postgres-replica

Logical replication in PostgreSQL is powerful and fiddly: publications, subscriptions, slots, and a REFRESH every time a table changes. I wanted a container that reads one YAML file and keeps a source in sync with many targets, so I could stop hand-typing the same SQL on every project.

For whom: small teams running Postgres in containers. Constraints: dynamic table add and remove without downtime, cross-schema support, an initial-sync toggle, and a health endpoint an orchestrator can poll.

• YAML in, replication out. The file is the desired state; the service reconciles publications and subscriptions towards it.

• Dynamic table changes without downtime. Adding or removing a table alters the publication and refreshes the subscription instead of rebuilding it.

• A health endpoint. Swarm and Kubernetes can tell the difference between "running" and "replicating".

• Docker Hub as the distribution channel. No install, one image, environment-driven.

1,441 image pulls on Docker Hub, 9 commits, and a Medium write-up. It went on to power the Postgres replication tool inside KwikNkap.

Nothing dramatic in production yet, which is what a nine-commit tool should say. The first thing I would fix: a new table added to the YAML is published immediately but only initial-syncs if the toggle is on, and the health endpoint does not distinguish "subscribed, lagging" from "subscribed, healthy". Lag reporting is the next commit.