Uploaded November 2026 | Updated November 2026, 3 days ago
Databricks Lakebase high availability runs your serverless Postgres on a primary plus up to three secondaries in other availability zones, with automatic failover behind the same connection string. In this video we turn it on while an app keeps writing, check the replicas, and use them for reads.
β° Chapters
00:00 - Lakebase High Availability Explained: Primary and Secondaries
00:53 - Turn On High Availability While the App Keeps Writing
01:44 - Check the Secondaries: Connection Strings, Roles and Freshness
02:49 - Read Offload: Analytics on Primary vs Read-Only Secondaries
03:44 - What Happens When the Primary Fails
04:28 - Results and When to Turn On High Availability
π‘οΈ How Lakebase high availability works
High availability is a setting on a Lakebase compute. It adds one to three secondaries, each in a different availability zone, kept in sync with the primary. If the primary fails, Lakebase promotes a secondary with every committed transaction preserved, and the same connection string points to the new primary. With read access on, the secondaries also serve read-only queries through a second -ro connection string. Scale to zero is not available with high availability, because the secondaries stay warm.
π οΈ What you'll see in the demo
Using a coffee shop project on 2 compute units, with an app inserting a row every quarter of a second:
β’ Switching Edit compute to High Availability with 3 instances, 1 primary and 2 read-only secondaries
β’ What the change cost the app: 1 failed write, with writes back after 1.7 seconds
β’ The compute table with roles and access, and the primary and -ro connection strings
β’ pg_is_in_recovery on each host, and the read-only replica refusing an INSERT
β’ Replica freshness: in 5 out of 5 tries, a new row was already on the replica at the first read
β’ 16 analytics clients on the primary at 15.3 queries per second, then on the secondaries at 29.9 queries per second, while the app's writes stayed at about 165 milliseconds
β’ What failover does, and what to configure in your app: retries, keepalives and two or more secondaries
Whether you're running a production app that can't go down, a platform engineer planning Postgres availability, or a team moving analytics off the primary, Lakebase high availability gives you standby secondaries that also earn their keep by serving reads.
π Resources
π High availability documentation: docs.databricks.com/aws/en/oltp/projects/high-availability
βοΈ Manage high availability: docs.databricks.com/aws/en/oltp/projects/manage-high-availability
π Databricks Lakebase: databricks.com/product/lakebase
π€ Stay connected
Website: databricks.com
Twitter: https://x.com/databricks/
LinkedIn: linkedin.com/company/databricks
Instagram: instagram.com/databricksinc
Facebook: facebook.com/databricksinc
Databricks Lakebase high availability runs your serverless Postgres on a primary plus up to three secondaries in other availability zones, with automatic failover behind the same connection string. In this video we turn it on while an app keeps writing, check the replicas, and use them for reads.
β° Chapters
00:00 - Lakebase High Availability Explained: Primary and Secondaries
00:53 - Turn On High Availability While the App Keeps Writing
01:44 - Check the Secondaries: Connection Strings, Roles and Freshness
02:49 - Read Offload: Analytics on Primary vs Read-Only Secondaries
03:44 - What Happens When the Primary Fails
04:28 - Results and When to Turn On High Availability
π‘οΈ How Lakebase high availability works
High availability is a setting on a Lakebase compute. It adds one to three secondaries, each in a different availability zone, kept in sync with the primary. If the primary fails, Lakebase promotes a secondary with every committed transaction preserved, and the same connection string points to the new primary. With read access on, the secondaries also serve read-only queries through a second -ro connection string. Scale to zero is not available with high availability, because the secondaries stay warm.
π οΈ What you'll see in the demo
Using a coffee shop project on 2 compute units, with an app inserting a row every quarter of a second:
β’ Switching Edit compute to High Availability with 3 instances, 1 primary and 2 read-only secondaries
β’ What the change cost the app: 1 failed write, with writes back after 1.7 seconds
β’ The compute table with roles and access, and the primary and -ro connection strings
β’ pg_is_in_recovery on each host, and the read-only replica refusing an INSERT
β’ Replica freshness: in 5 out of 5 tries, a new row was already on the replica at the first read
β’ 16 analytics clients on the primary at 15.3 queries per second, then on the secondaries at 29.9 queries per second, while the app's writes stayed at about 165 milliseconds
β’ What failover does, and what to configure in your app: retries, keepalives and two or more secondaries
Whether you're running a production app that can't go down, a platform engineer planning Postgres availability, or a team moving analytics off the primary, Lakebase high availability gives you standby secondaries that also earn their keep by serving reads.
π Resources
π High availability documentation: docs.databricks.com/aws/en/oltp/projects/high-availability
βοΈ Manage high availability: docs.databricks.com/aws/en/oltp/projects/manage-high-availability
π Databricks Lakebase: databricks.com/product/lakebase
π€ Stay connected
Website: databricks.com
Twitter: https://x.com/databricks/
LinkedIn: linkedin.com/company/databricks
Instagram: instagram.com/databricksinc
Facebook: facebook.com/databricksinc










