Data Tool Guides

Databases

8 Best AlloyDB ETL Tools for PostgreSQL CDC in 2026

Compare 8 AlloyDB ETL tools, from Google Datastream to Airbyte, Fivetran and Debezium, with notes on logical decoding, replication slots and BigQuery.

Nov 5, 2026

Summarize with AI:

Download this ebook

How regulated industries build AI-ready data infrastructure without sending records, credentials or keys outside their boundary.

TL;DR

Short answer: AlloyDB for PostgreSQL is Google Cloud's PostgreSQL-compatible database, so AlloyDB ETL tools are mostly PostgreSQL tools pointed at AlloyDB. Google's Datastream streams AlloyDB changes into BigQuery with a free monthly allowance, Airbyte connects through its PostgreSQL source to reach any warehouse, and Debezium suits real-time event streams. Every CDC option needs logical decoding enabled on the primary instance. Eight AlloyDB ETL tools compared:

Airbyte: connects to AlloyDB through its PostgreSQL source, with logical-replication CDC into any of its destinations.
Google Datastream: Google's serverless CDC service, with an automated AlloyDB to BigQuery flow.
Fivetran: a managed PostgreSQL connector that works with AlloyDB, billed on monthly active rows.
Debezium: open-source CDC that streams AlloyDB changes as events through Kafka or Debezium Server.
Estuary: a managed real-time CDC platform whose PostgreSQL capture works with AlloyDB.
Google Database Migration Service: Google's managed service for migrating PostgreSQL databases into AlloyDB.
dlt: an open-source Python library whose SQL database source reads AlloyDB tables into a warehouse.
pg_dump and COPY: PostgreSQL's built-in tools for one-off exports and bulk loads.
The AlloyDB constraint: change data capture reads the write-ahead log from the primary instance through logical decoding, which you enable with the alloydb.logical_decoding flag, and turning it on restarts the instance, so schedule it.

AlloyDB ETL tools move data out of AlloyDB for PostgreSQL, Google Cloud's managed, PostgreSQL-compatible database built for demanding transactional and analytical workloads, into warehouses such as BigQuery, Snowflake or Databricks. Because AlloyDB speaks the PostgreSQL protocol and supports standard logical replication, almost any tool with a PostgreSQL connector can read it. For PostgreSQL-specific tools, see our guide to PostgreSQL ETL tools.

Do you need to move data out of AlloyDB at all?

Sometimes not. AlloyDB's columnar engine can accelerate analytical queries on the database itself, and AlloyDB Omni runs the same engine outside Google Cloud. Moving data out still makes sense when you need to:

Join with other sources: combine application data with SaaS, event and finance data in one warehouse.
Protect the primary: keep heavy reporting off the transactional database.
Use warehouse tools: run BI, dbt models and machine learning where the rest of your analytics lives.

How do the AlloyDB ETL tools compare?

Tool	Approach	Destinations	Best for	Cost
Airbyte	PostgreSQL source with CDC	Any warehouse, lake or database	AlloyDB next to SaaS and other databases	Free self-hosted Core; paid Cloud plans
Datastream	Serverless CDC	BigQuery, Cloud Storage	Google-native replication into BigQuery	Free tier, then per GiB
Fivetran	Managed PostgreSQL connector	Major warehouses	Managed pipelines with no upkeep	Monthly active rows
Debezium	Log-based CDC events	Kafka and Debezium Server sinks	Real-time event streams	Free, open source
Estuary	Managed real-time CDC	Warehouses, lakes, streams	Low-latency managed CDC	Usage-based
Database Migration Service	Managed migration	AlloyDB	Moving PostgreSQL into AlloyDB	No charge for homogeneous migrations
dlt	Python SQL source	DuckDB, BigQuery, Snowflake and more	Code-first batch loads	Free, open source
pg_dump and COPY	Native tools	Files, other PostgreSQL databases	One-off exports	Free

Which AlloyDB ETL tools should you consider?

Eight AlloyDB ETL tools, from managed CDC to native PostgreSQL tools:

1. Airbyte

Airbyte is the open-source data connectivity platform for AI and analytics, and among AlloyDB ETL tools it is the one that reaches any destination. Airbyte no longer lists a separate AlloyDB connector; instead, AlloyDB connects through Airbyte's PostgreSQL source, which supports logical-replication CDC with the pgoutput plugin, so inserts, updates and deletes all reach the destination. For other CDC options, see our guide to CDC tools.

That route lands AlloyDB data in Snowflake, Databricks, Redshift or BigQuery next to data from 700+ other connectors. Airbyte runs in Airbyte Cloud, in your own environment with Airbyte Flex, or self-hosted with Airbyte Core, which can run inside your Google Cloud VPC.

Key features

Logical replication CDC: reads the write-ahead log through a publication and replication slot, capturing deletes.
Standard mode: tables without CDC can sync incrementally on a cursor column.
Private connectivity: SSH tunnelling and self-hosting reach AlloyDB on private IPs.
Many destinations: Snowflake, BigQuery, Databricks, Redshift, S3 and more.

Best for: teams loading AlloyDB into a warehouse other than BigQuery, or alongside many other sources.

