Wikimedia Release Engineering Team/Deployment pipeline/2017-05-09

2017-05-09

Who's here: Giuseppe Alexandros Marko Dan Antoine JR Tyler

Last Time

• Data center switchover work and Marko is a tech manager :)
• discussed: Requirements doc
• Front-end for deployers to pipline
  • could be scap, but probably shouldn't be scap :)
  • Helm
• Build things:
  • Service runner: https://github.com/wikimedia/node-service-builder
  • Blubber: https://github.com/marxarelli/blubber

Topics

• Base images per technology and blubber
  • blubber should be agnostic to base-image
  • "service-type" could specify a base image that provides basic things
  • base image is something like "service-node" module in puppet -- i.e. minimum things that everyone will be using
  • "service-type" is a PaaS-type approach -- must be careful about how granular we go with base images
  • "service-type" vs base-image specification

• Blubber and service runner duplication of effort
  • Different base-images for different node versions in blubber would work
    • Base image cleanup is a mess currently even inside docker
  • Needs pairity with service-runner
    • may not need ability to build deploy repo
    • Deploy repos may go away with docker work

• Prepare and maintain base container images
  • Looks like there is a bootstrap-vz yaml config for stretch

Team Updates

• Releng
  • Lots of POC done, just need to make it sane and live -- Dan & Tyler pairing?
  • Add post-merge to mediawiki/services/mathoid/deploy for next week

• Ops
  • Base image for nodejs

• Services
  • will work on the k8s manifest

As Always

• Release Pipeline Workboard
• Meeting notes