Databricks Apps Cost and Runtime Analysis
Open Apps in the sidebar to investigate the cost of Databricks Apps across workspaces. Select a time range and workspace filters before comparing apps.
This page monitors your Databricks Apps. It is separate from the option to deploy LakeSentry itself as a Databricks App.
Data requirements
Section titled “Data requirements”Apps analytics use ingested Databricks billing usage and pricing, with app identifiers and metadata where available. First complete connector setup and a successful sync. An app without usage in the selected range may not appear in the inventory; this is not a live list of every deployed app.
Overview tab
Section titled “Overview tab”Use Overview to compare app cost and activity. The trend can be grouped by app, workspace, or compute size, and switched between spend and active apps. Open an app to inspect its cost, runtime, creator, compute size, workspace, and last activity.
Interpret the activity fields carefully:
- Runtime reflects billed active hours.
- Utilization is the share of window hours with activity, not measured CPU or memory utilization.
- Status uses recent billing independently of the selected date range. Always-on, Intermittent, Idle, and Inactive describe billing patterns, not a live health check.
- If creator or compute metadata is missing, use the app identifier and workspace to find the app in Databricks.
Optimization tab
Section titled “Optimization tab”The Optimization tab combines open Apps insights with two review tables:
| Review | What to investigate |
|---|---|
| Off-hours Review | Apps incurring cost outside working hours or on non-working days; costs cover the selected range |
| Compute Sizing Review | Compute size, inferred concurrent instances, and estimated monthly savings from scaling in or downsizing |
Admins and owners can set the organization’s Working hours in Settings → General, including timezone and business hours. Workspace overrides are available in Settings → Workspaces. Without a configured schedule, detection uses a UTC-based estimate; check the basis shown for each app before treating activity as avoidable spend.
Inferred instances come from billed usage rates. They are estimates, not a measured replica count. Scale-in and downsize savings describe different alternatives; do not add them together as guaranteed savings. Confirm real load, availability requirements, and the actual Databricks configuration before changing an app.
The Insight Type Catalog describes the Apps findings: App Cost Spike and App Stop Schedule.
Troubleshooting
Section titled “Troubleshooting”| Symptom | Check and next action |
|---|---|
| No apps found | Widen the date range, clear filters, and check that billing extraction completed with app usage in scope |
| Status does not change with the date range | Status uses recent billing; costs and trends use the selected range |
| High utilization but little traffic | Utilization measures billed activity, not request volume; verify demand in Databricks |
| Off-hours costs seem wrong | Ask an admin to check organization working hours, timezone, and workspace overrides |
| Analytics or insight stats unavailable | Retry and check pipeline status |