Databricks publishes four-stage guide to scaling Databricks Apps

Databricks publishes a guide to scaling Databricks Apps in four stages: design, build, test and deploy, and operate.

The guide recommends setting latency and peak-load targets first, keeping session state rebuildable, and using load tests to choose instance size and count.

The guide estimates Lakebase database connections as instances × workers × pool size. It states that the default limit is five instances per app, and that a Databricks representative can raise the limit if load testing shows more is needed. It also says each user who queries Lakebase as themselves needs their own Postgres role, and that Lakebase allows 500 roles per branch.