Download
Kueue is a set of APIs and controllers for job queueing. It is a job-level manager that decides when a job should be admitted to start (as in pods can be created) and when it should stop (as in active pods should be deleted). Use Kueue to build a multi-tenant batch service with quotas and a hierarchy for sharing resources among teams in your organization. Based on the available quotas, Kueue decides when jobs should wait and when and where they should run. Kueue works in combination with standard kube-scheduler, cluster-autoscaler, and the rest of the Kubernetes ecosystem. This combination allows Kueue to run both on-prem and in the cloud, where resources can be heterogeneous, fungible, and dynamically provisioned.
Features
- Support job queueing based on priorities with different strategies: StrictFIFO and BestEffortFIFO
- Support resource fair sharing and preemption with a variety of policies between different tenants
- Quota borrowing or preemption in ClusterQueue and Cohort
- Built-in support for popular jobs, e.g. BatchJob, Kubeflow training jobs, RayJob, RayCluster, JobSet, plain Pod
- Build-in prometheus metrics to help monitor the state of the system, as well as Conditions
- A mechanism for internal or external components to influence whether a workload can be admitted
- Integration with cluster-autoscaler's provisioningRequest via admissionChecks
- A simple implementation of all-or-nothing scheduling
Project Samples
