Kubernetes KEPs Design Tracking
这页是 KEP 设计方案追踪表，服务于后续长期更新。源摘要和设计脉络见 src-kubernetes-keps-design-tracking；按 feature 合并后的覆盖矩阵见 kubernetes-keps-feature-coverage；逐个 KEP 的 Alpha/Beta/GA、是否实现、feature gates 和关键实现路径见 kubernetes-keps-implementation-matrix。
使用方式
- 先按 SIG 找到 KEP。
- 再看
分类判断它属于哪条设计线。 优先级用于决定后续是否需要单独深挖：P0：近期活跃、跨 SIG 影响大、或与 AI/HPC/GPU/平台工程强相关。P1：稳定基础设计，适合作为背景知识或项目对照。P2：历史、边缘或低频补充。
重要设计文档详解
| 设计文档 | 覆盖 KEP | 重点 |
|---|---|---|
| kubernetes-keps-feature-coverage | all P0/P1 groups | 回答“整理了哪些 feature”，按合并设计组列出 scheduling / autoscaling / node 的覆盖状态。 |
| kubernetes-keps-implementation-matrix | all P0/P1 KEPs | 逐个 KEP 追踪 status、stage、Alpha/Beta/GA milestones、feature gate、是否落地和关键实现路径。 |
| kubernetes-scheduler-core-design | 624, 785, 1451, 3521, 4247, 6132, 5598, 895, 4832 |
Scheduler framework、profiles、queue/requeue、topology placement、async preemption 和调度性能底座。 |
| kubernetes-workload-gang-scheduling-design | 4671, 5710, 6012, 6089, 5732 |
Workload / PodGroup API、gang scheduling、workload-aware preemption、topology-aware workload scheduling 和 controller API building blocks。 |
| kubernetes-dra-design-deep-dive | 4381, 3063, 5007, 5075, 4815, 4816, 5055 |
DRA structured parameters、ResourceSlice/ResourceClaim、scheduler plugin、kubelet plugin、CDI、autoscaler 可推理性和设备扩展 KEP。 |
| kubernetes-hpa-autoscaling-design | 4951, 2021, 5679, 1610, 5325, 5030 |
HPA tolerance、container resource metrics、pod selection accuracy、scale from zero、external metric fallback 和 attach-limit autoscaler。 |
| kubernetes-in-place-pod-resize-design | 1287, 5419, 5526, 5554, 6122, 5836 |
Pod /resize subresource、desired/allocated/actuated/actual 状态、Pod-level resources、static CPU manager、scheduler preemption for deferred resize。 |
| kubernetes-node-runtime-observability-security-design | 2040, 2221, 585, 3570, 1769, 693, 753, 127, 2033, 2371, 5394, 4680 |
kubelet/CRI、resource managers、sidecar lifecycle、user namespace/rootless kubelet、CRI stats、PSI 和 resource health。 |
SIG Scheduling
P0 活跃设计线
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| workload/gang | 4671-gang-scheduling |
implementable / beta / v1.37 | 用 Workload/PodGroup 表达整组 Pod 调度，解决 minCount、组装、整体可行性、preemption、binding 和失败回退。 | P0 |
| workload/gang | 5710-workload-aware-preemption |
implementable / beta / v1.37 | 抢占从单 Pod 扩展到 workload 级，避免高优先级 gang 抢占后仍无法整体运行。 | P0 |
| workload/gang | 6012-composite-podgroup-api |
implementable / alpha / v1.37 | 表达组合 PodGroup，适合多组件 workload。 | P0 |
| workload/gang | 6089-was-controller-apis |
implementable / alpha / v1.37 | Workload Aware Scheduling controller API，给 controller 暴露 workload-aware 调度状态。 | P0 |
| queue | 6132-prequeueing-hints |
implementable / beta / v1.37 | 在 Pod 入 activeQ 前通过 hint 判断事件是否足以唤醒调度。 | P0 |
| queue | 5598-opportunistic-batching |
implementable / beta / v1.37 | 批量化调度机会，降低重复计算，提高吞吐。 | P0 |
| DRA scheduling | 5075-dra-consumable-capacity |
implementable / beta / v1.37 | DRA 支持可消费容量，不再只处理离散设备实例。 | P0 |
| DRA scheduling | 5517-dra-node-allocatable-resources |
implementable / alpha / v1.37 | 暴露 DRA node allocatable，支撑调度和容量推理。 | P0 |
| DRA scheduling | 5729-resourceclaim-support-for-workloads |
implementable / beta / v1.37 | ResourceClaim 支持 workload 级使用，和 PodGroup/gang 交叉。 | P0 |
| DRA scheduling | 5941-dra-shared-consumable-capacity |
implementable / alpha / v1.37 | 多设备共享同一 consumable capacity。 | P0 |
| DRA scheduling | 5963-device-compatibility-groups |
