xlyk/docker-jnlp-slave-docker

Docker JNLP slave with docker installed

Jenkins JNLP Agent Docker image with docker inside

This is an image for Jenkins agent (FKA "slave"/"worker") using JNLP to establish connection. This agent is powered by the Jenkins Remoting library, which version is being taken from the base Docker Agent image.

It includes addionally to original image tools useful on CI/CD workers:

• docker installed locally for docker-inside-docker or docker-outside-docker scenario
• awscli
• j2cli
• kubectl
• helm

See Jenkins Distributed builds for more info.

Running

To run a Docker container

docker run korekontrol/jnlp-slave-docker -url http://jenkins-server:port <secret> <agent name>

Optional environment variables:

• JENKINS_URL: url for the Jenkins server, can be used as a replacement to -url option, or to set alternate jenkins URL
• JENKINS_TUNNEL: (HOST:PORT) connect to this agent host and port instead of Jenkins server, assuming this one do route TCP traffic to Jenkins master. Useful when when Jenkins runs behind a load balancer, reverse proxy, etc.
• JENKINS_SECRET: agent secret, if not set as an argument
• JENKINS_AGENT_NAME: agent name, if not set as an argument

Running docker inside docker

If you want to use Jenkins in a container for building and running docker images, you must run docker inside/outside docker. Two scenarios are possible - see: https://applatix.com/case-docker-docker-kubernetes-part/ Docker outside docker seems very slow...

Configuration specifics

Enabled JNLP protocols

By default, the JNLP3-connect is disabled due to the known stability and scalability issues. You can enable this protocol on your own risk using the JNLP_PROTOCOL_OPTS=-Dorg.jenkinsci.remoting.engine.JnlpProtocol3.disabled=false property (the protocol should be enabled on the master side as well).

In Jenkins versions starting from 2.27 there is a JNLP4-connect protocol. If you use Jenkins 2.32.x LTS, it is recommended to enable the protocol on your instance.

Amazon ECS

Make sure your ECS container agent is updated before running. Older versions do not properly handle the entryPoint parameter. See the entryPoint definition for more information.

Credits

Created by Marek Obuchowicz from KoreKontrol - kubernetes consulting

License

MIT