Features
- dra allocator implementation (#3082) #3082 (Jason Deal)
- add zone label to node/nodeclaim lifecycle metrics (#3096) #3096 (Derek Frank)
- Support CapacityBuffer API (#3028) #3028 (Sumukha Radhakrishna)
- consolidateAfter now applies to destination nodes during consolidation (#3006) #3006 (Derek Frank)
- dra consumable capacity + partitionable devices support (#3110) #3110 (Ryan Mistretta)
- add offering overrides and multi-allocatable scheduling support (#3109) #3109 (Jigisha Patil)
- dra allocator integration (#3113) #3113 (Jason Deal)
- implement Balanced consolidation policy (#2962) #2962 (James Thompson)
- account for instance type requirements for daemonset scheduling (#2975) #2975 (GaneshBannur)
- add NormalizedLabelValues for value translation during label normalization (#3104) #3104 (Andy Zhang)
- bump CapacityBuffer to v1beta1 and emit operator_* status metrics (#3129) #3129 (Derek Frank)
Bug Fixes
- use namespaced name as resourceclaim index (#3102) #3102 (Jason Deal)
- ResourceClaim hydration deadlock (#3122) #3122 (Ryan Mistretta)
- DRA counter tracking mutation race (#3116) #3116 (Ryan Mistretta)
- detect pod-level ResourceClaims in HasDRARequirements (#3124) #3124 (Andrew Mitchell)
- buffer computeReplicas must take the max of desired replicas and percentage (#3131) #3131 (Sumukha Radhakrishna)
- Ensure DaemonSets with '/initialized' and '/registered' are considered during nodeclaim calculations (#2161) #2161 (Wataru)
Documentation
- add Clever Cloud provider to README (#3099) #3099 (diodonfrost)
- add Hetzner to Karpenter implementations list (#3095) #3095 (Jannes Stubbemann)
- Support Capacity Buffers in Karpenter (#2898) #2898 (Sumukha Radhakrishna)
- add feature lifecycle, scope guidelines, RFC template, and updated contributing docs (#3049) #3049 (Derek Frank)
- InPlacePodVerticalScaling Support RFC (#3037) #3037 (Jigisha Patil)
Performance Improvements
- memory optimizations in cluster cost tracking (#3068) #3068 (Ryan Mistretta)
Tests
- replace metrics collection and calibrate resource utilization thresholds for perf tests (#3048) #3048 (Andrew Mitchell)
- add Capacity Buffers integ tests (#3044) #3044 (Sumukha Radhakrishna)
- set NodeClassRef on synthesized NodePool and migrate stale InstanceTypeOptions literals (#3117) #3117 (Andrew Mitchell)
Chores
- support consumable capacity and partitionable device fields in core DRA types (#3097) #3097 (Ryan Mistretta)
- add managed dimension to node and pod metrics (#3091) #3091 (Andrew Mitchell)
- make Buffer expectations provider/env agnostic (#3114) #3114 (Sumukha Radhakrishna)
- update go-opentelemetry dependencies (#3123) #3123 (Sumukha Radhakrishna)
- use status resources in pod resource calculation (#3065) #3065 (Jigisha Patil)
- make NodeClaim launch timeout configurable (#3112) #3112 (Andrew Mitchell)
- upgrade KWOK to 0.6.0 (#2970) #2970 (Alvin Bao)