完善预排队资源限制与作业准入校验#398
Closed
YiD11 wants to merge 2 commits intoraids-lab:mainfrom
Closed
Conversation
Contributor
YiD11
commented
Apr 28, 2026
YiD11
commented
- 新增作业准入校验，提交前基于节点可用资源、亲和性、污点容忍等约束判断作业是否可放置。
- 调整队列资源限制模型
- 将预排队候选作业数改为全局系统配置
- 提交普通作业时区分“单次请求超限”和“当前使用量超限”：请求本身超限直接拒绝，使用量超限则进入预排队。
Contributor
There was a problem hiding this comment.
Pull request overview
本 PR 旨在完善 Prequeue/队列配额体系：将“预排队候选作业数”提升为全局系统配置、调整队列配额模型（简化为仅保留 quota）、并在作业提交前新增基于调度约束的准入校验，同时同步更新前端管理界面与多语言文案。
Changes:
- 后端新增作业准入校验模块，并在作业提交流程中接入；同时将 prequeueCandidateSize 迁移为全局 Prequeue 配置项。
- 队列配额（queue quota）模型/接口收敛：移除 per-queue 的 enabled 与 prequeueCandidateSize 字段，迁移数据库字段并更新 swagger。
- 前端管理端同步：在账户（default）中配置/展示“队内上限（queueLimit）”，调整平台设置页的 Prequeue 配置项，更新表格分页与多语言。
Reviewed changes
Copilot reviewed 35 out of 35 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| frontend/src/utils/quota.ts | Quota 表单结构新增 queueLimit，并提供表单与 queue-quota 的转换函数 |
| frontend/src/services/api/system-config.ts | Prequeue 配置接口新增 prequeueCandidateSize 字段 |
| frontend/src/services/api/queue-quota.ts | QueueQuota 接口类型移除 enabled / prequeueCandidateSize 等字段 |
| frontend/src/routes/admin/more/index.tsx | 平台设置页移除旧“用户资源限制”管理，改为维护全局 prequeueCandidateSize |
| frontend/src/routes/admin/more/-components/user-resource-limit.tsx | 删除旧的用户资源限制 UI 组件 |
| frontend/src/routes/admin/more/-components/prequeue-settings.tsx | Prequeue 配置表单新增候选作业数输入项 |
| frontend/src/routes/admin/accounts/-components/account-table.tsx | 账户表格新增“队内上限”列并读取 default queue quota 展示 |
| frontend/src/routes/admin/accounts/-components/account-form-component.tsx | default 账户支持编辑/保存 default queue quota（queueLimit）并做 hydration/invalidate |
| frontend/src/i18n/locales/zhCN/translation.json | 新增/调整队内上限、Prequeue 候选作业数、弹性(backfill)相关文案及 toast 文案 |
| frontend/src/i18n/locales/enUS/translation.json | 同步新增/调整对应英文文案 |
| frontend/src/i18n/locales/ja/translation.json | 同步新增/调整对应日文文案 |
| frontend/src/i18n/locales/ko/translation.json | 同步新增/调整对应韩文文案 |
| frontend/src/components/query-table/index.tsx | 表格排序/过滤/全局搜索变更时重置页码，并关闭 autoResetPageIndex |
| backend/pkg/vcjob/admission/admission.go | 新增作业准入校验：基于节点约束与资源进行放置模拟 |
| backend/pkg/utils/resource.go | 资源工具增强：复制、覆盖判断、缺口计算、汇总展示、按副本累加等 |
| backend/pkg/reconciler/vcjob-reconciler.go | 事件/终止状态采集触发条件改为 isReleasedJobPhase |
| backend/pkg/prequeuewatcher/scheduling.go | 移除本地 node 约束匹配逻辑，复用 vcjob admission 的约束判断；可用资源计算保留 |
| backend/pkg/prequeuewatcher/resources.go | 复用 utils.ResourceListCovers，移除重复实现 |
| backend/pkg/prequeuewatcher/preemption.go | 复用 admission 的约束判断与 utils.ResourceDeficit；细化过滤条件 |
| backend/pkg/prequeuewatcher/activation.go | 激活批次使用全局 prequeueCandidateSize 限制候选扫描与激活数量 |
| backend/internal/service/vcjob/runtime.go | 资源统计改用 utils.CalculateReplicatedResources 复用逻辑 |
| backend/internal/service/prequeue_service.go | 队列配额结构/解析逻辑调整；新增 CheckRequestedResourceLimit 以区分“单次请求超限” |
| backend/internal/service/config_service.go | PrequeueRuntimeConfig 新增 prequeue_candidate_size 配置项并纳入校验/序列化 |
| backend/internal/handler/system_config.go | Prequeue 配置接口新增字段并强制 required,gt=0 校验 |
