Analyzes Kueue repository guidance to produce structured Kubernetes queuing knowledge for workload admission scenarios.
npx skills add https://github.com/tankpkg/packages --skill tank-kueue
Please help me install this Agent Skill. Skill: @tank/kueue Source: https://github.com/tankpkg/packages/tree/main/skills/kueue Command: npx skills add https://github.com/tankpkg/packages --skill tank-kueue
Kueue helps teams control when Kubernetes workloads start by turning ad hoc job submission into governed queueing with quotas, priorities, and admission checks, reducing cluster contention and wasted capacity.
It covers ResourceFlavor, ClusterQueue, LocalQueue, Workload, Cohort, AdmissionCheck, and WorkloadPriorityClass flows for batch, AI/ML, HPC, and multi-cluster operations. Use it to configure production job admission, share GPU or CPU capacity across teams, enable preemption and fair sharing, wire in autoscaling or federation, and troubleshoot stalled workloads with metrics and CLI guidance.
Ask the Skill to explain how to install, configure, and operate Kueue for a Kubernetes cluster with shared GPU workloads and fair queueing.
High-intent search queries and answers about installing and using this skill.
You manage Kubernetes quota for shared GPU workloads by configuring ClusterQueue and LocalQueue resources in Kueue, which governs job admission and capacity sharing across teams to reduce cluster contention.
Workload admission control in Kubernetes governs when submitted jobs start executing based on quotas, priorities, and admission checks. You need it for batch, AI/ML, and HPC scenarios where shared cluster capacity requires governed queueing to prevent resource contention.
You configure preemption and fair sharing for Kubernetes batch jobs by defining Cohort and WorkloadPriorityClass resources in Kueue, enabling prioritized workload admission and automatic suspension of lower-priority jobs when capacity is constrained.
Kueue supports multi-cluster workload queueing through MultiKueue configurations and integrates with cluster autoscaling, allowing batch and HPC workloads to be admitted and dispatched across federated Kubernetes clusters based on available ResourceFlavor capacity.
You troubleshoot stalled Kubernetes workloads by inspecting Kueue observability metrics and CLI outputs to diagnose admission check failures, ResourceFlavor shortages, or ClusterQueue quota limits blocking workload activation in the queueing system.
Yes, Kubernetes quota management supports topology-aware scheduling through Kueue ResourceFlavor configurations, allowing batch and AI/ML workloads to target specific hardware topologies for optimized GPU scheduling and HPC performance.
GoldenWing-360Analyze Kubernetes cluster configurations for security weaknesses and hardening best practices.
pluginagentmarketplaceIntegrate Prometheus, Grafana, Loki, and OpenTelemetry for Kubernetes observability.
scitixDiagnose Volcano-managed Pod scheduling failures via kubectl cluster state queries.
scitixDiagnose Kubernetes ResourceQuota and LimitRange admission rejections preventing pod creation.
