LWKD: Week Ending July 6, 2025
Week Ending July 6, 2025
Developer News
SIG-Architecture group proposes to form a new Working Group focused on AI Conformance Certification. The WG would define a standardized set of capabilities, APIs, and configurations that Kubernetes clusters must support to reliably and efficiently run AI/ML workloads.
Kubernetes has formed a dedicated Checkpoint/Restore Working Group to integrate native Checkpoint/Restore functionality, enabling container migration and workload pre-emption to improve resource efficiency and support advanced use cases like AI/ML.
Release Schedule
Next Deadline: Code and Test Freeze, July 24/25
Code and Test Freeze starts at 0200 UTC on Friday, July 25. Your PRs should all be merged by then. Vyom Yadav has shared mid-cycle status, including 72 tracked changes. Because this means an extra-long Release Blog, the Comms Team requests that leads submit their release highlights early, if you can.
Cherry-picks for the July Patch Releases are due on July 11.
Featured PRs
131641: DRA kubelet: add dra_resource_claims_in_use gauge vector
This PR introduces a new gauge vector metric, dra_resource_claims_in_use
, to the Kubelet; This metric tracks active DRA drivers and informs administrators when a driver is in use, ensuring safe removal of drivers without impacting pod operations; This metric is useful to determine if drivers have active ResourceClaims, preventing issues during the driver removal process.
KEP of the Week
This KEP adds support for distributed tracing in the kubelet to help diagnose node-level issues like pod creation latency or container startup delays.
It solves the problem of limited visibility into how the kubelet talks to the API server and container runtime by exporting trace data. The implementation uses OpenTelemetry
to generate and export spans in the OTLP format. An OpenTelemetry Collector
, typically deployed as a DaemonSet
, receives and forwards this data to a tracing backend.The feature is enabled through the KubeletTracing
feature gate and configured using the TracingConfiguration in the kubelet configuration file.
This KEP is tracked as stable in v1.34.
Other Merges
- logger.Error replaced with utilruntime.HandleErrorWithXXX where errors cannot be returned
- Fix for validation error when specifying resource requirements at the container level for a resource not supported at the pod level
- Declarative Validation enabled for CertificateSigningRequest
- Names of new Services are validated with NameIsDNSLabel() relaxing pre-existing validation when RelaxedServiceNameValidation feature gate is enabled
- allocationManager's IsPodResizeInProgress method unexported
- New dra_resource_claims_in_use kubelet metrics to inform about active ResourceClaims
- Statefulset now respects minReadySeconds
- CSIDriverRegistry cleaned up
- Function to translate named port to port number cleaned up to avoid duplication
- Unit tests for VolumePathHandler
Deprecated
- In a major refactoring effort, replaced the deprecated package 'k8s.io/utils/pointer' with 'k8s.io/utils/ptr' across multiple components
- Deprecated gogo protocol definitions removed from k8s.io/externaljwt and k8s.io/cri-api
Subprojects and Dependency Updates
- cluster-api v1.11.0-alpha.2: releases alpha version for testing
- cluster-api-provider-vspherev1.14.0-alpha.2 : releases alpha version for testing
- kustomize [shlex] (https//github.com/google/shlex) has been replaced with carapace-shlex, Bump to viper v1.20.0 and Drop usage of forked copies of goyaml.v2 and goyaml.v3
Shoutouts
- No shoutouts this week. Want to thank someone awesome in the community? Tag them in the #shoutouts channel.