PostgreSQL Conference Germany 2026

Upgrading PostgreSQL across hundreds of production databases without downtime sounds impossible,especially when logical replication slots, Debezium CDC pipelines, and outbox event streams are in play. At Fresha, we faced exactly that: mission-critical workloads still on PostgreSQL 12, and no safe way to reach 17 without freezing the business.

This talk walks through how we designed and automated a blue-green upgrade framework using logical decoding, controlled WAL overlap, and connector orchestration. We’ll dive into how Debezium connectors, replication origins, and PgBouncer pools were coordinated to guarantee continuity for both CDC and outbox topics with rollback and dry-run modes built in.

Attendees will learn practical techniques for:

• Orchestrating zero-downtime Postgres major upgrades on RDS or self-managed clusters

• Managing replication slots and Debezium connectors safely across clusters

• Handling sequence alignment, WAL overlap, and connector state transitions

• Designing reversible, testable database cutovers

This is a practical session from real production experience: no magic tools, just PostgreSQL internals, Debezium knowledge and careful planning, and a few well-placed bash scripts.