Pricing: self-hosted Airbyte Core is free; Airbyte Cloud is billed on volume, and Pro and Airbyte Flex on capacity.

Pros	Cons
Any destination, not only BigQuery	No separate AlloyDB-branded connector
Log-based CDC including deletes	Needs a dedicated replication slot
Cloud, Flex or self-hosted Core	Schema changes on CDC tables may need a resync

2. Google Datastream

Datastream is Google's serverless CDC service, and among AlloyDB ETL tools it is the most native. An automated flow, started from the AlloyDB cluster page, configures logical decoding, the publication, the replication slot and private networking, then streams changes from the primary instance into BigQuery or Cloud Storage. For more Google Cloud options, see our guide to GCP ETL tools.

Key features

Automated flow: one click creates connection profiles, private connectivity and the stream.
Backfill and CDC: existing rows are backfilled, then changes stream continuously.
BigQuery integration: changes are applied to BigQuery tables in near real time.
Free tier: up to 100 GiB of AlloyDB to BigQuery CDC data a month is free.

Best for: teams whose analytics live in BigQuery and who want Google-managed replication.

Pricing: a free tier of 100 GiB of CDC data a month for AlloyDB to BigQuery, then billed per GiB processed.

Pros	Cons
Automated setup from the AlloyDB console	Destinations limited to BigQuery and Cloud Storage
Free tier for AlloyDB to BigQuery	Enabling logical decoding restarts the instance
Serverless, no infrastructure	Streams only from the primary instance

3. Fivetran

Among managed AlloyDB ETL tools, Fivetran connects through its PostgreSQL connector, using logical replication to capture changes, with no infrastructure to run and automatic schema handling. Fivetran completed its merger with dbt Labs in June 2026.

Key features

Logical replication: captures inserts, updates and deletes from the write-ahead log.
Schema evolution: new columns flow downstream automatically.
Hybrid deployment: processing can run inside your network for regulated data.
Re-syncs: historical re-syncs can be triggered on demand.

Best for: teams already standardised on Fivetran who want AlloyDB in the same managed pipeline.

Pricing: billed on monthly active rows, with a free tier for low volumes.

Pros	Cons
Fully managed	Billed on monthly active rows
Automatic schema handling	Busy tables drive cost
Many warehouse destinations	Less control over sync behaviour

4. Debezium

Debezium is the open-source, event-driven choice among AlloyDB ETL tools: its PostgreSQL connector works with AlloyDB through standard logical decoding. It takes an initial snapshot, then emits a change event for every insert, update and delete, with before and after values, to Kafka or, through Debezium Server, to targets such as Pub/Sub.

Key features

pgoutput support: uses PostgreSQL's built-in logical decoding plugin.
Snapshots: an initial snapshot precedes streaming.
Debezium Server: sends events to Pub/Sub, Kinesis and other sinks without Kafka.
Event detail: each event carries the log position and transaction metadata.

Best for: engineering teams building event-driven systems on AlloyDB changes.

Pricing: free and open source under the Apache 2.0 licence.

Pros	Cons
Real-time change events	You run Kafka or Debezium Server
Free and open source	Warehouse loading is a separate step
Rich before and after values	Operational effort for slots and offsets

5. Estuary

Estuary is one of the lower-latency AlloyDB ETL tools, a managed real-time data platform whose PostgreSQL capture works with AlloyDB through logical replication. It backfills existing data, streams changes with low latency, and can deliver the same data to warehouses, lakes and streaming systems at once.

Key features

Logical replication capture: reads AlloyDB changes continuously.
Multiple materializations: deliver one stream to several destinations.
Transformations: derivations reshape data in flight.
Managed service: no Kafka or connectors to operate.

Best for: teams that want managed, low-latency AlloyDB replication.

Pricing: usage-based, with a free tier; check current pricing.

Pros	Cons
Low-latency managed CDC	Smaller ecosystem than the largest vendors
One capture, many destinations	Usage-based pricing to model
Backfill and streaming combined	Fewer SaaS connectors than general ELT tools

6. Google Database Migration Service

Database Migration Service moves data into AlloyDB rather than out of it, migrating PostgreSQL databases from on-premises, other clouds or Cloud SQL with continuous replication until cutover. It belongs on this list because many AlloyDB ETL projects start as a migration. For more migration options, see our guide to data migration tools.

Key features

Continuous migration: initial load followed by ongoing replication.
Cutover control: switch applications when the target is in sync.
Validation: compare source and destination before cutover.
Managed: no replication infrastructure to run.

Best for: teams moving existing PostgreSQL databases onto AlloyDB.

Pricing: no charge for homogeneous migrations into AlloyDB; underlying resources are billed.

Pros	Cons
Minimal-downtime migrations	Moves data into AlloyDB, not out
Continuous replication until cutover	PostgreSQL sources only for homogeneous migrations
No charge for homogeneous migrations	Not an ongoing analytics pipeline

7. dlt

dlt is an open-source Python library whose SQL database source reads AlloyDB tables through SQLAlchemy and loads them into DuckDB, BigQuery, Snowflake or Postgres, with incremental loading on a cursor column. It suits code-first batch pipelines. For more options, see our guide to Python ETL tools.

