What's Changed
Bug fixes
- [release-1.14] Fix Network-Topology-Aware scheduling soft mode for jobs and subjobs by @3th4novo in #5513
- [release-1.14] Fix scalar in-queue resource accounting to use milli-units by @WHOIM1205 in #5503
- [release-1.14] Clear device-related annotations when releasing pods by @archlitchi in #5486
- [release-1.14] Fix maxFloat and maxInt display in scheduler resource strings by @r0hansaxena in #5477
- [release-1.14] Fall back
minAvailableto replicas when partitionminPartitionsis omitted by @togettoyou in #5459 - [release-1.14] Check Ascend vNPU health using Allocatable resources by @DSFans2014 in #5437
- [release-1.14] Prevent unbounded growth of
job.Status.Conditionsby @avinxshKD in #5448 - [release-1.14] Fix HAMi vGPU scheduling failures in medium and large clusters by @Copilot in #5434
- [release-1.14] Restore
ImageStatesin scheduler node snapshots by @kitianFresh in #5356 - [release-1.14] Reprieve higher-priority pods first during preemption by @dengaosong in #5307
Full Changelog: v1.14.2...v1.14.3