Skip to content

Databricks Work Unit Cost Tracking

Work Units is the workload-centric view of LakeSentry. It groups related runs and usage into units you can investigate, compare, and attribute.

  • Jobs: Databricks Jobs and job tasks.
  • Pipelines: Lakeflow / Delta Live Tables pipelines.
  • SQL Analytics: SQL workloads associated with warehouses and statements.
  • Model Serving: serving endpoints and served entities where system tables are available.

The list shows cost, run/activity counts, success or uptime, owner, workspace, category, and trend. Search and filters help narrow by workspace, type, owner, principal, and cost.

For batch workloads, metrics focus on cost per run, run count, duration, and success rate. For continuous or serving-like workloads, metrics focus on cost per hour, uptime, and restart/activity signals.

At the top of the page, category cards summarize total cost and activity by workload type. Click a card to filter the list to that category.

Open a work unit to review:

  • Total and average cost modes.
  • Cost trend over the selected time range.
  • Run or activity history.
  • Associated cluster, SQL, serving, or pipeline details when available.
  • Owner and attribution information.
  • Related insights and action plans.

Other pages link back to Work Units when a cost driver, anomaly, query, principal, or cluster can be connected to a specific workload.