PostgreSQL integration

PostgreSQL CDC pipelines,
described in plain English.

Replicate PostgreSQL to any database, warehouse, or file store using real-time CDC logical replication. Sub-second latency, no polling, no data engineers required.

TL;DR

rsync.ai subscribes to PostgreSQL logical replication (pgoutput / wal2json) and streams every INSERT, UPDATE, and DELETE to your destination in sub-second latency. Works with self-hosted Postgres, Amazon RDS, Aurora, Google Cloud SQL, Supabase, and Neon. Destinations include MySQL, AWS S3, and Google Sheets. No polling — changes land as they happen.

Real-time CDC via logical replication — not polling
Works with RDS, Aurora, Cloud SQL, Supabase, Neon, self-hosted
Plain-English pipeline setup — no YAML, no DAGs
Self-hosted under Elastic License 2.0 — data never leaves your network

Pick your destination

Three production-ready destinations for your PostgreSQL CDC pipeline.

PostgreSQL → MySQL

Stream PostgreSQL changes to MySQL in real time via CDC logical replication. Automatic type mapping: UUID→CHAR(36), JSONB→JSON, TIMESTAMPTZ→DATETIME.

PostgreSQL → AWS S3

Export Postgres tables to S3 as Parquet or JSON — data lake ingestion, compliance archiving, cost-effective long-term storage. CDC streaming or scheduled snapshots.

PostgreSQL → Google Sheets

Push Postgres query results into Google Sheets on a schedule. For business users who live in spreadsheets — no SQL required on their end.

What rsync.ai captures from PostgreSQL

Logical replication gives rsync.ai a live event stream — schema-filtered, column-masked, resumable.

Logical replication (wal2json / pgoutput)Replication slots & publicationsWorks with RDS, Aurora, Cloud SQL, Supabase, NeonSelf-hosted PostgreSQL 10+Schema-level filteringSub-second change capturePII column maskingResume after downtime+ DDL propagation (ALTER TABLE) on roadmap

rsync.ai vs. Fivetran, Airbyte, AWS DMS for PostgreSQL CDC

An honest comparison for real-time PostgreSQL replication pipelines.

Feature

rsync.aiyou

Fivetran

Airbyte

AWS DMS

Real-time CDC via logical replication

Plain-English pipeline setup

Self-hosted (data stays in your network)

Source-available connector code (auditable)

PII column masking built-in

No per-row / per-MAR pricing

Resume after downtime (no data loss)

PostgreSQL CDC — frequently asked

What is PostgreSQL logical replication and why does rsync.ai use it?

Logical replication reads the PostgreSQL write-ahead log (WAL) at the logical level, meaning rsync.ai receives a stream of INSERT/UPDATE/DELETE events for exactly the tables you selected — in sub-second latency, with no polling. Unlike physical replication, it works across major versions and doesn't require the destination to be another Postgres instance. rsync.ai uses the pgoutput plugin (built into Postgres 10+) or wal2json where available. You set `wal_level=logical`, create a replication slot and publication, and rsync.ai subscribes.

Does CDC logical replication work with Amazon RDS or Aurora PostgreSQL?

Yes. On RDS and Aurora PostgreSQL you set the `rds.logical_replication` parameter to `1` in your parameter group (requires a reboot), then create the replication slot and publication as normal. rsync.ai uses the pgoutput plugin which is native to both. Aurora Multi-AZ is also supported — the replication slot is on the writer instance and survives failover. Minimum supported version is PostgreSQL 10 on RDS; Aurora PostgreSQL 10.7+.

How much WAL disk space will the replication slot consume?

A replication slot retains WAL segments until the consumer (rsync.ai) confirms it has processed them. If rsync.ai is down for an extended period, WAL accumulates. To protect against disk exhaustion, set `max_slot_wal_keep_size` (PostgreSQL 13+) to a safe limit — typically 5–20 GB depending on your write volume. rsync.ai checkpoints its LSN position frequently and resumes cleanly after a restart, so in normal operation WAL retention is minimal.

What happens if rsync.ai goes down — will I lose changes?

No. The replication slot on your PostgreSQL server holds all unflushed WAL segments until rsync.ai reconnects and confirms it has processed them. When rsync.ai restarts, it resumes from the last committed LSN. No changes are lost, and no duplicate rows are written — each change event carries a unique LSN that rsync.ai deduplicates on restart.

Can rsync.ai mask PII columns before they reach the destination?

Yes. In the pipeline configuration you mark any column as `masked` — rsync.ai replaces the value with a SHA-256 hash (or NULL, your choice) before writing to the destination. Masking happens in the rsync.ai pipeline process, so the raw value never appears in the destination. Common use cases: email, phone, SSN, card number columns in orders or users tables.

Is rsync.ai self-hosted for PostgreSQL CDC?

Yes. The full rsync.ai stack runs on your own infrastructure via `docker compose up`. Your PostgreSQL replication credentials never leave your network. The source database, rsync.ai process, and destination all live inside your VPC or on-premise environment. License is Elastic License 2.0 — free to self-host, cannot be resold as a managed service.

Ship your PostgreSQL pipeline today.

Pick a destination — MySQL, S3, or Google Sheets — and have real-time replication running in minutes.

PostgreSQL → MySQL PostgreSQL → AWS S3 PostgreSQL → Google Sheets