2017-03-14 edit

Who's here: Marko, Tyler, Dan, Antoine, JR

Last Time edit

Discussed when to build containers — as frequently as possible
Secrets — future discussion
e2e tests before things merge and are deployed to staging
Evaluate deployment-pipeline-jenkins plugin
- Build testing k8s setup in ci-staging
- Install Kubernetes Continuous Deployment Pipeline plugin on ci-staging-jenkins

Spin up new Jenkins on k8s per project
- gets environment from k8s create yaml file JENKINS_ADMIN_USER, JENKINS_ADMIN_PASSWORD, POD_IP, POD_NAMESPACE, POD_NAME, git repo, git private key
- k8s yaml for jenkins service
This deployment uses a base container
- There are 64 layers according to docker history [image]
Spawn a jenkins master for each project
Installs and pins Jenkins plugins via a shell script + curl
- jobdsl is the big one
init.groovy.d
- Contains scripts that create seed-job if it doesn't exist: http://ci-staging-jenkins.wmflabs.org/ci/job/seed-job/
seed-job runs a bunch of groovy scripts which themselves contain a bunch of HEREDOC'd groovy scripts
- https://github.com/thcipriani/deployment-pipeline-jenkins-plugin-setup/tree/master/jobdsl
seed-job makes more jobs: http://ci-staging-jenkins.wmflabs.org/ci/
- Pulls down repo defined in k8s
  - build.bitesize, environments.bitesize, containers.bitesize
- builds jobs based on those files + inception groovy scripts

Lots of thought went into this, it's well automated
There is a lot going on in their Docker base container (exhumed from base container)
We have a lot of repos (1080 configured in CI as of March 14th). A Jenkins for every repo is a lot of Jenkinses
"enterprise class"
Some good ideas here

TODO releng team meeting bring up specifications/requirements for deployment pipeline for next quarter