Kubernetes 1.37, with Dipesh Rawat
Source: YouTube · The Kubernetes Podcast from Google · published Aug 26, 2026 · 30:53
BLUF: The Kubernetes 1.37 release, themed "Garhwal," introduces significant enhancements for AI workloads, cost efficiency, and security, alongside critical deprecations that require cluster operators to update their configurations 4:21.
Key Takeaways:
• The release theme "Garhwal" symbolizes the collaborative spirit of the open-source community, reflecting the release lead's heritage and the interconnectedness required to survive challenging terrains 4:21.
• HPA Scale to Zero has graduated to beta, allowing workloads using external metrics to scale down to zero replicas, which is crucial for optimizing costs in event-driven or batch processing scenarios 14:31.
• Gang Scheduling and Workload-Aware Preemption have moved to beta, enabling Kubernetes to schedule modern AI and distributed workloads as cohesive groups rather than individual pods, preventing resource wastage during partial resource availability 16:18.
• Several important deprecations are in effect, including the IPVS mode in kube-proxy, the -f flag for kubectl run, and the kube-dns project, urging users to migrate to CoreDNS and newer networking modes 7:43.
• Static pods can no longer reference ConfigMaps or Secrets due to a security bug fix, and Cgroups V1 support is being phased out, requiring operators to audit their static manifests and node configurations 12:16.
• KML (Kubernetes Markup Language) has graduated to stable, offering a less ambiguous YAML subset for manifest definitions, while the metrics.k8s.io API also reaches stability for pod and node metrics 22:47.
Upgrading to Kubernetes 1.37 is essential for leveraging these new workload-aware scheduling capabilities and maintaining security compliance. Operators should carefully review the release notes and migration guides to address the deprecations and ensure a smooth transition to the latest supported version.
Sources:
- 4:21 Explanation of the "Garhwal" release theme and its significance to community collaboration.
- 14:31 Details on HPA Scale to Zero graduating to beta for cost efficiency.
- 16:18 Overview of Gang Scheduling and Workload-Aware Preemption features for AI workloads.
- 7:43 Discussion on deprecating IPVS mode,
kubectl run -f, and kube-dns. - 12:16 Warning about static pods losing ConfigMap/Secret access and Cgroups V1 removal.
- 22:47 Announcement of KML and
metrics.k8s.ioAPI graduating to stable.
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 0.5 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Cloud Security. Commonly maps to: Security Architecture and Engineering, Communication and Network Security. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
Hello and welcome to the Kubernetes podcast from Google. I'm your host, Kathleen Fields. >> [music] >> In this episode, we explore the Kubernetes 1.37 release with release lead Devesh Rawat. We talk about the release theme, deprecations and removals, and highlight a few of the most exciting new enhancements. But first, let's get to the news. >> [music] >> Kubeflow has officially reached graduated status following TOC approval. The project started at Google in 2017 primarily to run TensorFlow on Kubernetes. It entered the CNCF at the incubation level in 2023. Congratulations to the Kubeflow team. Buildpacks also graduated. The project was jointly launched by Pivotal and Heroku in 2018 and incubated by the CNCF the same year. Buildpacks automatically detects application language and compiles…