Skip to main content
A Lakebase project can keep a minimum amount of compute running at all times, which bills even when no one is connected. nOps finds projects whose always-on minimum runs almost every hour while sitting idle outside business hours, and estimates the saving from letting the compute scale to zero.

Why it matters

Development, test and other non-production databases are often created with an always-on minimum and then used only occasionally.

How nOps finds it

nOps reads the last 30 days of Databricks billing for each Lakebase project’s always-on minimum compute. A resource is flagged when it was billed in at least 90% of the hours of the last 30 days (or since your first Databricks day, if later), was still running on the last complete day, and sat at its minimum level (within 5% of its 10th-percentile hourly usage) in at least 80% of off-hours. Off-hours are weekends and weekday hours outside 13:00-24:00 UTC. The saving is what that minimum cost in those idle off-hours, scaled to 30 days, and must be at least $10 a month. nOps also shows how much autoscaled compute the project used beyond the minimum, as a sign of whether it is in use.

Before you start

Know which applications connect to the project and whether a short delay on the first query after idle is acceptable.

How to implement

  1. Check whether the project is still used. If it shows little or no compute beyond the minimum, ask the owner whether it can be deleted.
  2. Otherwise, open the Lakebase project in Databricks, edit the compute settings for the branch, and turn on scale to zero (suspend after a period of inactivity).
  3. Start with development, test and other non-production branches.
  4. Check after a week that the always-on charge has dropped.

Risks and considerations

  • The first query after an idle period waits while compute resumes.
  • Applications that hold connections open or poll the database can keep it awake, which reduces the saving.
  • To roll back, turn scale to zero off again.