PgQue – Zero-Bloat Postgres Queue

Skill by ara.so — Daily 2026 Skills collection.

PgQue is a pure PL/pgSQL reimplementation of the battle-tested PgQ (Skype/Postgres) architecture. It uses snapshot-based batching and TRUNCATE-based table rotation instead of row-level locking, delivering zero dead-tuple bloat, predictable performance under sustained load, and native fan-out — all from a single SQL file on any Postgres 14+ instance including managed providers (RDS, Aurora, Cloud SQL, Supabase, Neon).

Key Concepts

• Tick: A periodic snapshot that closes a batch of events. Nothing is delivered until a tick fires.
• Batch: A group of events captured in one tick, consumed atomically by a subscriber.
• Subscriber/Consumer: A named cursor on the event log. Multiple consumers get independent copies of every batch (fan-out).
• Zero bloat: Events are stored in rotating tables and cleared via TRUNCATE, never DELETE. No dead tuples.
• Latency trade-off: End-to-end delivery is ~1–2 s (one tick interval + poll). Per-call function latency is microseconds.

Installation

Requirements

• Postgres 14+
• pg_cron (recommended) or an external scheduler calling pgque.ticker() every second