implementable / alpha / v1.37 | 设备兼容组，表达多设备组合约束。 | P0 |
| DRA scheduling | 6080-dra-derived-attributes |
provisional / alpha / v1.37 | 从设备属性派生拓扑/匹配属性，解决 driver 属性表达不一致。 | P0 |
| resize/preemption | 5836-scheduler-preemption-for-ippr |
implementable / alpha / v1.37 | scheduler 处理 in-place pod resize 引发的资源冲突和抢占。 | P0 |
P1 基础设计线
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| framework | 624-scheduling-framework |
implemented / stable / v1.19 | scheduler plugin 扩展点体系，是后续所有调度 KEP 的底座。 | P1 |
| config | 785-scheduler-component-config-api |
implemented / stable / v1.25 | versioned scheduler ComponentConfig。 | P1 |
| config | 1451-multi-scheduling-profiles |
implementable / beta / v1.19 | 一个 scheduler 实例支持多个 profile。 | P1 |
| queue | 3521-pod-scheduling-readiness |
implemented / stable / v1.30 | scheduling gates 控制 Pod 是否进入调度。 | P1 |
| queue | 4247-queueinghint |
implemented / stable / v1.34 | plugin-specific QueueingHint，减少无效 requeue。 | P1 |
| queue | 5142-pop-backoffq-when-activeq-empty |
implementable / beta / v1.33 | activeQ 空时提前处理 backoffQ。 | P1 |
| queue | 5501-reflect-preenqueue-rejections-in-pod-status |
implementable / beta / v1.35 | PreEnqueue 拒绝原因进入 Pod status。 | P1 |
| topology | 895-pod-topology-spread |
implemented / stable / v1.19 | Pod 拓扑分布基础设计。 | P1 |
| topology | 1258-default-pod-topology-spread |
implementable / stable / v1.24 | 默认拓扑分布策略。 | P1 |
| topology | 3022-min-domains-in-pod-topology-spread |
implemented / stable / v1.30 | PodTopologySpread 最小 domain。 | P1 |
| topology | 3094-pod-topology-spread-considering-taints |
implemented / stable / v1.33 | skew 计算考虑 taints/tolerations。 | P1 |
| topology | 3243-respect-pod-topology-spread-after-rolling-upgrades |
implementable / beta / v1.34 | rolling upgrade 后继续尊重 spread。 | P1 |
| affinity | 2249-pod-affinity-namespace-selector |
implementable / stable / v1.24 | Pod affinity 支持 namespace selector。 | P1 |
| affinity | 3633-matchlabelkeys-to-podaffinity |
implemented / stable / v1.33 | affinity/anti-affinity 支持 matchLabelKeys。 | P1 |
| scoring | 2458-node-resource-score-strategy |
implementable / beta / v1.22 | 节点资源打分策略可配置。 | P1 |
| preemption | 4832-async-preemption |
implementable / beta / v1.33 | 异步化 preemption。 | P1 |
| preemption | 5278-nominated-node-name-for-expectation |
implementable / beta / v1.35 | 改善 nominated node 期望表达。 | P1 |
| DRA scheduling | 5007-device-attach-before-pod-scheduled |
implementable / beta / v1.36 | 设备绑定条件进入调度路径。 | P1 |
| DRA scheduling | 5004-dra-extended-resource |
implementable / stable / v1.37 | Extended resources 通过 DRA driver 处理。 | P1 |
| DRA scheduling | 5055-dra-device-taints-and-tolerations |
implementable / stable / v1.37 | 设备级 taints/tolerations。 | P1 |
| DRA scheduling | 4815-dra-partitionable-devices |
implementable / beta / v1.36 | 可分区设备。 | P1 |
| DRA scheduling | 4816-dra-prioritized-list |
implementable / stable / v1.36 | 设备请求优先级备选列表。 | P1 |
| DRA scheduling | 5234-dra-resourceslice-mixins |
implementable / alpha / v1.34 | ResourceSlice mixins。 | P1 |
| DRA scheduling | 5491-dra-list-types-for-attributes |
implementable / alpha / v1.37 | DRA 属性支持 list 类型。 | P1 |
| placement | 5732-topology-aware-workload-scheduling |
implementable / beta / v1.37 | workload 级拓扑感知调度。 | P1 |
| scheduler/API | 5229-asynchronous-api-calls-during-scheduling |
implementable / beta / v1.34 | 调度期间 API 调用异步化。 | P1 |
