After-Hours Resource Governance

Your Dev Environment
Shouldn't Run All Weekend

Clusters left running, warehouses nobody's querying, apps nobody's using. Kostavo scans for resources that are active outside their expected operating window and stops them automatically.

The After-Hours Problem

Clusters & Warehouses Running Overnight

A developer starts a cluster on Friday afternoon and forgets to stop it. It runs all weekend. Databricks won't stop it for you.

Apps Running With No Users

Databricks Apps keep running after work hours even when no one is connected. Dev apps don't need to be available at 3am.

Nobody Notices Until Monday

Without continuous scanning, nobody knows what's running over the weekend, and forgotten resources quietly run for days.

What Kostavo Does

App Scheduling (Start & Stop)

Kostavo can automatically stop Databricks Apps after operating hours and start them again before your team arrives. The only resource type with full start/stop scheduling.

  • Stop apps after operating hours
  • Start apps before business hours
  • Configurable per app via policy profiles

Running Resource Detection

Kostavo scans for clusters and warehouses that are currently running. On finding them active, it can notify the resource owner or stop them automatically.

  • Running cluster detection
  • Running warehouse detection
  • Notify or stop automatically

App Compute Size Policies

Beyond scheduling, Kostavo also verifies that Databricks Apps aren't running on compute that's larger than allowed: flags oversized app compute and notifies the owner.

  • Max compute size per app
  • Overage notifications
  • Owner + admin alerts

Kostavo flags compute that is still running, like a GPU cluster idle since Friday evening.

Resources Covered

Databricks Apps

Full start/stop scheduling: stop after hours, start before work

Compute Clusters

Running cluster detection: notify or stop automatically

SQL Warehouses

Running warehouse detection: notify or stop automatically

Model Serving Endpoints

Provisioned throughput and scale-to-zero disabled detection

Vector Search Endpoints

Running Vector Search endpoints: notify or flag for removal

Continuous Jobs & Pipelines

Continuously-running jobs and declarative pipelines in continuous mode: notify or stop

High-Frequency Alerts

SQL Alerts scheduled to run too frequently against a warehouse

Stop Paying for the Weekend

Connect a workspace and see what is still running this weekend.

Start Free Trial