Review Databricks Runtime support policy

Check Databricks Runtime support dates and plan version selection and updates for long-running operations.

Description

A Databricks Runtime release past its support period no longer receives security fixes or support, which can increase operational risk. Non-LTS releases can still be supported; select a runtime using its actual end-of-support date and workload requirements.

databricks_spark_version selects a Databricks Runtime identifier, not just an Apache Spark version. Consider LTS for long-lived deployments while also managing library compatibility and updates.

Potential impact

  • After support ends, security fixes and troubleshooting support are no longer available.
  • Shorter support periods may require more frequent compatibility testing and upgrades.

Remediation

Choose a supported Databricks Runtime that suits the operating period and upgrade before support ends. Use long_term_support = true when extended support is needed; in provider v1.133.0, this filter includes LTS and ESR releases. Test the selected version’s compatibility and plan the restarts needed to apply updates.

Examples

These excerpts show GPU/ML runtime selection and omit other cluster settings such as the node type. The before example does not filter for long-term support; that alone does not mean it selects an unsupported runtime.

Before

hcl
data "databricks_spark_version" "gpu_ml" {
  gpu = true
  ml  = true
}

resource "databricks_cluster" "example" {
  cluster_name  = "Research Cluster"
  spark_version = data.databricks_spark_version.gpu_ml.id
}

After

hcl
data "databricks_spark_version" "gpu_ml" {
  gpu               = true
  ml                = true
  long_term_support = true
}

resource "databricks_cluster" "example" {
  cluster_name  = "Research Cluster"
  spark_version = data.databricks_spark_version.gpu_ml.id
}

The after example restricts candidates to LTS and ESR releases. This filter does not complete an upgrade automatically, so check the selected version and support period and apply updates to the actual cluster.

References