Skip to main content
PostgreSQL · logical replication

Replicate PostgreSQL to your cloud warehouse.

Describe the sync in plain English, approve it once, and rsync.ai streams every change from PostgreSQL into Snowflake, BigQuery, Databricks, or MySQL — read straight from the write-ahead log via logical replication, sub-second and no polling.

Real-time logical replicationYou approve every syncNo per-row pricing
Connect PostgreSQL free →See how a sync runs
Reads the Postgres WAL — no polling You approve before anything moves
Destinations

Send PostgreSQL data where your team already works.

Pick where it lands. rsync.ai maps Postgres types automatically and builds the pipeline — you approve the plan before anything moves.

Cloud warehouse

Land typed, query-ready tables for BI and analytics — incremental after the first snapshot.

BigQueryDatabricksSnowflakeRedshift

Another database

Replicate Postgres into MySQL — or into a fresh Postgres for a migration — with automatic type mapping and real-time CDC.

MySQLPostgreSQL

Storage, lake & sheets

Write Parquet, JSON, or CSV to object storage for your lake, or push query results to a live spreadsheet.

Amazon S3GCSAzure BlobGoogle Sheets
What syncs

Every PostgreSQL object, mapped to typed columns.

Tables
Views
Primary keys
Logical replication · WAL
Snapshot + incremental
JSONB → JSON
UUID → STRING
TIMESTAMPTZ → TIMESTAMP
How it works

Four steps. You are in the loop for each one.

01

Describe it

Name the tables to replicate in plain English — no SQL, no replication-slot setup, no connector config.

02

Review the plan

rsync.ai discovers the schema, maps Postgres types to the destination, and flags any PII before a row moves.

03

Approve

Nothing runs until you say yes. When the schema changes, the pipeline pauses for the same review.

04

Stream

rsync.ai subscribes to the WAL and keeps the destination live within seconds; a failed run resumes from its last committed LSN.

Compare

Why teams move PostgreSQL this way.

What you care aboutpg_dump / cronAWS DMSPer-row ETLrsync.ai
Set up without an engineerNoConsole formsRarelyPlain English
Real-time logical replicationBatch dumpsYesVariesYes
Types mapped automaticallyBy handPartialSometimesAutomatic
You approve before it runsn/aNoNoYes
Cost as volume growsYour timePer hour + I/ORises per rowPer GB, not per row

Straight about status: the PostgreSQL source and its MySQL, BigQuery, Databricks, storage, and Google Sheets destinations are live in production. Snowflake and Redshift are in preview. Self-hosting in your own VPC is coming soon — today rsync.ai runs as a managed cloud.

FAQ

PostgreSQL replication — common questions

How fresh is the data?
You choose the schedule, from a few times a day down to near real time. rsync.ai subscribes to PostgreSQL logical replication (the pgoutput plugin, built into Postgres 10+), so it receives every INSERT, UPDATE, and DELETE for the tables you selected in sub-second latency — no polling, no watermark queries. A failed run resumes from its last committed LSN instead of starting over.
Does CDC work on Amazon RDS, Aurora, Cloud SQL, Supabase, and Neon?
Yes. On RDS and Aurora, set rds.logical_replication=1 in the parameter group (a reboot applies it), then create the replication slot and publication. rsync.ai uses pgoutput, which is native to both. After an Aurora failover the replication slot may need to be re-established; rsync.ai resumes from the last committed LSN and back-fills any gap. Google Cloud SQL, Supabase, and Neon are all supported — you set wal_level=logical (or the provider's equivalent flag) and rsync.ai subscribes. Minimum supported version is PostgreSQL 10.
How much WAL disk will the replication slot use, and what if rsync.ai goes down?
A replication slot retains WAL segments until rsync.ai confirms it has processed them, so no changes are lost during an outage. To protect against disk exhaustion in a long outage, set max_slot_wal_keep_size (Postgres 13+) to a safe limit — typically 5–20 GB depending on write volume. rsync.ai checkpoints its LSN frequently and deduplicates on restart, so in normal operation WAL retention stays minimal and no duplicate rows are written.
Do I need to write SQL or configure a connector?
No. Describe the sync in plain English. rsync.ai discovers the schema, maps Postgres types to the destination — JSONB, UUID, TIMESTAMPTZ, NUMERIC, arrays, and more — flags PII, and shows you the plan. Nothing runs until you approve it, and you can mask or hash sensitive columns (SHA-256) before they ever leave your network.
How are you priced?
On data volume (GB moved), not per row or Monthly Active Rows — so your bill tracks bytes moved, not row counts. Each plan includes a GB allowance, and pipelines pause at the limit rather than silently running up charges; beyond the allowance it's a flat $3/GB.
Who controls what moves?
You do. Every sync waits for your approval, PII is flagged before it lands, and later schema changes pause for review. Self-hosting entirely in your own VPC is coming soon for teams that need data to never leave their network.

Put your PostgreSQL data in the warehouse.

Connect your database free and send your first live sync to BigQuery or Databricks today.

Connect PostgreSQL free →See how it works
Live on rsync.ai Cloud today · self-hosting coming soon