rebuild the images via:

make clean
make build

I get errors about missing node_modules content when I try to build / run a Node.js based image

Make sure to run yarn in Lagoon's root directory, since some services have common dependencies managed by yarn workspaces.

I get an error resolving the nip.io domains

Error response from daemon: Get https://registry.172.18.0.2.nip.io:32080/v2/: dial tcp: lookup registry.172.18.0.2.nip.io: no such host

This can happen if your local resolver filters private IPs from results. You can work around this by editing /etc/resolv.conf and adding a line like nameserver 8.8.8.8 at the top to use a public resolver that doesn't filter results.

Example workflows

Here are some development scenarios and useful workflows for getting things done.

Editing kubectl-build-deploy-dind

This example shows a workflow for editing the Lagoon deploy logic.

Edit kubectl-build-deploy-dind

In this example we want to add some functionality to the Lagoon deploy logic in the kubectl-build-deploy-dind image.

  1. Start a local KinD cluster with Lagoon installed from locally built images, and smoke-test it by running a single test suite:

make -j8 kind/test TESTS='[features-api-variables]'
  1. Edit images/kubectl-build-deploy-dind/build-deploy-docker-compose.sh.

--- a/images/kubectl-build-deploy-dind/build-deploy-docker-compose.sh
+++ b/images/kubectl-build-deploy-dind/build-deploy-docker-compose.sh
@@ -1,5 +1,7 @@
#!/bin/bash
+echo HELLO WORLD
+
function cronScheduleMoreOftenThan30Minutes() {
#takes a unexpanded cron schedule, returns 0 if it's more often that 30 minutes
MINUTE=$(echo $1 | (read -a ARRAY; echo ${ARRAY[0]}) )
  1. Now rebuild the kubectl-build-deploy-dind image with the edits included.

rm build/kubectl-build-deploy-dind
make -j8 build/kubectl-build-deploy-dind
  1. Push the newly built image into the cluster registry. It will now be used for future deploys.

make kind/push-images IMAGES=kubectl-build-deploy-dind
  1. Rerun the tests.

make kind/retest TESTS='[features-api-variables]'
  1. See the edits have been applied.

$ kubectl -n ci-features-api-variables-control-k8s-lagoon-api-variables logs lagoon-build-lat2b | grep -A2 build-deploy-docker-compose.sh
+ . /kubectl-build-deploy/build-deploy-docker-compose.sh
++ echo HELLO WORLD
HELLO WORLD

Add tests

  1. Repeat the first step above.

  2. Edit tests/tests/features-api-variables.yaml and add a test case.

  3. Rebuild the tests image.

rm build/tests
make -j8 build/tests
  1. Push the new tests image into the cluster registry.

make kind/push-images IMAGES=tests
  1. Rerun the tests.

make kind/retest TESTS='[features-api-variables]'