P2 历史和补充
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| extender | 1819-scheduler-extender |
implemented / stable / v1.19 | HTTP extender 历史扩展路径。 | P2 |
| priority | 268-priority-preemption |
implementable / alpha | Priority/preemption 基础线。 | P2 |
| priority | 902-non-preempting-priorityclass |
implemented / stable / v1.24 | 高优先级但不抢占。 | P2 |
| preemption | 3280-guarantee-pdb-when-preemption-happens |
implementable / alpha / v1.27 | 抢占时保护 PDB。 | P2 |
| job | 2926-job-mutable-scheduling-directives |
implemented / stable / v1.27 | Job scheduling directives 可变。 | P2 |
| pod | 3838-pod-mutable-scheduling-directives |
implementable / stable / v1.30 | Pod scheduling directives 可变。 | P2 |
| taint | 382-taint-node-by-condition |
implemented / stable | Node condition taint。 | P2 |
| taint | 3902-decoupled-taint-manager |
implemented / stable / v1.34 | TaintManager 从 NodeLifecycleController 解耦。 | P2 |
| daemonset | 548-schedule-daemonset-pods |
implemented / stable | DaemonSet pods 交给 scheduler。 | P2 |
| quota | 986-resource-quota-scope-selectors |
implemented / stable | ResourceQuota scope selectors。 | P2 |
| quota | 2372-node-labels-quota |
provisional | 基于 node labels 的 ResourceQuota。 | P2 |
| score | 964-binpacking-priority |
implementable / alpha | RequestedToCapacityRatio 支持 extended resource binpacking。 | P2 |
| historical | 583-coscheduling |
provisional | gang scheduling 前身。 | P2 |
| historical | 5832-decouple-podgroup-api |
implementable / alpha / v1.36 | PodGroup API 解耦。 | P2 |
| config | 2891-simplified-config |
implementable / beta / v1.23 | scheduler config 简化。 | P2 |
| placement | 5471-enable-sla-based-scheduling |
implementable / alpha / v1.35 | threshold-based placement。 | P2 |
| DRA scheduling | 5027-dra-admin-controlled-device-attributes |
provisional / alpha / v1.33 | 管理员控制设备属性。 | P2 |
SIG Autoscaling
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| HPA behavior | 4951-configurable-hpa-tolerance |
implementable / stable / v1.37 | per-HPA / per-direction tolerance，避免全局 10% 过粗。 | P0 |
| HPA behavior | 853-configurable-hpa-scale-velocity |
implemented / stable | HPA behavior policies、stabilization window、scale up/down velocity。 | P1 |
| HPA API | 2702-graduate-hpa-api-to-GA |
implemented / stable / v1.23 | HPA v2 API GA。 | P1 |
| metrics | 1610-container-resource-autoscaling |
implemented / stable / v1.30 | container-level resource metrics，解决 sidecar 稀释主容器指标。 | P0 |
| metrics | 117-hpa-metrics-specificity |
implemented / stable | custom/external metrics label selector。 | P1 |
| metrics | 5325-hpa-pod-selection-accuracy |
implementable / v1.35 | 更准确的 HPA Pod selection，避免读到非 target Pod 指标。 | P0 |
| scale from zero | 2021-scale-from-zero |
implementable / beta / v1.37 | HPA 基于 object/external metrics 从 0 副本启动。 | P0 |
| failure policy | 5679-external-metric-fallback |
implementable / alpha / v1.36 | 外部指标失败时 fallback，降低 metrics adapter 故障影响。 | P0 |
| cluster autoscaler | 5030-attach-limit-autoscaler |
implementable / beta / v1.37 | CSI volume attach limit 进入 Cluster Autoscaler 推理。 | P0 |
SIG Node
P0 活跃设计线
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| DRA/device | 4381-dra-structured-parameters |
implemented / stable / v1.35 | DRA structured parameters，取代 opaque allocation 路线，是 DRA 主线。 | P0 |
| DRA/device | 3063-dynamic-resource-allocation |
