October 10, 2024

LWKD: Week Ending October 6, 2024

Week: 2024-10-06

Developer News

The second pair of New Contributor Orientations will take place on October 15th. Please recommend them to the aspiring new contributor in your life.

Due to dependency changes in Debian LTS images, SIG-Release is proposing to stop building artifacts for some architectures. Discussions continue in SIG-Release.

Aldo Culquicondor has stepped down from WG Batch chair, and has nominated Kevin Hannon to take his place.

The NA Contributor Summit schedule is live. Remember to register, and if you're going consider volunteering to help run the Summit.

Release Schedule

Next Deadline: Enhancements Freeze, October 10

We're in the final week for Enhancement Opt-In. Please check your features on the project board and make sure their status is correct; if not, get in touch with your SIG lead ASAP.

Speaking of deadlines, patch release cherry picks for the October patches are due Friday.

KEP of the Week

KEP-3973: Consider Terminating Pods in Deployments

This KEP updates how the Deployment controller behaves when reporting Pod count when Pods are in terminating state. Currently Deployments have inconsistent behaviour in how the terminating Pods are handled based on the rollout strategy. This KEP introduces a new .spec.podReplacementPolicy with which users can configure this behaviour. This is similar to the spec.podReplacementPolicy in the Job spec and allows users to define if the Deployment should wait for its Pods to terminate before creating new Pods or if it should create them immediately.

This KEP is tracked for alpha in the ongoing v1.32 release.

Other Merges

• The NestedNumberAsFloat64 method converts int64 values to float64 when accessing numbers in unstructured objects
• Assert the flag behavior on kube-proxy using unit test
• API impact verification: Added a new -b flag for explicitly requesting build checks
• DRA Scheduler: mock workqueue for controller unit test
• Prevent a race condition from causing erroneous volume unmounts for flex volume plugins on kubelet restart
• Use generics for Workqueue metrics
• kubectl apply --server-side now supports --subresource
• In TestCoreResourceEnqueue, verify whether Pods are requeued to the activeQ after specific cluster events are triggered
• Fixed: TestPersistentVolumeProvisionMultiPVCs
• MVP cleanup #2
• New Tests: dra helper function, TestGenerateLinuxContainerResources, VolumeGroupSnapshot
• The scheduler adds QueueingHint in the VolumeBinding plugin's CSIDriver event
• system-reserved-compressible and kube-reserved-compressible to set only compressible (e.g. CPU) type resources on system and kube reserved slice respectively
• Reduce memory usage/allocations during wait for volume attachment

Version Updates

After increasing golang versions to 1.23.1 and 1.22.7, revert back to 1.23.0 until the question of alternate architectures is resolved.