The Immortal Postgres Web
summary
Shows how to build a massively resilient Postgres web stack using HAProxy, Patroni, and ZooKeeper wired together: Patroni exposes a REST health port, ZooKeeper provides quorum for leader election, HAProxy routes traffic to the active primary and round‑robins reads to replicas. Demonstrates dynamic templating of HAProxy from the KV store so clusters auto‑register, uses PG‑rewind for rejoin, and yields automatic failover in seconds — with the caveat that high‑frequency, microsecond‑safe block‑level replication (e.g., DRBD) may be required for some financial workloads.
AI-written summary · about
chapters (20)
- 0:00Intro
- 2:11Who is This Guy?!
- 3:43What isn't High Availability?
- 6:28Bits and Pieces
- 10:08Ridiculous HA
- 12:32Enter Sandman Governor
- 13:57Ch-Ch-Changes
- 18:08Out with the Governor!
- 18:51Key Differences and Enhancements
- 20:28It's Complicated
- 22:29Zookeeper Differences
- 23:34Component Summary
- 23:51Layering Up
- 25:16Initial Proxy Pool: HAProxy
- 28:54Getting Fancy
- 30:40Handy Horizontal Horizons
- 32:46Ceaseless Entertainment
- 35:51Sharing the Load
- 37:38Hiding the Seams
- 41:04Questions?
related talks
13mBuilding reliable infrastructure with Postgres
57mReliability features in Postgres Pro
46mDeploying Maximum HA Architecture With Postgres
50mPatroni Deployment Patterns
47mPostgreSQL High Availability: The Considerations and the Candidates
41mEnsuring High Availability in Self-Managed PostgreSQL with Patroni
36mDesigning Layered High‑Availability Architectures for PostgreSQL on a Budget
25mFollow my Leader connecting Postgres client applications after server failover