feat(worker): 实现集群限流与自适应负载
保留平台模型 RPM、TPM 和并发策略语义，增加 PostgreSQL 集群级租约、饱和候选重选和多平台自动负载，避免突发任务固定等待首个平台。\n\n新增 Worker 实时负载采样、自适应 active/heavy 容量、心跳与管理端指标，并扩展本地 acceptance runner，覆盖三 Worker、同模型三平台 2/4/6 并发和 48 个带图视频突发任务。\n\n验证：go test ./...、go vet ./...、PostgreSQL 跨 Store 集成测试、gofmt、bash -n、ShellCheck 及本地集群 provider-burst 验收通过；48/48 成功，无越限、重复提交、重复计费、重复回调或终态资源泄漏。
This commit is contained in:
@@ -2575,6 +2575,30 @@
|
||||
}
|
||||
}
|
||||
},
|
||||
"/api/admin/runtime/workers": {
|
||||
"get": {
|
||||
"security": [
|
||||
{
|
||||
"BearerAuth": []
|
||||
}
|
||||
],
|
||||
"produces": [
|
||||
"application/json"
|
||||
],
|
||||
"tags": [
|
||||
"acceptance"
|
||||
],
|
||||
"summary": "获取集群 Worker 实时负载与领取容量",
|
||||
"responses": {
|
||||
"200": {
|
||||
"description": "OK",
|
||||
"schema": {
|
||||
"$ref": "#/definitions/store.WorkerClusterRuntime"
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
"/api/admin/system/acceptance/capacity-profiles": {
|
||||
"get": {
|
||||
"security": [
|
||||
@@ -16347,6 +16371,108 @@
|
||||
"$ref": "#/definitions/store.GatewayWalletAccount"
|
||||
}
|
||||
}
|
||||
},
|
||||
"store.WorkerClusterRuntime": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"capturedAt": {
|
||||
"type": "string"
|
||||
},
|
||||
"queue": {
|
||||
"$ref": "#/definitions/store.WorkerQueueRuntime"
|
||||
},
|
||||
"workers": {
|
||||
"type": "array",
|
||||
"items": {
|
||||
"$ref": "#/definitions/store.WorkerInstanceRuntime"
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
"store.WorkerInstanceRuntime": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"activeLeases": {
|
||||
"type": "integer"
|
||||
},
|
||||
"allocatedCapacity": {
|
||||
"type": "integer"
|
||||
},
|
||||
"capacityLimit": {
|
||||
"type": "integer"
|
||||
},
|
||||
"drainingAt": {
|
||||
"type": "string"
|
||||
},
|
||||
"finalizingTasks": {
|
||||
"type": "integer"
|
||||
},
|
||||
"hardCapacityLimit": {
|
||||
"type": "integer"
|
||||
},
|
||||
"heartbeatAt": {
|
||||
"type": "string"
|
||||
},
|
||||
"heavyCapacity": {
|
||||
"type": "integer"
|
||||
},
|
||||
"instanceId": {
|
||||
"type": "string"
|
||||
},
|
||||
"loadSampledAt": {
|
||||
"type": "string"
|
||||
},
|
||||
"podName": {
|
||||
"type": "string"
|
||||
},
|
||||
"podUid": {
|
||||
"type": "string"
|
||||
},
|
||||
"preparingTasks": {
|
||||
"type": "integer"
|
||||
},
|
||||
"pressureReason": {
|
||||
"type": "string"
|
||||
},
|
||||
"pressureState": {
|
||||
"type": "string"
|
||||
},
|
||||
"reportedActiveTasks": {
|
||||
"type": "integer"
|
||||
},
|
||||
"revision": {
|
||||
"type": "string"
|
||||
},
|
||||
"runningTasks": {
|
||||
"type": "integer"
|
||||
},
|
||||
"safeCapacity": {
|
||||
"type": "integer"
|
||||
},
|
||||
"site": {
|
||||
"type": "string"
|
||||
},
|
||||
"status": {
|
||||
"type": "string"
|
||||
},
|
||||
"waitingUpstreamTasks": {
|
||||
"type": "integer"
|
||||
}
|
||||
}
|
||||
},
|
||||
"store.WorkerQueueRuntime": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"oldestWaitSeconds": {
|
||||
"type": "number"
|
||||
},
|
||||
"queued": {
|
||||
"type": "integer"
|
||||
},
|
||||
"running": {
|
||||
"type": "integer"
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
"securityDefinitions": {
|
||||
|
||||
@@ -4196,6 +4196,73 @@ definitions:
|
||||
primaryAccount:
|
||||
$ref: '#/definitions/store.GatewayWalletAccount'
|
||||
type: object
|
||||
store.WorkerClusterRuntime:
|
||||
properties:
|
||||
capturedAt:
|
||||
type: string
|
||||
queue:
|
||||
$ref: '#/definitions/store.WorkerQueueRuntime'
|
||||
workers:
|
||||
items:
|
||||
$ref: '#/definitions/store.WorkerInstanceRuntime'
|
||||
type: array
|
||||
type: object
|
||||
store.WorkerInstanceRuntime:
|
||||
properties:
|
||||
activeLeases:
|
||||
type: integer
|
||||
allocatedCapacity:
|
||||
type: integer
|
||||
capacityLimit:
|
||||
type: integer
|
||||
drainingAt:
|
||||
type: string
|
||||
finalizingTasks:
|
||||
type: integer
|
||||
hardCapacityLimit:
|
||||
type: integer
|
||||
heartbeatAt:
|
||||
type: string
|
||||
heavyCapacity:
|
||||
type: integer
|
||||
instanceId:
|
||||
type: string
|
||||
loadSampledAt:
|
||||
type: string
|
||||
podName:
|
||||
type: string
|
||||
podUid:
|
||||
type: string
|
||||
preparingTasks:
|
||||
type: integer
|
||||
pressureReason:
|
||||
type: string
|
||||
pressureState:
|
||||
type: string
|
||||
reportedActiveTasks:
|
||||
type: integer
|
||||
revision:
|
||||
type: string
|
||||
runningTasks:
|
||||
type: integer
|
||||
safeCapacity:
|
||||
type: integer
|
||||
site:
|
||||
type: string
|
||||
status:
|
||||
type: string
|
||||
waitingUpstreamTasks:
|
||||
type: integer
|
||||
type: object
|
||||
store.WorkerQueueRuntime:
|
||||
properties:
|
||||
oldestWaitSeconds:
|
||||
type: number
|
||||
queued:
|
||||
type: integer
|
||||
running:
|
||||
type: integer
|
||||
type: object
|
||||
info:
|
||||
contact: {}
|
||||
description: |-
|
||||
@@ -5846,6 +5913,20 @@ paths:
|
||||
summary: 更新 Runner 策略
|
||||
tags:
|
||||
- runtime
|
||||
/api/admin/runtime/workers:
|
||||
get:
|
||||
produces:
|
||||
- application/json
|
||||
responses:
|
||||
"200":
|
||||
description: OK
|
||||
schema:
|
||||
$ref: '#/definitions/store.WorkerClusterRuntime'
|
||||
security:
|
||||
- BearerAuth: []
|
||||
summary: 获取集群 Worker 实时负载与领取容量
|
||||
tags:
|
||||
- acceptance
|
||||
/api/admin/system/acceptance/capacity-profiles:
|
||||
get:
|
||||
produces:
|
||||
|
||||
Reference in New Issue
Block a user