Keynote: The Terrifying Migration of 420 Services to Kubernetes
summary
420 services move to Kubernetes using an existing Python DSL that generates CloudFormation and Kubernetes manifests, and the migration reveals concrete failures: developers choke on YAML vs DSL, missing IRSA/firewall rules cause intermittent 500s to the database, and upgrades fail without pod disruption budgets or ordered termination. Recovery requires rolling back and hybrid mode; building an infrastructure testing environment and enforcing migration ready checklists (docs, discovery sessions, rollback plan, communication channels) proves effective at preventing repeat outages.
AI-written summary · about
chapters (8)
- 0:00Start
- 0:51Developers running around aimlessly
- 5:00Knowledge gap
- 9:51Mindset Migration
- 10:24Create communication channels with the developers
- 11:11Not just production ready, be migrations ready
- 11:54Plan what happens when we get to shore
- 12:48Infrastructure testing env - Upgrades & breaking changes
related talks
20mKeynote: The Challenges of Migrating 150+ Microservices to Kubernetes
20mKeynote: Kubernetes - Don't Stop Believin'
17mKeynote: Kubernetes: Finally...A True Cloud Platform
17mKeynote: Nine lessons from Kubernetes in Enterprise
14mKeynote: Kubernetes as a Foundational Layer of Infrastructure
35mPerforming Infrastructure Migrations at Scale
08mSponsored Keynote: Free Range Kubernetes: Feeding, Watering & Herding - Brian Goodman-Jones
27mA Safari of Kubernetes and its Natural Habitat