|
| 1 | +## v0.16.1 |
| 2 | + |
| 3 | +Changes since `v0.16.0`: |
| 4 | + |
| 5 | +## Changes by Kind |
| 6 | + |
| 7 | +### Feature |
| 8 | + |
| 9 | +- KueueViz backend and frontend resource requests/limits are now configurable via Helm values (kueueViz.backend.resources and kueueViz.frontend.resources). (#8981, @david-gang) |
| 10 | + |
| 11 | +### Bug or Regression |
| 12 | + |
| 13 | +- Fix Visibility API OpenAPI schema generation to prevent schema resolution errors when visibility v1beta1/v1beta2 APIServices are installed. |
| 14 | + |
| 15 | + The visibility schema issues result in the following error when re-applying the manifest for Kueue 0.16.0: |
| 16 | + `failed to load open api schema while syncing cluster cache: error getting openapi resources: SchemaError(sigs.k8s.io/kueue/apis/visibility/v1beta1.PendingWorkloadsSummary.items): unknown model in reference: "sigs.k8s.io~1kueue~1apis~1visibility~1v1beta1.PendingWorkload"` (#8901, @vladikkuzn) |
| 17 | +- Fix a bug where finished or deactivated workloads blocked ClusterQueue deletion and finalizer removal. (#8936, @sohankunkerkar) |
| 18 | +- LeaderWorkerSet: Fix the bug where rolling updates with maxSurge could get stuck. (#8886, @PannagaRao) |
| 19 | +- LeaderWorkerSet: Fixed bug that doesn't allow to delete Pod after LeaderWorkerSet delete (#8882, @mbobrovskyi) |
| 20 | +- Metrics certificate is now reloaded when certificate data is updated. (#9099, @MaysaMacedo) |
| 21 | +- MultiKueue & ElasticJobs: fix the bug that the new size of a Job was not reflected on the worker cluster. (#9055, @ichekrygin) |
| 22 | +- Observability: Fix Prometheus ServiceMonitor selector and RBAC to enable metrics scraping. (#8980, @IrvingMg) |
| 23 | +- Observability: Fixed a bug where workloads that finished before a Kueue restart were not tracked in the gauge metrics for finished workloads. (#8827, @mbobrovskyi) |
| 24 | +- Observability: fix the bug that the "replica-role" (leader / follower) log decorator was missing in the log lines output by |
| 25 | + the webhooks for LeaderWorkerSet and StatefulSet . (#8820, @mszadkow) |
| 26 | +- PodIntegration: Fix the bug that Kueue would occasionally remove the custom finalizers when |
| 27 | + removing the `kueue.x-k8s.io/managed` finalizer. (#8903, @mykysha) |
| 28 | +- RayJob integration: Make RayJob top level workload managed by Kueue when autoscaling via |
| 29 | + ElasticJobsViaWorkloadSlices is enabled. |
| 30 | + |
| 31 | + If you are an alpha user of the ElasticJobsViaWorkloadSlices feature for RayJobs, then upgrading Kueue may impact running live jobs which have autoscaling / workload slicing enabled. For example, if you upgrade Kueue, before |
| 32 | + scaling-up completes, the new pods will be stuck in SchedulingGated state. (#9039, @hiboyang) |
| 33 | +- TAS: Fix a bug that TAS ignored resources excluded by excludeResourcePrefixes for node placement. (#8990, @sohankunkerkar) |
| 34 | +- TAS: Fixed a bug that pending workloads could be stuck, not being considered by the Kueue's scheduler, |
| 35 | + after the restart of Kueue. The workloads would be considered for scheduling again after any update to their |
| 36 | + ClusterQueue. (#9056, @sohankunkerkar) |
| 37 | + |
| 38 | +### Other (Cleanup or Flake) |
| 39 | + |
| 40 | +- KueueViz: It switches to the v1beta2 API (#8804, @mbobrovskyi) |
| 41 | + |
1 | 42 | ## v0.16.0 |
2 | 43 |
|
3 | 44 | Changes since `v0.15.0`: |
|
0 commit comments