The Kubernetes container orchestration has reached version 1.28. Under the name Planternetes – because the work on it can be compared to a garden – the second release of the year stabilizes twelve existing features and transfers 14 to beta and 19 to alpha status.
One of the alpha features concerns the pod replacement policy for jobs. A new field in the job API now lets you specify whether the control plane should continue to create new pods as before as soon as the previous pods start scheduling, or instead apply the new, optional behavior until fully scheduled to maintain the pods. According to the Kubernetes team, the previous behavior could lead to difficulties with machine learning frameworks such as TensorFlow or JAX, since these require unique pods per index.
Stable features for node shutdowns and storage class
Version 1.28 allows cleaning up after a node unexpectedly exits and restarting stateful workloads on another node as a stable feature. This also applies when a node is in a non-recoverable state. Reasons for this and for the abrupt termination of a node can be hardware errors or a faulty operating system.
The automatic, retrospective assignment of a standard storage class is now also stable. With a StorageClass, admins can describe the “classes” of storage they offer, for example in terms of quality of service levels or backup policies. Kubernetes automatically sets a storageClassName for a PersistentVolumeClaim (PVC) if developers don’t specify a value. Also, the control plane specifies a StorageClass for an existing PVC with undefined storageClassName. In version 1.28, this pre-existing feature has transitioned to stable status and is automatic and active by default.
Other stable features in Kubernetes 1.28 include support for third-party device monitoring plugins.
Open source available for download
Kubernetes 1.28 is open source for download on GitHub. The Kubernetes team encourages interested parties to actively participate in the development process – for example through one of the Special Interest Groups (SIGs), which specialize in topics such as architecture, autoscaling, docs, scalability, testing or user interface (UI).
All details about the new release can be found on the Kubernetes blog and the release notes on GitHub.
Go to home page
#Container #orchestration #Kubernetes #stabilizes #handling #node #shutdowns