Bamboo Agent using Docker

This repository contains a Dockerfile to build a fully-fledged, self-contained Bamboo Agent image targeted at Java-/Web-based development.

It currently features:

• JDK 1.8 and 1.10, Groovy,

• Node 4.x, 6.x and 8.x

• Maven 3, Ant 1.8, 1.9 and 1.10

• (Headless) chromium browser and chromedriver

• grunt-cli, yarn and npm

• IntelliJ IDEA 2018.1 and idea-cli-inspector

It installs a set of dependencies and declares them under custom capability names. So overall you probably want to (heavily) adjust/trim/expand this image according to your needs.

Building the image

Simply run the following command. Please note: The prefix myregistry.mycorp.com:5000 is there assuming that you have a private registry where you will publish you images for the clients to pick it up. Simply run

sudo docker build --pull -t myregistry.mycorp.com:5000/bentolor-bamboo-agent .

then you can push the newly build image with the following command

sudo docker push myregistry.mycorp.com:5000/bentolor-bamboo-agent

into your registry running on myregistry.mycorp.com:5000.

NOTE

Please note, that these images tend to get quite large. So probably you want be able to delete outdated images from you registry. This can get a little complicated, so you might be aware upfront how to do this.

Running a new Agent

Using out published docker image it’s easy to start new instances of the Docker ("Container") as follows:

sudo docker run -d --cap-add SYS_ADMIN --init --name bentolor-bamboo-agent -h docker-host1 -e BAMBOO_SERVER=https://bamboo.mycorp.com:1234/ myregistry.mycorp.com:5000/bentolor-bamboo-agent

The hostname is what you will see later-on in the Bamboo UI. If you do not pass any value here, Bamboo will generate a random ID on every container creation.

INFO

The option --cap-add SYS_ADMIN is only necessary for (Headless) operation of the Chromium browser. Otherwise chromium will crash, as it will fail to create a sandbox with Failed to move to new namespace: PID namespaces supported, Network namespace supported, but failed: errno = Operation not permitted The --init advises Docker to start a dedicated init process within the containers, so that processes whose parent died get reap. Otherwise you could see lot of defunct zombie processes i.e. of chrome <defunct> or node <defunct>.

Controlling the agent

You can start and stop the container with the usual docker commands:

sudo docker stop bentolor-bamboo-agent
sudo docker start bentolor-bamboo-agent

Stop and delete the instance with the following command

sudo docker kill bentolor-bamboo-agent
sudo docker rm bentolor-bamboo-agent

How to (automatically) redeploy an updated image

One could use some big-class solutions like Kubernetes or OpenShift of the container orchistration. As a typically build agent will probably run on a limited set of bare metal computers (probably some old laptops & PCs no longer in use) I find a smaller solution better. I found watchtower to be the most intriguing solution sof this purpose: It can periodically check for updates of the current running containers and the automatically stop, pull and restart them with the exact same options as before.

It’s as simply as:

sudo docker run -d \
    --name watchtower \
    -v /var/run/docker.sock:/var/run/docker.sock \
    v2tec/watchtower \
    --schedule "0 0-59/3 9-19 * * MON-FRI" \
    --cleanup

This will check every 3 minutes between 9am and 7pm between monday and friday for updates and then redeploy your agent (and other images).

See the manual of watchtower to learn more about the options you have here.

Maintaining the agent

Simply adjust & update the Dockerfile and bamboo-capabilities.properties accordingly.

IDEA IntelliJ Handling

The Dockerfile downloads and installs IntelliJ IDEA as .tar.gz. To get it running (headlessly) IntelliJ needs some sort of minimum configuration like required plugins or licence server definition.

Therefore you’ll find a directory bamboo/.IntelliJIdea2018.1 which gets copied into the container.

Creating the minimal configuration from scratch

1. Start and definde licenceserver

2. Define JDKs within IDEA (important!)

3. Add optional plugins like handlebars, js-karma, NodeJS, Scala, VueJS, …

4. Hand-pick the required files in bamboo/.IntelliJIdea2018.1 and add them to Git

5. Delete all unversionated files (i.e. via git clean -fdx)

Interactively running IDEA inside the container for this purpose

To create or maintain the first configuration from a Linux-based host I simply forward my X11-connection into a newly started host and do the necessary adjustemts interactivly via:

xhost +
docker run -it --rm --name debug-bamboo-agent -h bentolor-bamboo-agent \
           --dns 8.8.8.8 \
           -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix \
           -v `pwd`/bamboo/.IntelliJIdea2018.1:/root/.IntelliJIdea2018.1 \
           -v `pwd`/bamboo/.java:/root/.java  bentolor-bamboo-agent  \
           /srv/idea.latest/bin/idea.sh
sudo chown -R $USER:$USER .
git add -u
git commit

FAQ

1. Something went wrong on the agent. How can I diagnoes it

   Simply try to reproduce the steps take by you build plan inside the container, i.e. with:

docker exec -it myagent /bin/bash

I’m having issues to access my local licencse server or other domain names

Probably your container uses a wrong DNS configuration. Create and adjust a /etc/docker/daemon.json to tell you Docker containers about the correct DNS servers:

{
    "dns": ["192.168.144.18", "8.8.8.8"]
}

I can’t push or pull from my local registry myregistry.mycorp.com:5000

You are probably using an insecure registry with no valid SSL certificate. Therefore you must add the following stance into /etc/docker/daemon.json to support so called insecure registries.

{
    "insecure-registries" : ["myregistry.mycorp.com:5000"]
}