| backend/internal/handler/queue_quota.go | QueueQuota API 请求/响应移除 enabled 与 prequeueCandidateSize 字段 |
| backend/internal/handler/vcjob/tensorflow.go | 资源统计改用 utils.CalculateReplicatedResources，移除 aitaskctl 依赖 |
| backend/internal/handler/vcjob/pytorch.go | 同上：资源统计改用 utils.CalculateReplicatedResources |
| backend/internal/handler/vcjob/lifecycle.go | 提交流程接入 admission 校验与队列配额“请求超限/使用超限”区分逻辑 |
| backend/docs/swagger.yaml | Swagger 定义同步：新增 prequeueCandidateSize、调整 queue quota 结构 |
| backend/docs/swagger.json | Swagger JSON 同步更新 |
| backend/docs/docs.go | swag 生成模板同步更新 |
| backend/dao/query/queue_quotas.gen.go | gorm-gen 查询字段同步移除 enabled / prequeue_candidate_size |
| backend/dao/model/queue_quota.go | QueueQuotaLimit 模型移除 enabled / prequeueCandidateSize 字段 |
| backend/dao/model/prequeue_config.go | Prequeue 配置新增 prequeue_candidate_size key 与默认值常量 |
| backend/cmd/gorm-gen/models/migrate.go | 新增迁移：删除 queue_quotas 表的 enabled 与 prequeue_candidate_size 列 |
Comment on lines
+663
to
+667
| type="string" | ||
| placeholder={t('accountForm.queueLimitPlaceholder')} | ||
| className="font-mono" | ||
| {...form.register(`resources.${index}.queueLimit`, { | ||
| valueAsNumber: true, |
There was a problem hiding this comment.
【核心规范】该输入框使用了无效的 HTML input 类型 type="string"，但同时又启用了 valueAsNumber。浏览器会把未知类型当作 text 处理，用户输入非数字或清空时会被转成 NaN，最终 convertFormToQueueQuota 会忽略该字段，导致“队内上限”无法稳定保存。建议改为 type="number"（并设置 min/step），同时对空值用 setValueAs/onChange 显式转成 undefined。
Comment on lines
+46
to
+49
| remaining := make([]v1.ResourceList, len(nodeList.Items)) | ||
| for i := range nodeList.Items { | ||
| remaining[i] = utils.CopyResources(nodeList.Items[i].Status.Allocatable) | ||
| } |
There was a problem hiding this comment.
【核心规范】这里用 node.Status.Allocatable 作为“剩余可用资源”进行准入判断，但 allocatable 是节点可分配总量，不包含当前已被 Pod 请求占用的资源；因此可能会误判“可准入”，提交后仍长期 Pending，和 PR 描述的“基于节点可用资源”不一致。建议改为基于 allocatable - 已分配（可参考 prequeuewatcher 的按节点列 Pod 再 SumRequests 的实现）计算 available 后再做放置模拟。
Comment on lines
67
to
+71
| scheduleTypeInt, err := strconv.ParseInt( | ||
| job.Annotations[vcjobservice.AnnotationKeyScheduleType], 10, 64, | ||
| ) | ||
| if err != nil { | ||
| return fmt.Errorf("invalid schedule type annotation value: %w", err) | ||
| return 0, fmt.Errorf("invalid schedule type annotation value: %w", err) |
There was a problem hiding this comment.
【核心规范】调度类型解析这里直接 strconv.ParseInt 读取注解，注解缺失（空字符串）会直接报错；同时也不兼容历史上可能存在的字符串值（例如 backfill），而代码库里已有 vcjobservice.ParseJobScheduleMetadata / model.ParseScheduleType 处理默认值与兼容格式。建议复用已有解析函数，避免因注解格式/缺失导致作业提交被拒绝。
| score float64 | ||
| } | ||
|
|
||
| func CheckJobAdmission(ctx context.Context, k8sClient client.Client, job *batch.Job) (*Result, error) { |
There was a problem hiding this comment.
【优化建议】新增的作业准入放置模拟逻辑（资源计算 + 亲和/污点匹配 + best-fit 选择）属于关键调度路径，建议补充单测覆盖典型场景（nodeSelector/required nodeAffinity、NoSchedule taint + toleration、多 task/多副本、资源不足的拒绝原因）。可以参考 backend/pkg/prequeuewatcher/watcher_test.go 使用 controller-runtime fake client 构造 Node/Pod/Job。
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.