Changelog Master Feed The Docker Swarm story (Ship It! #47)
Apr 8, 2022
Andrea Luzzardi, the former lead at Docker Swarm, delves into the fascinating evolution of Docker, sharing insights from its infancy at dotCloud to the inception of Swarm. He discusses the technical challenges of multi-host networking and the surprising scale testing results from running thousands of nodes. Luzzardi reflects on the differences between Swarm and Kubernetes, emphasizing managed services for most users. He also shares memorable moments from the Swarm team, highlighting the collaborative spirit that fueled its development.
AI Snips
Chapters
Transcript
Episode notes
Docker Grew From dotCloud's Internal Platform
- Andrea describes Docker's origin as an internal platform at dotCloud that evolved from an internal container runtime and tools.
- That internal stack later became open-sourced as Docker, including the registry, build, and Swarm.
From Stateless Multiplexer To Stateful Orchestrator
- Swarm began as a stateless API multiplexer to make the single-node Docker API work across machines.
- Swarm mode later added state (Raft) and a new services API to support multi-host orchestration.
SwarmKit Brought Consensus Inside Docker
- SwarmKit bundled a Raft implementation and a services API to avoid requiring external etcd clusters.
- The goal was one-command clustering with docker swarm init and join for simplicity and good DX.
