PostgreSQL High Availability: The Considerations and the Candidates

on YouTube ↗47m

" summary

Defines HA by concrete RPO/RTO targets, compares actual tools and tradeoffs, and shows what each needs to run. Notes Patroni + etcd leader lock as the popular but config-heavy template; says repmgr requires rsync and passwordless SSH and embeds in Postgres; flags PG Bouncer is single-threaded unless kernel SO_REUSEPORT and describes pg_stat_monitor as a released Percona observability extension. Gives enough to pick a candidate by operational constraints and failure-recovery behavior.

AI-written summary · about

" chapters (18)

1. 0:00Intro
2. 0:27About Myself More than two decades of professional software development • I'm part of Percona's PostgreSQL Engineering team
3. 2:40Initial Thoughts . There are no good or bad HA solutions. There are suitable, unsuitable ones.
4. 3:18Presentation Outline • Architectural Considerations
5. 4:24Why High Availability is Needed? "High Availability" system ensures service continuity
6. 7:11RPO: Recovery Point Objective Recovery point objective is the acceptable data loss. • What is the acceptable data loss in case of failure?
7. 8:36RTO: Recovery Time Objective • Acceptable service down time after the incident has occurred. Service downtime during the recovery phase. • Considerations
8. 11:12How Many PostgreSQL Instances? • Quorum size?
9. 12:46The Dreaded Split Brain and Cluster Degradation • Cluster with a single primary can end up with multiple ones.
10. 14:31Multi-Primary Cluster Considerations Is logical replication an option? • Proprietary solutions available
11. 16:17Configuration Hardware configuration
12. 17:30Fixing Cache Poisoning Should you immediately restart after a PostgreSQL crash? This is a good practice, however, it becomes a consideration with an HA system where you want to prevent unnecessary auto-scaling
13. 18:55Fixing Rogue Connections • Long running queries can become a pain
14. 23:03Performing Upgrades • Minor upgrades
15. 24:42Virtual IP Address and Domain Name Reassignment • ip command Use it to release and assign the IP address to a new system. • arp command Let the network know about the reassignment • DNS changes
16. 25:48HAProxy - Load Balancer • Simple to install
17. 27:49PgBouncer - Connection Pooler • Configuring Pgbouncer
18. 39:57RepMgr - Workings Setup

" related talks

• 36mDesigning Layered High‑Availability Architectures for PostgreSQL on a BudgetUmair Shahid · SREcon26 Americas
• 46mDeploying Maximum HA Architecture With PostgresDenish Patel · Postgres Open 2012
• 57mReliability features in Postgres ProLeonid Albrekht · PGConf India 2025
• 50mEvolution Of PostgreSQL Replication and High Availability (upto ~ 99.999%)Stacy Scoggins · PGConf NYC 2021
• 55mHigh Availability with PostgreSQL and PacemakerShaun M. Thomas · Postgres Open 2012
• 52mOverview of Solutions for PostgreSQL High AvailabilityAleksey Lesovskiy · Backend Conf 2019 (РИТ++)
• 50mMoving from MSSQL to pgSQL – High Availability and Disaster RecoveryRick Lowe · Postgres Conference Orlando 2025
• 13mBuilding reliable infrastructure with PostgresAndrey Fleyta · PGConf India 2025