github.com/kubewharf/katalyst-core

What's Changed

• test: fix flaky unit tests by @tituschewxj in #1205
• feat(cpu): add metrics for cpu weight and system exclusive pool by @junyu-peng in #1213
• fix(kcc): use JSON-based checksum for checkpoint compatibility by @luomingmeng in #1206
• fix(cpuburst): skip updating burst when cpu quota is max by @JustinChengLZ in #1217
• feat(sysadvisor): fix-mbw-overhead-calculation by @syc4704413 in #1192
• chore: add CodeRabbit configuration by @tituschewxj in #1220

New Contributors

• @tituschewxj made their first contribution in #1205

Full Changelog: v0.5.50...v0.5.51

What's Changed

• feat(tmo): support reservedInactiveFile by @lubinszARM in #1135
• feat: implement rdma associated device allocation logic by @JustinChengLZ in #1083
• feat(lifecycle): add label keep list for CNR lifecycle controller by @luomingmeng in #1183
• fix(gpu): short-circuit when device request count is 0 by @JustinChengLZ in #1194
• fix(gpu reporter): handle devices with no NUMA nodes by @JustinChengLZ in #1196
• feat(gpu-qrm): add failure metric for get associated device topology hints by @JustinChengLZ in #1197

Full Changelog: v0.5.49...v0.5.50

What's Changed

• feat(qrm): introduce resource package core foundation by @luomingmeng in #1152
• feat: qrm cpu numa rp optimization by @luomingmeng in #1177
• feat: qrm system and topology enhancements by @luomingmeng in #1179
• refactor: qrm allocation hooks by @luomingmeng in #1178
• fix bugs by @luomingmeng in #1181
• feat: set vf readonly state by @junyu-peng in #1176
• feat(gpu qrm): add more metrics for gpu plugins by @JustinChengLZ in #1182
• fix(gpu): clear inactive pods from state and add duration metrics by @JustinChengLZ in #1185
• feat(eviction): add eviction annotations for explicit trigger by @JustinChengLZ in #1157
• stateful spd baseline based on shard by @ddjjia in #1028
• refactor cpu usage eviction by @luomingmeng in #1187
• feat(resctrl): add manager for resctrl filesystem operations by @luomingmeng in #1088
• feat(gpu_memory): support milligpu allocation logic by @JustinChengLZ in #1131
• refactor(node metric reporter): adapt ResourceMetric to ResourceList by @zj666-lgtm in #1186
• feat: support system exclusive pool by @junyu-peng in #1162
• feat: support cpu weight by @junyu-peng in #1174

New Contributors

• @zj666-lgtm made their first contribution in #1186

Full Changelog: v0.5.48...v0.5.49

What's Changed

• fix(gpu-reporter): bypass cache when getting pod metadata by @luomingmeng in #1143
• fix: correct 17 klog format-string bugs across the codebase by @pierluigilenoci in #1134
• refactor(irq-tuning): handle known failure of get GetNicQueue2Irq by @zhanghaoyu1986 in #1122
• feat(gpu): ignore machine state of pod if generator doesnt exist by @JustinChengLZ in #1153
• feat: implement hugepages resource allocation by @JustinChengLZ in #1084
• fix: adjust hints to avoid affinity error if not restricted by @junyu-peng in #1156
• feat(device affinity): generate priority dimensions from calculating device dimensions by @JustinChengLZ in #1146
• feat(sysadvisor): add disk metrics by @syc4704413 in #1154
• refactor(memory-qrm): refactor huge page adjustment for host watermark and test by @luomingmeng in #1161
• feat(eviction): add system CPU pressure eviction plugin by @luomingmeng in #1141
• feat(qrm): add cpu threshold check for shared_cores with numabinding GetTopologyHint by @liunxaa in #1155
• fix(qrm): qrm-mb-plugin exits start if resctrl FS not mounted yet by @h-w-chen in #1164
• chore: watch events of cgroup file CREATE and REMOVE by @luomingmeng in #1168
• metrics(qrm): mbm throttle metrics by @h-w-chen in #1159
• fix(sysadvisor): fix mbm calculation logic by @syc4704413 in #1171
• fix(gpu): fix reporter to trigger periodically and filter out inactive pods by @JustinChengLZ in #1173

New Contributors

• @pierluigilenoci made their first contribution in #1134
• @liunxaa made their first contribution in #1155

Full Changelog: v0.5.47...v0.5.48

What's Changed

• fix(gpu): add retry mechanism for report building failures by @luomingmeng in #1142

