Azure Databricks Runtime 10.4 LTS reaches end of life; clusters and jobs must move to a supported Databricks Runtime version

Microsoft is retiring an old managed runtime on Azure Databricks. In an Azure Updates notice, Microsoft says Azure Databricks Runtime 10.4 LTS will reach end of life on 1 November 2026. This is the final step in a lifecycle that has been winding down for a while: the runtime reached end of support on 18 March 2025, and end of life now removes it altogether.

The wording is unusually blunt for a runtime notice. After 1 November 2026, Microsoft says, Databricks Runtime 10.4 LTS “will no longer be available or usable, and workloads remaining on this version will no longer run.” That is a hard shutoff, not a support cut-off — a job pinned to 10.4 LTS does not degrade quietly, it stops starting.

Who this affects

Anyone with Azure Databricks clusters, scheduled jobs, or pipelines still pinned to Databricks Runtime 10.4 LTS (10.4.x). Because it is a long-term-support release, it is exactly the kind of version that gets set once in a cluster policy or job definition and then left alone — which is precisely how a workload ends up running on a 2022-era runtime three years later. Interactive clusters, Delta Live Tables or Jobs pinned to 10.4, and Databricks Asset Bundles that hard-code the runtime are all in scope.

What to do

Move affected workloads to a supported Databricks Runtime version before 1 November 2026:

1. Inventory every cluster policy, job definition, and bundle whose Spark/runtime version is 10.4.x LTS.
2. Re-target each to a current supported LTS runtime. A gap this wide can bring Spark, Python, and library behaviour changes, so treat it as a real upgrade rather than a version bump.
3. Test each workload on the new runtime, then cut over. 10.4 LTS keeps running until the date, so you can migrate and validate one job at a time instead of in a single big-bang cutover.