Kueue plugin for Headlamp traces why a Kubernetes Job is still pending
Victoria Shinder2 min readThe Kubernetes blog published a walkthrough on 14 October 2026 for the Kueue plugin for Headlamp. The question it answers is a common one in clusters that use Kueue for batch queuing: why a Job stays Pending.
Admission before scheduling. The article's starting point is that Kueue and the Kubernetes scheduler answer different questions. Kueue decides whether a Workload should start now, given queue policy, quota, ResourceFlavors, admission checks and other waiting Workloads; the scheduler then places Pods. If the Workload is not admitted, the investigation stays in Kueue; if it is admitted but Pods are still Pending, it moves to Headlamp's normal Job and Pod views and scheduler events.
What the plugin shows. Release 0.1.0-alpha provides read-only views for Kueue v1beta2 Cohorts, ClusterQueues, LocalQueues, ResourceFlavors and Workloads. It adds no create, edit, stop, resume or delete workflows of its own; Headlamp's standard YAML and resource actions remain available where permissions allow. The plugin checks permissions and shows an access message when the current identity cannot read a resource.
The path through the UI. The walkthrough starts at Kueue → Workloads, finds the Workload Kueue created for the Job, and reads its conditions. The two that matter most are QuotaReserved and Admitted: Admitted=True hands the problem to scheduling, while QuotaReserved=False with a reason such as Pending means quota was not reserved, and the condition message says why — for example, that a ResourceFlavor could not be assigned because the ClusterQueue lacks CPU quota. An Inadmissible status means the Workload cannot fit the current queue configuration. From there the article follows links to the LocalQueue, the namespace entry point, and the ClusterQueue behind it to compare the request with capacity and usage.
A detail worth noticing. The Workload page shows PodSets with the total request. A Job with parallelism: 3 and one CPU per Pod asks Kueue for three CPUs; reading only one Pod template underestimates it. In the alpha, owner references appear as text rather than links, so the article suggests the namespace filter and table search to find the Workload.
Prerequisites listed: a cluster with Kueue installed, a Headlamp installation, the 0.1.0-alpha plugin, and an identity allowed to read the Kueue resources.
Source: Kubernetes Blog, "Why Is My Kubernetes Job Still Pending? Tracing Kueue Admission with Headlamp" — https://kubernetes.io/blog/2026/10/14/tracing-kueue-admission-with-headlamp/