withdrawn / alpha / v1.32 | 早期 opaque DRA controller 协商方案，因 autoscaler 不可推理和 apiserver 往返复杂而撤回。 | P0 |
| DRA/device | 4817-resource-claim-device-status |
implementable / stable / v1.37 | ResourceClaim status 标准化设备信息。 | P0 |
| DRA/device | 5304-dra-attributes-downward-api |
implementable / beta / v1.37 | DRA device attributes 暴露到 Pod。 | P0 |
| DRA/device | 5677-dra-resource-availability-visibility |
implementable / alpha / v1.37 | DRA resource availability 可见性。 | P0 |
| DRA/device | 5945-dra-optional-node-preparation |
implementable / alpha / v1.37 | DRA NodePrepare 可选化。 | P0 |
| DRA/device | 6072-dra-standard-numanode |
implementable / stable / v1.37 | DRA 标准 numaNode 属性。 |
P0 |
| resize/resources | 1287-in-place-update-pod-resources |
implemented / stable / v1.35 | Pod resource requests/limits 原地更新。 | P0 |
| resize/resources | 5419-pod-level-resources-in-place-resize |
implementable / beta / v1.36 | Pod-level resources 的 in-place resize。 | P0 |
| resize/resources | 5526-pod-level-resource-managers |
implementable / beta / v1.37 | Pod-level resource managers。 | P0 |
| resize/resources | 5554-in-place-update-pod-resources-alongside-static-cpu-manager-policy |
implementable / alpha / v1.37 | in-place resize 与 static CPU Manager policy 协作。 | P0 |
| resize/resources | 6122-configurable-scaling-delay-with-pod-resource-exposure |
implementable / alpha / v1.37 | Pod resource exposure 后的 scaling delay。 | P0 |
| observability | 2371-cri-pod-container-stats |
implementable / beta / v1.37 | CRI-full container/pod stats，减少 cAdvisor 依赖。 | P0 |
| observability | 5394-psi-node-conditions |
implementable / alpha / v1.36 | PSI 信号进入 Node Conditions。 | P0 |
| observability | 4680-add-resource-health-to-pod-status |
implementable / beta / v1.36 | Device Plugin/DRA resource health 进入 Pod status。 | P0 |
| security | 2033-kubelet-in-userns-aka-rootless |
implementable / beta / v1.37 | rootless kubelet。 | P0 |
| security | 5607-hostnetwork-userns |
implementable / alpha / v1.36 | HostNetwork Pod 使用 user namespaces。 | P0 |
| lifecycle | 4438-container-restart-termination |
implementable / alpha / v1.37 | Pod 终止期间 sidecar restart 行为。 | P0 |
| lifecycle | 4563-eviction-request-api |
implementable / alpha / v1.37 | EvictionRequest API。 | P0 |
P1 基础设计线
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| resource manager | 3570-cpumanager |
implemented / stable / v1.26 | CPU Manager cpuset 分配。 | P1 |
| resource manager | 1769-memory-manager |
implemented / stable / v1.32 | Memory Manager 和 Topology Manager hint 协作。 | P1 |
| resource manager | 693-topology-manager |
implemented / stable / v1.27 | CPU/Memory/Device NUMA hint 聚合。 | P1 |
| resource manager | 3545-improved-multi-numa-alignment |
implemented / stable / v1.32 | 多 NUMA 对齐改进。 | P1 |
| resource manager | 2837-pod-level-resource-spec |
implementable / stable / v1.37 | Pod-level resource spec。 | P1 |
| resource manager | 2570-memory-qos |
implementable / beta / v1.37 | Memory QoS with cgroup v2。 | P1 |
| resource manager | 2902-cpumanager-distribute-cpus-policy-option |
implementable / beta / v1.33 | CPUManager 分散 CPU policy。 | P1 |
| resource manager | 4540-strict-cpu-reservation |
implemented / stable / v1.35 | strict CPU reservation。 | P1 |
| device | 3573-device-plugin |
implemented / stable / v1.26 | Device Plugin API。 | P1 |
| device | 4009-add-cdi-devices-to-device-plugin-api |
