Understanding transaction visibility in PostgreSQL clusters with read replicas
a year ago
- #PostgreSQL
- #Database
- #AWS
- Jepsen report highlights transaction visibility behavior in Amazon RDS for PostgreSQL Multi-AZ clusters.
- Issue involves Long Fork anomaly, affecting visibility order between primary and replicas.
- No data loss or corruption; issue known to PostgreSQL community since 2013.
- Example scenario: Alice and Bob see different rankings due to visibility order differences.
- AWS is working with PostgreSQL community to propose a fix using Commit Sequence Numbers (CSN).
- Long Fork anomaly impacts distributed systems, query routing, and data synchronization.
- AWS commits to PostgreSQL's success, with a dedicated PostgreSQL Contributors Team.
- Recommendations: review application assumptions, use explicit synchronization, contact AWS Support.
- Further reading available on PostgreSQL community work and Aurora solutions.