No video of the event yet, sorry!

*The Challenge *

High availability in PostgreSQL environments is often assumed to be solved by managed cloud services. In practice, many default implementations rely on DNS-based failover mechanisms that can take up to five minutes to reroute application traffic. They do not guarantee zero data loss across regions, and they offer limited control over leader election and replication behaviour.

For organisations running mission-critical workloads with strict RPO and RTO requirements, these limitations represent unacceptable risk. When a failover event occurs, every second of downtime and every lost transaction carries real business impact.

The challenge is clear: how do you achieve deterministic failover, synchronous replication and cross-region resilience for PostgreSQL without relying on cloud-managed defaults that were never designed for these guarantees?

*The Approach *

This talk presents a production-grade PostgreSQL high availability architecture designed to solve exactly this problem. The solution is built on three core technologies:

• Patroni for cluster orchestration and automated failover
• etcd for distributed consensus, leader election and split-brain protection
• Barman for cross-region backup management and WAL replication

The architecture follows a share-nothing topology with synchronous replication between primary and standby nodes, deployed across multiple availability zones and geographic regions.

This session also explores the design trade-offs and operational realities encountered when implementing this architecture across geographically distributed Australian infrastructure.

Throughout the session, attendees will walk through the complete design, including:

• Multi-region infrastructure layout with Sydney as the primary site and Melbourne for disaster recovery
• Synchronous replication configuration to eliminate transaction loss between nodes
• How etcd prevents split-brain scenarios where both nodes accept writes simultaneously
• Load balancer integration using Patroni REST API health checks for intelligent read/write routing
• Separation of read and write workloads across primary and standby instances
• Why a standby instance must never be treated as a substitute for a proper backup strategy
• Cross-region latency considerations and their impact on application design patterns
• Security hardening through controlled network access and elimination of direct host exposure

*What You Will Take Away *

Attendees will leave this session with a clear, replicable blueprint for PostgreSQL high availability that goes beyond cloud-managed defaults. Specifically:

• An architecture delivering recovery in seconds during node-level failures
• Zero transaction loss between primary and standby through synchronous replication
• Cross-region disaster recovery capability using Barman and WAL streaming
• Practical guidance on performance trade-offs, security posture and operational resilience

This is not a theoretical exercise. The architecture presented addresses real-world constraints and requirements encountered in enterprise PostgreSQL environments demanding the highest levels of uptime, data integrity and recoverability.

---

Note: This abstract represents the planned scope and structure for a 40-minute talk. The detailed presentation will be developed upon selection.

Date:

Duration:

40 min

Room:

Conference:

PG Down Under 2026

Language:

Track:

Ops and Administration

Difficulty:

Medium