implemented / stable / v1.31 | Device Plugin 返回 CDI devices。 | P1 |
| device | 3695-pod-resources-for-dra |
implemented / stable / v1.36 | PodResources API 包含 DRA resources。 | P1 |
| runtime | 2040-kubelet-cri |
implementable / beta / v1.23 | kubelet CRI 支撑。 | P1 |
| runtime | 2221-remove-dockershim |
implemented / stable / v1.24 | 移除 dockershim。 | P1 |
| runtime | 585-runtime-class |
implemented / stable / v1.20 | RuntimeClass。 | P1 |
| runtime | 4033-group-driver-detection-over-cri |
implemented / stable / v1.37 | CRI 发现 cgroup driver。 | P1 |
| runtime | 5825-cri-pagination |
implementable / alpha / v1.36 | CRI list streaming/pagination。 | P1 |
| lifecycle | 753-sidecar-containers |
implemented / stable / v1.33 | Sidecar containers。 | P1 |
| lifecycle | 277-ephemeral-containers |
implemented / stable / v1.25 | Ephemeral containers。 | P1 |
| lifecycle | 2000-graceful-node-shutdown |
implementable / beta / v1.21 | Graceful node shutdown。 | P1 |
| lifecycle | 2712-pod-priority-based-graceful-node-shutdown |
implementable / beta / v1.24 | Priority-based graceful shutdown。 | P1 |
| lifecycle | 5307-container-restart-policy |
implementable / beta / v1.35 | Container restart policy。 | P1 |
| security | 127-user-namespaces |
implemented / stable / v1.36 | User namespaces。 | P1 |
| security | 2413-seccomp-by-default |
implementable / stable / v1.27 | Seccomp by default。 | P1 |
| security | 2862-fine-grained-kubelet-authz |
implementable / stable / v1.36 | Kubelet API 细粒度授权。 | P1 |
| security | 2254-cgroup-v2 |
implementable / stable / v1.25 | cgroup v2。 | P1 |
| security | 5573-remove-cgroup-v1 |
implementable / beta / v1.35 | 移除 cgroup v1。 | P1 |
| observability | 727-resource-metrics-endpoint |
implemented / stable / v1.29 | Kubelet resource metrics endpoint。 | P1 |
| observability | 4205-psi-metric |
implemented / stable / v1.36 | PSI metrics。 | P1 |
| observability | 5067-pod-generation |
implemented / stable / v1.35 | Pod generation。 | P1 |
| observability | 5328-node-declared-features |
implementable / stable / v1.37 | Node declared features。 | P1 |
P2 补充清单
| 分类 | KEP | 状态 / 阶段 | 设计方案 | 优先级 |
|---|---|---|---|---|
| storage/resource | 1029-ephemeral-storage-quotas |
implementable / beta | Ephemeral storage quota。 | P2 |
| storage/resource | 1539-hugepages |
implemented / stable | HugePages。 | P2 |
| storage/resource | 1967-size-memory-backed-volumes |
implemented / stable | memory-backed volume size。 | P2 |
| storage/resource | 6030-dynamic-resize-of-memory-backed-volumes |
implementable / alpha | memory-backed volume 动态 resize。 | P2 |
| runtime | 1547-building-kubelet-without-docker |
implemented / stable | Dockerless kubelet。 | P2 |
| runtime | 4216-image-pull-per-runtime-class |
implementable / alpha | RuntimeClass 维度镜像拉取。 | P2 |
| runtime | 4191-split-image-filesystem |
implementable / beta | split image filesystem。 | P2 |
| runtime | 4210-max-image-gc-age |
implemented / stable | image GC max age。 | P2 |
| runtime | 5365-ImageVolume-with-image-digest |
implementable / alpha | ImageVolume with digest。 | P2 |
| probe/log | 2727-grpc-probe |
implemented / stable | gRPC probe。 | P2 |
| probe/log | 4939-grpc-probe-with-tls |
implementable / alpha | gRPC probe with TLS。 | P2 |
| probe/log | 5999-h2c-container-probes |
implementable / alpha | h2c probes。 | P2 |
| probe/log | 2411-cri-container-log-rotation |
implemented / stable | CRI log rotation。 | P2 |