Full Changelog: v0.5.46...v0.5.47

What's Changed

• feat(qrm): qrm-mb-plugin: recalibrate ccd cap by @h-w-chen in #1138

Full Changelog: v0.5.45...v0.5.46

What's Changed

• fix: allow non-consecutive affinity priority levels by @JustinChengLZ in #1098
• feat(*): add switch to support disabling some reporters by @fantastic-hf in #1114
• feat(*): add switch to support disabling cnr lifecycle by @fantastic-hf in #1115
• feat(qrm): support set memory.high to throttle reclaimed_cores by @cheney-lin in #1125
• feat(gpu): strict device affinity requirement and refactor by @JustinChengLZ in #1121
• enhance(qrm): mem bandwidth throttler qrm plugin supports multiple resctrl groups sharing same priority by @h-w-chen in #1111
• feat(fetcher): add checkpoint support for inner reporter plugins by @luomingmeng in #1116
• feat(sriov): skip dynamic VF allocation by pod annotations by @junyu-peng in #1112
• feat(gpu): start device topology registry in base plugin by @luomingmeng in #1108
• fix(vpa-status): skip inactive pods when checking recommendation applied by @luomingmeng in #1129
• feat: support setting of cpu burst only for main container by @JustinChengLZ in #1123
• Revise memory headroom calculation to reduce OOM by @jinxin32 in #1128
• skip verify when Upload coverage to Codecov by @xu282934741 in #1137
• test case fix: race among test cases by @h-w-chen in #1139
• fix(hintoptimizer): handle checkpoint not found error gracefully by @luomingmeng in #1140
• feat(gpu reporter): support reading of kubelet state checkpoint by @JustinChengLZ in #1133

New Contributors

• @fantastic-hf made their first contribution in #1114

Full Changelog: v0.5.44...v0.5.45

What's Changed

• fix(gpu/staticpolicy): handle ensureState failures gracefully by @luomingmeng in #1109
• fix: implement interface methods in the unsupportedManagers of cgv1 and cgv2 by @JustinChengLZ in #1110
• refactor gpu reporter by @luomingmeng in #1119
• feat: enable reporting of allocations for gpu by @JustinChengLZ in #1118

Full Changelog: v0.5.43...v0.5.44

What's Changed

• fix(sysadvisor): fix zen4 mb data adjustment by @xdc0527 in #1054
• feat: implement canonical fallback if there is no device affinity by @JustinChengLZ in #1093
• fix(irq): fix the issue of forbidden core not working by @JulyWindK in #1085
• fix: custom metric emitter by @JustinChengLZ in #1087
• refactor(resource-package): enhance resource package handling with config and attributes by @luomingmeng in #1092
• fix(qosaware): remove overload check logic to avoid excessive eviction by @jinxin32 in #1081
• fix(sysadvisor): SPD tmo indicator overrides PodEnableReclaim config by @Zera-Algorithm in #1079
• Add a new PoolNameConfig field for tmo, to configure a specific tmo reclaim policy for each pool name by @cheney-lin in #1102
• feat(*): remove sysadvisor metacache persistence by @JulyWindK in #1061
• feat(resource-pool): support numa level resource pool by @Bourbon-Wang in #1077
• refactor(device): simplify device topology interface by removing numa ready flag by @luomingmeng in #1090
• feat(qrm-plugin): support reclaimed cores vpa by @junyu-peng in #1078
• feat: report borwein strategy upper and lower bound metrics by @jinxin32 in #1096
• feat(*): adjust watermark for hugepages by @JulyWindK in #1101
• fix(evictionmanager): add CNR taint health check registration by @luomingmeng in #1103
• refactor gpu topology by @luomingmeng in #1106
• feat: support sriov vf by @junyu-peng in #1064
• mbm - memory bandwidth qrm plugin by @h-w-chen in #886
• feat: enable default cpu burst for shared cores pods by @JustinChengLZ in #1080
• feat(cpu): new cpu enhancements in dynamic policy by @JustinChengLZ in #1063
• feat(qrm, sysadvisor): Periodically (every 300s) clean up excessive (>2000) dying mem cgroups by @cheney-lin in #1107
• fix(gpu/staticpolicy): handle ensureState failures gracefully by @luomingmeng in #1109
• fix: implement interface methods in the unsupportedManagers of cgv1 and cgv2 by @JustinChengLZ in #1110
• refactor gpu reporter by @luomingmeng in #1119
• feat: enable reporting of allocations for gpu by @JustinChengLZ in #1118

New Contributors

• @xdc0527 made their first contribution in #1054
• @Zera-Algorithm made their first contribution in #1079

Full Changelog: v0.5.42...v0.5.44

What's Changed

• fix: bug when storing cpu burst value in file when it is 0 by @JustinChengLZ in #1056
• fix(vpa): skip qrm hinter check when pod request is scaling down by @Bourbon-Wang in #1062
• feat(borwein): Add Borwein V3 model support by @jinxin32 in #1066
• fix(cpu): reject inplace update from non-snb to snb allocation by @luomingmeng in #1071
• fix(network): handle zero resource requests in NIC selection by @luomingmeng in #1057
• refactor(cpu): rename sharedCoresNUMABindingResultAnnotationKey to NUMABindingResultAnnotationKey by @luomingmeng in #1060
• refactor(cpu): improve cpu usage calculation and quota adjustment by @luomingmeng in #1070
• feat(gpu): implement gpu plugins by @JustinChengLZ in #1008
• feat(memory): support vm.watermark_scale_factor by @lubinszARM in #1075
• feat(memory): support SetMemTHP by @lubinszARM in #1073
• Revert "fix(cpu-advisor): skip validation for non-numa_exclusive allocations" by @luomingmeng in #1089

New Contributors

• @jinxin32 made their first contribution in #1066

Full Changelog: v0.5.40...v0.5.42