If you've ever stared at a Job that never leaves Pending because the controller never creates Pods, congratulations you've experienced one of Kueue's operational blind spots. Headlamp's new Kueue plugin removes that blind spot by surfacing admission and workload-state information in the UI so you can see why a Job is suspended before a single Pod exists.
That's the practical change people will use today. The plugin exposes Kueue admission traces and the workload CRD state in Headlamp's UI, so operators no longer have to toggle between kubectl, multiple CRDs, and controller logs to answer the simple question: has the scheduler accepted this workload or is it waiting for resources or constraints? Observability for control-plane decisions belongs in the same place teams look for what to run next.
Kubernetes published an alpha snapshot that includes ongoing scheduling and workload-management work; the alpha itself doesn't change Kueue semantics but reflects continued ecosystem activity in this area. The Headlamp Kueue integration comes from the community need to reduce debugging round-trips: when admission happens at the Kueue layer, there are no Pods to inspect and pod-centric tooling comes up short.
Why this matters: teams building batch orchestrators, ML training controllers, or CI platforms on top of Kueue now get first-class signals about admission decisions, preemption, and constraints that prevented Pod creation. Fewer paging incidents should start with better UX for control-plane state and Headlamp delivering that UX is a sensible, practical improvement.
runc 1.6.0-rc.1 was released recently; the change set continues the runtime project's work toward stable cgroup v2 behavior and tighter alignment with how Kubernetes expects runtimes to behave under a unified hierarchy. Kubernetes has been converging on cgroup v2 in recent releases, and platform operators should treat the move away from cgroup v1 as an operational project: kernel config, kubelet flags, CRI implementations like containerd, and CI images need validation under cgroup-v2 semantics.
Notably absent in this snapshot: no new containerd release, no OCI spec changes called out, and no major CNCF graduations. That silence matters; ecosystem upgrades are happening in pieces. runc moves; Kubernetes documents cgroup v2 as the future; containerd, distro packages, and downstream tools will be the gating factors for many clusters.
If you want a short checklist of what to do next: (1) instrument CI to boot nodes with cgroup v2 and run kubelet/CNI/CSI smoke tests, (2) validate privileged tooling and init systems under unified-hierarchy semantics, and (3) add Kueue admission traces to your on-call playbooks so you're not debugging by grepping controller logs at 2 a.m.
Final thought: Headlamp's Kueue plugin is small but consequential it fixes the right UX problem. runc 1.6-rc1 is the loud reminder that the runtime and kube are converging on cgroup v2, and the friction will land squarely in platform teams responsible for kernels, distributions, and the CRI ecosystem. If you haven't already started converting tests and CI to cgroup v2, the time to begin is now.
For a deeper look at runc implications for Kubernetes platform teams, see my piece on runc 1.6.0-rc.1 and cgroup v2 alignment, and if you're prepping cluster changes, consult Kubernetes guidance on migrating away from cgroup v1.