Databricks launches Genie Migrate to move classic workloads to serverless

Databricks has launched Genie Migrate for Serverless in Beta, an agent that scans classic jobs and notebooks, proposes the changes needed for serverless compute, and applies approved migrations with a 30-day rollback. The company says the workflow turns migration into a review-and-approve process instead of weeks of manual troubleshooting.
- Genie Migrate for Serverless is available in Beta.
- The agent scans workloads and shows proposed changes as reviewable diffs.
- Migrations use cloned jobs and include a 30-day rollback.
- Bulk job migration requires Databricks CLI version 1.24.0 or above.
- Sandbox runtime validation and notebook support for bulk migration are coming soon.
A managed path from classic to serverless
Genie Migrate for Serverless is designed for organizations that still operate classic compute workloads but want to use Databricks serverless compute. Databricks says serverless handles provisioning, scaling and optimization, reducing the infrastructure work required to run data and AI workloads.
The agent runs from the /compute migrate slash command inside Genie on any job or notebook. It is fully managed and requires no setup to begin scanning. According to Databricks, the agent reads job configuration and workload code inside the customer’s own workspace while leaving workload code untouched unless an approved migration requires a change.
Assess, translate and review every change
Genie Migrate for Serverless first scans Jobs and Notebooks and classifies their readiness. It identifies workloads that are suitable for migration, those that need an administrator, and those that are not ready. Workloads whose source of truth is outside the workspace, including jobs managed by Databricks Asset Bundles or notebooks in a Git folder, cannot be migrated yet.
For an eligible workload, the agent creates a labeled, side-by-side diff before writing anything. The proposed translation can include mapping a classic cluster configuration to serverless compute, resolving libraries against the serverless runtime, synthesizing an environment definition and carrying over applicable Spark configurations.
Configurations that may not apply are surfaced for the user’s decision instead of being silently removed. Databricks also says the agent can rewrite code patterns that do not run on serverless, including APIs incompatible with Spark Connect, with those edits shown in the reviewable diff.
Conservative application and bulk migration
Before applying a migration, the agent checks that the translation is coherent, library dependencies resolve against the serverless runtime and the configuration applies cleanly. It then applies changes to a cloned job rather than preempting the running workload. Users can run the clone on demand to check its health before promoting the migration to the original job.
Databricks says the current validation is static and does not run the workload. A deeper Sandbox validation capability is coming, allowing a migrated workload to run in an isolated environment and compare its output with a classic baseline. The company also cautions that rollback does not undo data written to production tables by a test run using changed versions.
For larger fleets, the Databricks CLI can start migrations for multiple jobs through the compute-migrations command group. The agent creates serverless working copies without an open Genie session, while the original job remains unchanged until approval. Bulk migration supports jobs today, with notebook support coming soon, and requires Databricks CLI version 1.24.0 or above. Approved migrations retain the job’s ID, schedule and permissions.
What this means for Databricks users
Genie Migrate for Serverless targets the translation work that often makes moving workloads difficult, including access-mode incompatibilities, Spark configuration differences and library resolution. Its reviewable diffs, cloned jobs and reversible workflow give teams control over what changes before a production migration is approved.
The tool is available in Beta, and Databricks says users can get started today. Its current limits, including external-source workloads and static rather than runtime validation, mean teams still need to review proposed changes and monitor migrated runs.
Databricks: Introducing Genie Migrate for Serverless


