Skip to main content
Last updated on

Get started with Lakebase Autoscaling

info

Lakebase Autoscaling is the latest version of Lakebase, with autoscaling compute, scale-to-zero, branching, and instant restore. For supported regions, see Region availability. If you are a Lakebase Provisioned user, see Lakebase Provisioned.

By the end of this guide, you'll have a running Postgres database with sample data, connected to Unity Catalog, with data flowing between Lakebase and the Databricks lakehouse.

Steps: ① Create a project → ② Connect → ③ Create a table → ④ Register in Unity Catalog → ⑤ Serve data → ⑥ Replicate

Step 1: Create your first project​

Open the Lakebase App from the apps switcher.

Apps switcher

Select Autoscaling to access the Lakebase Autoscaling UI.

Click New project. Give your project a name and select your Postgres version. Your project is created with a single production branch, a default databricks_postgres database, and compute resources configured for the branch.

Create project dialog

It may take a few moments for your compute to activate. The compute for the production branch is always on by default (scale-to-zero is disabled), but you can configure this setting if needed.

The region for your project is automatically set to your workspace region.

Learn more: Create a project | Autoscaling | Scale to zero

Step 2: Connect to your database​

From your project, select the production branch and click Connect. Connection strings work with any standard Postgres client (psql, pgAdmin, DBeaver, or application frameworks).

Connect dialog

To connect with your Databricks identity, copy the psql snippet from the connection dialog and paste the OAuth token when prompted:

Bash
psql 'postgresql://your-email@databricks.com@ep-abc-123.databricks.com/databricks_postgres?sslmode=require'

Learn more: Connection quickstart | psql | pgAdmin | Postgres clients

Step 3: Create your first table​

The Lakebase SQL Editor comes preloaded with sample SQL. From your project, select the production branch, open the SQL Editor, and run the provided statements to create a playing_with_lakebase table and insert sample data.

SQL Editor with preloaded sample SQL

Learn more: SQL Editor | Tables Editor | Postgres clients

Step 4: Register in Unity Catalog​

Your Lakebase database is running, but it's invisible to the rest of the Databricks platform until you register it in Unity Catalog. Once registered, you can query Lakebase tables from Databricks SQL, join operational data with lakehouse analytics, and apply unified governance.

In Catalog Explorer, create a new catalog with Lakebase Autoscaling as the type, pointing at your project's production branch and databricks_postgres database.

Register LKB database in Unity Catalog

You can now query from a SQL warehouse:

SQL
SELECT * FROM lakebase_catalog.public.playing_with_lakebase;

Learn more: Register in Unity Catalog

Step 5: Serve lakehouse data in your app​

Synced tables bring analytical data from Unity Catalog into your Lakebase database so applications can query it with low-latency transactional reads. Create a sample Unity Catalog table, then sync it to Lakebase.

In a SQL warehouse or notebook, create a source table:

SQL
CREATE TABLE main.default.user_segments AS
SELECT * FROM VALUES
  (1001, 'premium', 2500.00, 'high'),
  (1002, 'standard', 450.00, 'medium'),
  (1003, 'premium', 3200.00, 'high'),
  (1004, 'basic', 120.00, 'low')
AS segments(user_id, tier, lifetime_value, engagement);

Now sync this table into Lakebase. In Catalog Explorer, create a synced table from user_segments with Snapshot mode, targeting your project's databricks_postgres database. Snapshot mode copies the data once. For continuous updates, use Triggered or Continuous mode.

Once the sync completes, the data is available in Lakebase as default.user_segments_synced. Query it in the Lakebase SQL Editor:

SQL
SELECT * FROM "default".user_segments_synced WHERE engagement = 'high';
note

default must be quoted because it is a PostgreSQL reserved keyword. The synced table schema inherits the Unity Catalog schema name, so if your schema is named default, you must always quote it in queries. Quotes around other identifiers are optional.

Query synced user segments in Lakebase SQL Editor

Your lakehouse analytics are now ready to serve from your transactional database.

Learn more: Synced tables | Sync modes | Data type mapping

Step 6: Replicate to the lakehouse​

note

Lakehouse Sync is in Beta.

Lakehouse Sync continuously replicates your Lakebase Autoscaling tables out to Unity Catalog as Delta tables using change data capture (CDC). Each change is appended as a new row, preserving full history. No external pipelines, jobs, or Spark required.

Postgres needs to record full row data in the write-ahead log for CDC to work. Set replica identity to full on the playing_with_lakebase table you created earlier:

SQL
ALTER TABLE playing_with_lakebase REPLICA IDENTITY FULL;

Then configure the sync from the Lakehouse sync tab in your branch overview. Choose the public schema as the source and a Unity Catalog catalog and schema as the destination.

All tables in the source schema are synced automatically. The playing_with_lakebase table appears in your destination catalog as lb_playing_with_lakebase_history. Query it from a SQL warehouse:

SQL
SELECT * FROM <catalog>.<schema>.lb_playing_with_lakebase_history LIMIT 10;

Each row includes system columns _change_type, _timestamp, _lsn, and _xid that track how and when the data changed.

Learn more: Lakehouse Sync | Data type mapping | Current-state mirror view

Next steps​

On this page