LoginSign Up
← Back to all sparks
Kubernetes logo

Kubernetes

DEVOPSINFRA · APIS
Velocity5.0

Container orchestration

kubernetes.io ↗

Kubernetes moves conformance from the cluster to the workloads running on it.

conformanceai conformanceworkload standardscgroup v2governance
◆Current state
Kubernetes is between releases, and its feed is v1.37 feature deep-dives, migration guidance and governance. The newest item confirms the AI Conformance working group has been archived and its work is now a permanent subproject under SIG Architecture, with certified platforms up from 18 to 31 within a few months. It follows the chartering of a Workload Conformance Working Group to test applications, not clusters.
◆Where it's heading
The project keeps codifying operational knowledge that operators used to hold in their heads: unused-PVC conditions, pod-level resource managers, and now a planned spec for how workloads behave during drains, upgrades and resource use. Conformance is widening from Certified Kubernetes to AI Conformance to workloads, and AI Conformance moving under the group that owns overall conformance makes it a standing program rather than a temporary effort.
◆Prediction
Expect automated AI Conformance verification, already on the program's roadmap, and a first Workload Conformance requirements draft for community input. In the meantime, the blog will likely keep shipping migration FAQs and community spotlights.

◆Recent moves

  1. 16d ago

    From Working Group to SIG Architecture: spotlight on AI Conformance

    A SIG Architecture interview confirms the AI Conformance working group met its goals and was archived, with the work now a permanent subproject; certified platforms grew from 18 to 31 within a few months. No code ships, but it turns the conformance widening of recent weeks into standing governance.

    View source ↗
  2. 18d ago

    Announcing the Workload Conformance Working Group

    A new Workload Conformance Working Group will write a per-minor-version spec for how applications behave on Kubernetes, covering disruption tolerance, privileges, resource use and API usage, plus upstream tests to check it. It extends the conformance idea from clusters to workloads and gives operators a shared baseline instead of repeating hand audits before every upgrade. Nothing is specified yet; the target is a stable first spec in two to three releases.

    View source ↗
  3. 1mo ago

    Why Is My Kubernetes Job Still Pending? Tracing Kueue Admission with Headlamp

    A walkthrough of the 0.1.0-alpha Kueue plugin for Headlamp, which shows read-only Workload, queue, ResourceFlavor and Cohort views so a stuck Job's admission can be traced in the UI instead of across YAML dumps. It extends the recent run of operator guidance, this time for batch and AI jobs queued behind quota.

    View source ↗
  4. 1mo ago

    The Shift to cgroup v2 in Kubernetes: What You Need to Know

    A migration FAQ for cgroup v2, prompted by v1.35 making the kubelet fail on cgroup v1 nodes by default and kubeadm preflight now erroring on them. It is the prerequisite for the release's memory features, which assume the unified hierarchy, and it signals that full v1 removal is coming.

    View source ↗
  5. 1mo ago

    Scaling Kubernetes Workloads with Node Swap

    A benchmark write-up on backing node swap with NVMe SSDs to page out idle memory and fit more pods per node, framed around agentic AI workloads that sit resident between prompts. It is guidance on an existing GA feature, but it ties the v1.37 memory work to a concrete density argument.

    View source ↗
  6. 1mo ago

    Spotlight on SIG Apps

    A community profile of SIG Apps, the group behind Deployments, StatefulSets and Jobs. No product change.

    View source ↗