Key features

SQL database source: reflects tables and loads them with minimal code.
Incremental: loads only rows changed since the last run.
Schema evolution: new columns are added automatically.
Many destinations: DuckDB locally, warehouses in production.

Best for: Python teams with batch needs that do not require deletes.

Pricing: free and open source under the Apache 2.0 licence; you pay only for where it runs and where the data lands.

Pros	Cons
Pipelines as code	Batch only, no log-based CDC
Incremental loading on a cursor	Deletes are not captured
Free and open source	You run and schedule it

8. pg_dump and COPY

When you need no AlloyDB ETL tools at all, PostgreSQL's own utilities do the job: because AlloyDB is PostgreSQL-compatible, pg_dump exports schemas and data, and COPY writes tables or query results to CSV for loading elsewhere. These suit one-off exports, seeding a test environment or a quick reconciliation, not ongoing pipelines.

Key features

pg_dump: exports schemas and data for restore into another PostgreSQL-compatible database.
COPY: writes tables or query results to CSV files.
Selective exports: choose tables, schemas or query results.
Standard tooling: works with any PostgreSQL client.

Best for: one-off exports and migrations between PostgreSQL-compatible databases.

Pricing: free; you pay only for compute and storage.

Pros	Cons
Built in and free	Manual or scripted, not a pipeline
Exact control over what you export	Full exports load the primary
Standard PostgreSQL tooling	No change capture

What should you check before syncing AlloyDB data?

Whichever AlloyDB ETL tools you choose, check these first:

Logical decoding: set the alloydb.logical_decoding flag to on before creating streams, and plan for the instance restart.
Replication slots: give each CDC tool its own slot and drop slots you stop using, because an abandoned slot retains write-ahead log and fills storage.
Replica identity: tables without a primary key need REPLICA IDENTITY FULL, or updates and deletes will not replicate.
Publications: publish only the tables you need, and add new tables to the publication when they are created.
Networking: AlloyDB uses private IPs, so plan private connectivity, a proxy or self-hosted tooling inside the VPC.
Read pools: run bulk batch extracts against a read pool instance rather than the primary where possible.

How do you start pulling AlloyDB data with Airbyte?

If you decide to test Airbyte, setting up an AlloyDB pipeline takes three steps.

Step 1: Prepare AlloyDB

Enable logical decoding, create a user with REPLICATION and SELECT privileges, then create a publication and a replication slot using the pgoutput plugin. For the full procedure, see the Postgres source documentation's CDC section.

Step 2: Set up a PostgreSQL source in Airbyte

In Airbyte, open Sources, choose New source and select Postgres. Enter AlloyDB's private IP or proxy host, port 5432, database, user and password, choose logical replication, and enter the publication and slot names.

Step 3: Set up a destination and connection

Pick a destination, such as Snowflake, BigQuery or Databricks, select tables and schedule syncs. The process is the same whether you start with Airbyte Cloud or run self-hosted Airbyte Core.

Which AlloyDB ETL tool should you choose?

Choosing between AlloyDB ETL tools starts with the destination. If it is BigQuery, Datastream is the most native route and has a free allowance. If it is Snowflake, Databricks or another warehouse, or AlloyDB sits alongside many other sources, use Airbyte or Fivetran, with Estuary for low-latency managed CDC. Debezium suits event-driven systems, Database Migration Service handles moves into AlloyDB, and dlt or pg_dump cover batch and one-off needs. For BigQuery-side tooling, see our guide to BigQuery ETL tools.

Replicate AlloyDB with Airbyte

Sync AlloyDB for PostgreSQL into any warehouse with Airbyte's PostgreSQL source and log-based CDC, in Airbyte Cloud, in your own environment with Airbyte Flex, or self-hosted with Airbyte Core.

Try Airbyte Cloud free for 14 days or talk to our team.

Frequently Asked Questions (FAQs)

1. What are the best AlloyDB ETL tools?

Google Datastream is the most native route into BigQuery, while Airbyte and Fivetran reach any warehouse through PostgreSQL connectors. Debezium and Estuary suit real-time CDC, Database Migration Service handles migrations into AlloyDB, and dlt or pg_dump cover batch and one-off exports.

2. Does Airbyte support AlloyDB?

Yes, through its PostgreSQL source. Airbyte no longer lists a separate AlloyDB connector, but AlloyDB is PostgreSQL-compatible, so the PostgreSQL source connects to it, including log-based CDC with pgoutput.

3. How do you enable CDC on AlloyDB?

Set the alloydb.logical_decoding flag to on for the primary instance, which restarts it, grant REPLICATION to the user, then create a publication and a replication slot with the pgoutput plugin.

4. Is Datastream free for AlloyDB?

Datastream offers a free tier of up to 100 GiB of change data capture data a month for AlloyDB to BigQuery streams, after which it is billed per GiB processed.

5. Can you replicate from an AlloyDB read pool?

CDC streams read from the primary instance. Batch extracts can run against a read pool instance to keep load off the primary.

8 Best AlloyDB ETL Tools for PostgreSQL CDC in 2026

Related posts