Works, so now what?
summary
OpenStack deployment and development at scale actually works: the infra runs 10–20,000 VMs a day across multiple public clouds using only OpenStack APIs, lands about 10,000 changes every 42 days, and records ~1.7 million test jobs in six months with ~18 TB of logs. The team builds a daemonary nodepool that keeps a ready set of single-use nodes, builds and uploads images, and exposes a Python library shade (now backing Ansible modules) that encodes all vendor quirks so multi‑cloud automation is reproducible.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 1:08What is Meta
- 1:50OpenStack
- 2:50Agenda
- 3:36What is OpenStack
- 5:10The Internet
- 6:45Node Pool Activity
- 8:00OpenZac Infosystems
- 8:45Gated Commits
- 10:55Diagrams
- 12:50Garrett
- 13:28Single VM
- 17:15Multicloud
- 18:25Graphs
- 19:25Whats next
- 20:30Get a base image
- 21:35Download prebuilt images
- 22:25Rackspace
- 23:35Image API
- 24:30External Routing
- 25:35Old Bug
- 25:55Config Library
- 26:35Client Config
- 26:50Shade
- 27:10Python API
- 28:10Summary
related talks
57mSnakes on a cloud: the OpenStack project
57mSnakes on a cloud: the OpenStack project
33mOpenStack Compute's automated testing
08mBuilding a Global Object Storage Platform
28mAn introduction to nodepool
34mAutomating infrastructure at SA Home Loans with Python (and friends)
39mArchitecture of a cloud hosting service using python technologies: django, ansible and celery
54mSynnefo: A Complete Cloud Stack Over Google Ganeti