Changes since v0.10.4:
Changes by Kind
Bug or Regression
- TAS: Add support for Node Selectors. (#5087, @mwysokin)
- Fix LocalQueue's status message to reference LocalQueue, rather than ClusterQueue, when its status is Ready (#4957, @PBundyra)
- Fix RBAC configuration for the Topology API to allow reading and editing by the service accounts using the Kueue Batch Admin role. (#4865, @KPostOffice)
- Fix a bug which caused Kueue's Scheduler to build invalid SSA patch in some scenarios when using
admission checks. This patch would be rejected with the following error message:
Workload.kueue.x-k8s.io "job-xxxxxx" is invalid: [admissionChecks[0].lastTransitionTime: Required value (#5088, @alexeldeib) - Fix bug which resulted in under-utilization of the resources in a Cohort.
Now, when a ClusterQueue is configured withpreemption.reclaimWithinCohort: Any,
its resources can be lent out more freely, as we are certain that we can reclaim
them later. Please see PR #4813 for detailed description of scenario. (#4823, @gabesaba) - Fixed bug that allow to create Topology without levels. (#5018, @mbobrovskyi)
- Helm: fix ServiceMonitor selecting the wrong service. This previously led to missing Kueue metrics, even with
enablePrometheusset totrue. (#5084, @j-vizcaino) - PodSetTopologyRequests are now configured only when TopologyAwareScheduling feature gate is enabled. (#4840, @mykysha)
- Revert making ClusterQueue.Preemption.BorrowWithinCohort as no-op when used with FairSharing, by reverting
#4165, which was done to prevent infinite-preemption loops as
described in #3779. However, it causes issues with upgrading for
users who use that combination. This configuration remains deprecated and can be removed without any notice
in future releases, so please consider alternatives. (#4819, @mimowo) - TAS: Fix bug where scheduling panics when the workload using TopologyAwareScheduling has container request value specified as zero. (#4976, @qti-haeyoon)