/
New feature

Compute metrics

Compute metrics is a Databricks compute / bi capability, introduced April 2023.

Ganglia metrics Compute metrics

The native metrics tool in the Databricks UI - the Metrics tab on all-purpose and jobs compute - showing near-real-time hardware and Spark metrics for the whole cluster or per node, collected every minute and kept for 30 days.

  • It arrived on April 14, 2023 alongside Databricks Runtime 13.0, promising a live view of cluster performance plus custom minute-by-minute history instead of periodic snapshots.
  • The metrics live in Databricks-managed storage rather than yours, so there is nothing to mount, nothing to export by hand, and nothing to delete by accident.
  • Serverless compute does not get it at all - notebooks and jobs on serverless are sent to query insights instead.

Limitations: History covers only the last 30 days at one-minute granularity; GPU metrics need Databricks Runtime ML 13.3 and above and are shown only at the individual-node level; the By node view cannot be downloaded; serverless compute for notebooks and jobs uses query insights instead of the metrics UI.

Open in REbricked →
Category
Compute / BI
Introduced
April 2023
Also known as
Cluster metrics, Cluster metrics UI, Metrics tab
Verified
2026-09-11

Sources

Related in Compute / BI