High‑Availability Cluster Solutions for PostgreSQL

High availability (HA) refers to the amount of time a service remains operational, typically defined by a business, and relies on redundancy to continue functioning after failures.

Continuous recovery requires restoring backups and applying WAL logs; if logs are archived on a standby server and applied promptly (e.g., every minute), recovery time can be reduced to near‑real‑time.

A standby database is a replica that maintains the same data as the primary and can be classified by replication nature (physical vs. logical), transaction synchrony (asynchronous vs. synchronous), and usage mode (hot standby with or without read‑only connections).

Clustering groups multiple hosts to act as a single logical server, providing horizontal scalability, fault tolerance, and transparent operation. Two clustering models exist: shared‑storage and non‑shared storage.

Several PostgreSQL HA clustering options are discussed:

DRBD : a Linux kernel module for synchronous block replication; requires additional software (e.g., Corosync + Pacemaker) for failover and monitoring.

ClusterControl : agent‑less management and automation platform that deploys, monitors, and scales database clusters, offering features such as automated failover, backup scheduling, and real‑time monitoring.

Rubyrep : an asynchronous multi‑master, multi‑platform replication tool implemented in Ruby/JRuby, focusing on simplicity but lacking DDL and permission support.

Pgpool II : middleware providing connection pooling, replication, load balancing, automatic failover, and parallel query execution.

Bucardo : trigger‑based asynchronous multi‑master/replica system written in Perl, supporting various source and target databases but unable to handle DDL or large objects.

Postgres‑XC : an open‑source, write‑scalable, synchronous, symmetric cluster solution with a global transaction manager, offering transparent data distribution and high availability.

Citus : adds built‑in HA with automatic sharding and replication; redirects writes/queries to healthy nodes when failures occur.

PostgresXL : a shared‑nothing multi‑master cluster with a Global Transaction Manager (GTM), supporting diverse workloads such as OLTP, BI, GIS, and multi‑tenant environments.

The article concludes that many other products exist, but users must consider factors like product maturity, licensing costs, implementation complexity, and infrastructure single points of failure before selecting a solution.