Deep Dive: etcd
summary
etcd provides a Raft implementation that exposes a step function and returns a Ready structure of state, entries and messages while deliberately implementing no network transport and no storage layer so users must persist state and deliver messages. leader election, log replication and quorum math (e.g. five nodes require three) are shown end-to-end, membership supports learners and joint consensus work is underway, and the Raft package is reused by projects like CockroachDB and TiKV.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:03Agenda
- 1:41Raft Recap
- 3:21The Problem
- 6:37Leader Election
- 7:51Leader Log Management
- 13:05Design Philosophy
- 14:04State Machine
- 15:05Code
- 17:11Output
- 19:27Proposed
- 27:40More details
related talks
40mDeep Dive: etcd
21mUsing Raft in Rust
42metcd: distributed locking and service discovery
41mConsistent, Distributed Elixir
42mPractical Raft Consensus
59mSolving Raft's practical problems in Tarantool. What, how and why
54mGovernor: Postgres HA Orchestator
44mHow Raft consensus algorithm will make replication even better in MongoDB 3 2