Skip to main content
Install Kueue on your CKS cluster from the CoreWeave-published Helm chart (coreweave/cks-kueue), then define ResourceFlavor, ClusterQueue, and LocalQueue resources to control how jobs flow through the cluster. To divide capacity among teams and let them borrow from each other, group the ClusterQueues in a Cohort. The chart also enables Topology-Aware Scheduling by default, which you can configure with CKS Node labels to co-locate Pods on InfiniBand or NVLINK topologies. For full details, see Kueue and the worked example in Run Ray with Kueue.
Workload Scheduling
Last modified on September 29, 2026