// the find
sorintlab/stolon
PostgreSQL cloud native High Availability and more.
Stolon is a PostgreSQL HA manager that uses etcd, Consul, or the Kubernetes API as a distributed store for leader election and cluster state. It wraps native streaming replication with a sentinel/keeper/proxy architecture to handle failover automatically. The target audience is teams running PostgreSQL on Kubernetes or bare VMs who want something lighter than Patroni but more opinionated than DIY replication.
The three-component split (sentinel computes the view, keeper enforces it, proxy cuts off stale connections) is a clean design that avoids split-brain without requiring fencing agents. Support for pg_rewind means failed primaries can rejoin as standbys quickly instead of needing a full base backup. Standby cluster support lets you run a warm replica cluster at a second site, which covers both DR and near-zero-downtime migrations. The integration test suite actually spins up real PostgreSQL instances and covers partition scenarios — that's rare and worth crediting.
Last meaningful activity is mid-2024 and the pace has been declining for years; the PostgreSQL version support list stops at 15, missing 16 and 17 entirely, which is a concrete gap for anyone on a current version. The proxy is a mandatory single point of entry — you must route all client connections through it, which adds a network hop and a new failure surface that many users don't realize until they're already committed. There is no built-in support for connection pooling (PgBouncer or equivalent) and the docs don't guide you through integrating one, so production deployments usually bolt it on awkwardly. The Kubernetes example YAMLs are StatefulSet-era vintage and don't reflect how most people deploy PostgreSQL on Kubernetes today (Helm charts, operators).