Fun with Postgres High Availability Poker

on YouTube ↗33m

" summary

Live demos use playing cards to model WAL, Heap and checkpoints and show that a standalone yields RPO zero but long RTO when recovering from the last checkpoint. Asynchronous replicas confirm commits immediately and risk losing recent transactions; the demo shows missing diamonds when a primary fails. Synchronous commit prevents data loss but doubles latency and requires at least three copies to avoid blocking. The hands‑on build exposes WAL transport, synchronous commit, replica apply lag and network as real failure modes and it works.

AI-written summary · about

" chapters (6)

1. 0:00Introduction
2. 2:21Core database concepts
3. 9:29Standalone database model
4. 9:46Asynchronous replication
5. 17:30Synchronous replication
6. 24:35Conclusions and Q&A

" related talks

• 22mSecuring Postgres with streaming replicationJan Karremans · POSETTE: An Event for Postgres 2025
• 55mHigh Availability with PostgreSQL and PacemakerShaun M. Thomas · Postgres Open 2012
• 02h 26mTutorialDenish Patel · PGConf Silicon Valley 2015
• 26mMyths and Truths about Synchronous Replication in PostgreSQLAlexander Kukushkin · POSETTE: An Event for Postgres 2025
• 48mReplication RefreshSimon Riggs · PostgresOpen 2013
• 42mPostgreSQL Replication Conflicts: Avoiding PitfallsHamid Quddus Akhtar · PGConf NYC 2022
• 36mDesigning Layered High‑Availability Architectures for PostgreSQL on a BudgetUmair Shahid · SREcon26 Americas
• 47mPostgreSQL High Availability: The Considerations and the CandidatesHamid Quddus Akhtar · Percona Live 2022