Documentation
ProductPricingDocsBlog

Overcommit

Suggest Edits

Overcommit

When running Hydrolix on a very small cluster for exploration and testing, the configured memory and CPU resource requests and limits of each container can cause Kubernetes to refuse to deploy new pods. In resource-constrained deployments where many containers request more resources than they actually need, this manifests in new pods unnecessarily being held in the "Pending" state when resources are actually available.

The top-level overcommit tunable avoids this situation by causing the scheduler to ignore any resource requests or limits configured for the containers within the pods.

❗️

Not for production

This setting is not recommended for production environments.

Edit your cluster's configuration to use this override. For example, to set it to true:

.....
spec:
  .....
  overcommit: true
    .....

There are four allowable values for this setting:

ValueDescription
falseScheduler behaves normally, abiding by container requests and limit configuration. Default behavior
trueScheduler ignores both request and limit configuration.
requestsScheduler ignores requests, but considers limits. Introduced in v5.4.
limitsScheduler ignores limits, but considers requests. Introduced in v5.4.

Learn more about Kubernetes resource requests and limits in the Kubernetes documentation.

Updated 1 day ago