This document describes what actually happens during a Lagoon build and deployment. It is heavily simplified from what actually happens, but it will help you to understand what is happening under the hood every time that Lagoon deploys new code for you.
Watch the video below for a walk-through of the deployment process.
First, Lagoon checks if the OpenShift project/Kubernetes namespace for the given environment exists and is correctly set up. It will make sure that we have the needed service accounts, create secrets, and will configure environment variables into a ConfigMap
lagoon-env which is filled with information like the environment type and name, the Lagoon project name, and so on.
Next, Lagoon will check out your code from Git. It needs that to be able to read the
docker-compose.yml and any
.env files, but also to build the Docker images.
Note that Lagoon will only process these actions if the branch/PR matches the branch regex set in Lagoon. Based on how the deployment has been triggered, different things will happen:
If the deployment is triggered automatically via a Git webhook and is for a single branch, Lagoon will check out the Git SHA which is included in the webhook payload. This will trigger a deployment for every Git SHA pushed.
If you trigger a branch deployment manually via the REST API (via the UI, or GraphQL) and do NOT define a
SHA in the POST payload, Lagoon will just check out the latest commit in that branch and deploy it.
If the deployment is a pull request (PR) deployment, Lagoon will load the base and the HEAD branch and SHAs for the pull request and will:
Check out the base branch (the branch the PR points to).
HEAD branch (the branch that the PR originates from) on top of the base branch.
Lagoon will check out and merge particular SHAs which were sent in the webhook. Those SHAs may or may not point to the branch heads. For example, if you make a new push to a GitHub pull request, it can happen that SHA of the base branch will not point to the current base branch HEAD.
If the merge fails, Lagoon will also stop and inform you about this.
For each service defined in the
docker-compose.yml Lagoon will check if images need to be built or not. If they need to be built, this will happen now. The order of building is based on the order they are configured in
docker-compose.yml , and some build arguments are injected:
LAGOON_SSH_PRIVATE_KEY - The SSH private key that is used to clone the source repository. Use
RUN /lagoon/entrypoints/05-ssh-key.sh to convert the build argument into an actual key at
/home/.ssh/key which will be used by SSH and Git automatically. For safety, remove the key again via
RUN rm /home/.ssh/key.
LAGOON_GIT_SOURCE_REPOSITORY - The full Git URL of the source repository.
Also, if this is a pull request build:
Additionally, for each already built image, its name is also injected. If your
docker-compose.yml is configured to first build the
cli image and then the
nginx image, the name of the
nginx image is injected as
Next, Lagoon will configure OpenShift/Kubernetes with all services and routes that are defined from the service types, plus possible additional custom routes that you have defined in
In this step it will expose all defined routes in the
LAGOON_ROUTES as comma separated URLs. It will also define one route as the "main" route, in this order:
If custom routes defined: the first defined custom route in
The first auto-generated route from a service defined in
The "main" route is injected via the
LAGOON_ROUTE environment variable.
Now it is time to push the previously built Docker images into the internal Docker image registry.
For services that didn't specify a Dockerfile to be built in
docker-compose.yml and only gave an image, they are also tagged and will cause the internal Docker image registry to know about the images, so that they can be used in containers.
Lagoon will now create persistent storage (PVC) for each service that needs and requested persistent storage.
For each service that requests a cron job (like MariaDB), plus for each custom cron job defined in
.lagoon.yml, Lagoon will now generate the cron job environment variables which are later injected into the Deployment.
Now Lagoon will check the
.lagoon.yml file for defined tasks in
pre-rollout and will run them one by one in the defined services. Note that these tasks are executed on the pods currently running (so cannot utilize features or scripts that only exist in the latest commit) and therefore they are also not run on first deployments.
If any of them fail, Lagoon will immediately stop and notify you, and the rollout will not proceed.
This is probably the most important step. Based on the defined service type, Lagoon will create the Deployment, Statefulset or Daemonsets for the service. (Note that Deployments are analogous DeploymentConfigs in OpenShift)
It will include all previously gathered information like the cron jobs, the location of persistent storage, the pushed images and so on.
Creation of these objects will also automatically cause OpenShift/Kubernetes to trigger new deployments of the pods if necessary, like when an environment variable has changed or an image has changed. But if there is no change, there will be no deployment! This means if you only update the PHP code in your application, the Varnish, Solr, MariaDB, Redis and any other service that is defined but does not include your code will not be deployed. This makes everything much much faster.
Now Lagoon waits! It waits for all of the just-triggered deployments of the new pods to be finished, as well as for their health checks to be successful.
If any of the deployments or health checks fail, the deployment will be stopped here, and you will be informed via the defined notification systems (like Slack) that the deployment has failed.
Now Lagoon will check the
.lagoon.yml file for defined tasks in
post-rollout and will run them one by one in the defined services.
If any of them fail, Lagoon will immediately stop and notify you.
If all went well and nothing threw any errors, Lagoon will mark this build as successful and inform you via defined notifications. ✅
There may be a case where you want to push without a deployment. Make sure your commit message contains "
[skip deploy]" or "
[deploy skip]" and Lagoon will not trigger a deployment from that commit.