| probe/log | 3288-separate-stdout-from-stderr |
implementable / alpha | stdout/stderr split stream。 | P2 |
| lifecycle | 3960-pod-lifecycle-sleep-action |
implemented / stable | PreStop sleep action。 | P2 |
| lifecycle | 4818-allow-zero-value-for-sleep-action-of-prestop-hook |
implemented / stable | PreStop sleep zero value。 | P2 |
| lifecycle | 4960-container-stop-signals |
implementable / beta | Container stop signals。 | P2 |
| lifecycle | 5532-restart-all-containers-on-container-exits |
implementable / beta | Restart all containers on exits。 | P2 |
| lifecycle | 5593-configure-the-max-crashloopbackoff-delay |
implementable / beta | CrashLoopBackOff max delay。 | P2 |
| security | 135-seccomp |
implemented / stable | Seccomp GA。 | P2 |
| security | 24-apparmor |
implementable / stable | AppArmor。 | P2 |
| security | 1898-hardened-exec |
implementable / alpha | Harden exec against SSRF。 | P2 |
| security | 213-run-as-group |
implemented / stable | RunAsGroup。 | P2 |
| security | 3619-supplemental-groups-policy |
implemented / stable | SupplementalGroups policy。 | P2 |
| security | 4265-proc-mount |
implementable / stable | ProcMount option。 | P2 |
| security | 5474-enable-writable-cgroups |
implementable / alpha | Writable cgroups。 | P2 |
| security | 5855-noexec-emptyDir |
implementable / alpha | emptyDir mount options。 | P2 |
| node status | 589-efficient-node-heartbeats |
implemented / stable | Efficient node heartbeat。 | P2 |
| node status | 3085-pod-conditions-for-starting-completition-of-sandbox-creation |
implementable / stable | Pod networking ready condition。 | P2 |
| node status | 5683-lifecycle-conditions |
implementable / alpha | Node lifecycle conditions。 | P2 |
| node feature | 793-node-os-arch-labels |
implementable / alpha | OS/arch labels GA。 | P2 |
| node feature | 4742-node-topology-downward-api |
implementable / beta | Node topology labels via Downward API。 | P2 |
| node config | 281-dynamic-kubelet-configuration |
removed | Dynamic kubelet config，已移除。 | P2 |
| node config | 3983-drop-in-configuration |
implemented / stable | kubelet drop-in config dir。 | P2 |
| node config | 4580-deprecate-kubelet-runonce |
provisional / alpha | Deprecate kubelet runonce。 | P2 |
| misc | 2400-node-swap |
implemented / stable | Node swap。 | P2 |
| misc | 2535-ensure-secret-pulled-images |
implementable / beta | Ensure secret pulled images。 | P2 |
| misc | 4639-oci-volume-source |
implemented / stable | OCI images as VolumeSource。 | P2 |
| misc | 5758-per-container-ulimits-configuration |
implementable / alpha | Per-container ulimits。 | P2 |
| misc | 5823-pod-level-checkpoint-restore |
implementable / alpha | Pod-level checkpoint/restore。 | P2 |
| misc | 6035-exec-session-identity |
implementable / alpha | Exec session identity propagation。 | P2 |
| misc | 6063-pod-pid-limit |
implementable / alpha | Per-Pod PID limit。 | P2 |
下一轮要补的数据
- 从
prod-readiness/sig-scheduling/*.yaml、prod-readiness/sig-autoscaling/*.yaml、prod-readiness/sig-node/*.yaml补 feature gates、metrics、rollback、scalability、upgrade/downgrade。 - 增加 “版本视图”：按
latest-milestone输出 v1.35、v1.36、v1.37 的活跃 KEP 列表。 - 增加 “跨项目视图”：把 KEP 和 kueue、karpenter、scheduler-plugins、metrics-server、prometheus-adapter、node-feature-discovery、k8s-gpu-device-stack 对齐。
- 从 P2 中挑选是否需要新增专题：quota/capacity governance、probe/logging、storage/resource local capacity。