Building a Continuous Integration Pipeline with Docker Alex Zbarcea - Apache Committer (alexz@apache.org, in/azbarcea, @azbarcea)

ABOUT THIS PRESENTATION Continuous Delivery

Business is a System inputs outputs feedback

Business is a System

Jenkins Artifacts Jenkinsfile + artfifacts artifacts

remote API dckr01 remote API dckr02 tutum jenkins archiva registry nexus Jenkins Builds

Automated Service Operations App App Vault Nexus Internal on-premise services Customer facing services github Continuous Integration Pipeline

The Machine to build the Machines

Computer architecture

No content

No content

rhel tomcat java app Another Distro One package manager - automated patches Reuse Isolated app. One hierarchy security Conf. Mngt.

Artifacts .jar .war .npm .pip .apk .deb .rpm .docker .-box.tar.gz .git - Chef - Ansible - .docker compositions - Pods .wiki .swagger.json apps/libs OS node config. docs .docker .docker compositions .rancher compositions only markdown (text) artifacts are allowed

Demo

Challenges

Pipeline Challanges - everybody sees and understands the pipeline - optimize for the global scope - inventory of INs and OUTs - artifact warehouse management - containerize everything

- Not all image can be build in docker-in-docker (dind) - Caching artifacts (e.g.: /.m2/repository folder) - Use data containers Docker Challanges

Operations Challenges - FAIL is good. Trust the upgrade button - complexity - generates: lack of trust - difficult to investigate - .git/social becomes a core system - emergent properties - focus is on scaling - deployments as a non-event

Thank you MCJUG apifocal