This code is related to this blog post:
This project demonstrates how one could use the Jenkins Pipeline/Workflow plugin and Docker to create containerized build environments. Please note that NOT everything in this project works as desired. It's an absolute work-in-progress.
The Jenkins master is not set up to save state. To start it, do the following:
cd jenkins-master && docker-compose up
The Jenkins master will be available at http://192.168.99.100:8080 . Before bringing up the agent, there's a few steps to setting up Jenkins. First, set up security so that users have to login and enable the JNLP agents.
Next, go to "Manage Jenkins" and look for "Manage Nodes" and create a new slave called "test-agent"
Setup the new agent with a remore directory of /var/jenkins_home
and set the launch method to JNLP:
Once the node is created, click agent name to reveal the -secret
:
Copy this value so you can pass it to the agent.
Execute the following:
cd jenkins-jnlp-agent
export JENKINS_SECRET="the-value-you-copied-earlier" | docker-compose up
You should now see the agent on the main screen.
The Jobs folder contains 3 Groovy scripts that define different pipelines. These jobs are here to represent what is working and what isn't. Here's what they do and the expected outcomes of each:
jobs/sans-docker.groovy
: Executes the build in on the test agent container without running inside a docker container. This pipeline works and the project build successfully.jobs/docker-inside.groovy
: This is my ideal state but this doesn't work at all and this will hang indefinately due to the fact that the build container can't see the volume. This pipeline does work if the agent is not in a container.jobs/docker-in-shell.groovy
: This pipeline demonstrates that it is possible to getdocker.inside()
to work if using a data volume container. The pipeline partially works and it can see code, but the build fails on updating the git submodule.
I didn't have a great way providing a Jenkins master with pre-configured jobs. Thus, you'll have to create each job manually.