Continuous integration with Docker and Ansible

Continuous integration with Docker and Ansible

Dmitriy Slupytskyi

About project

• 70+ repositories

• 10+ microservices

• 4+ technologies (golang, php, ruby, nodejs)

• 65 build and deploy scripts

Docker

Virtual Machines Containers

Why Docker?• Easily to manager containers

• Build specify environment for project

• Isolated workspace

• Package it and be portable

• Replicable - lower barrier to entry for new environments, teams

• No “dependence hell”

Deployment process

Start Clone Build Push DeployTest

Jenkins

CloudBees Docker Build and Publish plugin

Build• Use Dockerfile

• Keep layers to minimum

• Use tags for version control

• Use .dockerignore(the same syntax as .gitignore)

• One container per service

• Manage output

• GNU Make is optional

Just for presentation, make better onefor your services

😉

Build

Test

Private Docker registry

• Version 0.9.1 (version 2 required docker version 1.6.0 and higher)

• Nginx + LDAP authentication (https://hub.docker.com/r/h3nrik/nginx-ldap/)

• Dockerized

Ansible

• Inventory

• Playbooks

• Tasks

• Modules

• docker-py (<=1.1.0)

• docker module(http://docs.ansible.com/ansible/docker_module.html)

• docker-compose

docker-composehttps://docs.docker.com/compose/

Deploy

Jenkins + Ansible plugin

Logspout(https://hub.docker.com/r/gliderlabs/logspout/)

Papertrail

Profit• easily to update and maintenance images

• Dockerfile

• portability

• isolated

• logspout

• docker registry timeout

• ansible freezes

• time to build

Developers

Pros

• Logs

• Build specify environment for target service

• Service management (start, stop, restart)

• Manage environment variables (one container for few environments)

Cons• SSH

• Time to build (can be depends on various factor, for example docker cache)

• Jenkins plugin exit with 0, even if fails, that bring up misunderstanding

• Unmounted volumes