Overcommit

Overview

In small or test clusters, Kubernetes may block new pods if their configured CPU and memory requests exceed available resources, even if the pods don’t actually need that much.

The overcommit setting avoids this by telling the scheduler to ignore some or all resource requests and limits.

⚠️

Not for production

Overcommit is only for testing or exploration. Don’t enable it in production.

Configure overcommit

Set overcommit in the top-level spec of your cluster configuration:

spec:
  overcommit: true

Allowed values

Choose the level of enforcement you want. These settings control whether Kubernetes respects resource requests, limits, both, or neither:

ValueBehavior
falseDefault. Kubernetes schedules pods normally, respecting both requests and limits.
trueIgnores both requests and limits. Pods run even if declared resources exceed cluster capacity.
requestsIgnores requests but enforces limits. Pods can start on constrained nodes but can’t exceed defined limits. (Added in v5.4)
limitsIgnores limits but enforces requests. Pods start only if requests fit, but may burst above defined limits. (Added in v5.4)

Related