Horizontal scaling for Databricks Apps is now generally available

Horizontal scaling for Databricks apps allows you to distribute app traffic across multiple instances to ensure high availability and higher concurrency, providing zero-downtime updates and session affinity. ### Use cases Imagine you have a production enterprise app receiving simultaneous accesses that cannot fail during code updates. With this feature, you can: - Distribute requests across 1 to 5 instances. - Perform zero-downtime deployments with updates rolling out to build instances first. - Keep temporary per-user data using best-effort session affinity. ### How to use To route API clients to the same instance, include the routing header in your requests: bash curl -X GET https://<your-app>.aws.databricksapps.com/api/endpoint -H "Authorization: Bearer YOUR_BEARER_TOKEN" -H "X-Routing-Key: f8822466-3b1e-423a-988b-54c9e639c250" What happens in this code: 1. The HTTP GET request is sent to your Databricks app URL. 2. The authorization token validates the access. 3. The X-Routing-Key header ensures the request is routed to the same app instance. ### Limitations and requirements - The feature is generally available (GA). - Each workspace can have at most 20 horizontally scaled apps. - Each horizontally scaled app can have at most 5 instances (contact support to raise limits). - Newly created horizontally scaled apps do not include pre-installed Python libraries; declare dependencies in requirements.txt or pyproject.toml.