生命周期管理架构¶
本文档详细介绍 MCPStore 生命周期管理的内部架构、组件设计和工作原理。
🏗️ 整体架构图¶
graph TB
subgraph "用户接口层"
UserAPI[用户API]
Context[MCPStoreContext]
end
subgraph "生命周期管理层"
LifecycleManager[ServiceLifecycleManager<br/>生命周期管理器]
StateMachine[ServiceStateMachine<br/>状态机]
HealthManager[HealthManager<br/>健康管理器]
ReconnectionManager[SmartReconnectionManager<br/>智能重连管理器]
end
subgraph "状态处理器"
InitProcessor[InitializingStateProcessor<br/>初始化处理器]
EventProcessor[StateChangeEventProcessor<br/>事件处理器]
ContentManager[ContentManager<br/>内容管理器]
end
subgraph "监控系统"
HealthCheck[健康检查<br/>30秒间隔]
ToolsUpdate[工具更新<br/>2小时间隔]
StateMonitor[状态监控<br/>实时]
PerformanceTracker[性能跟踪器]
end
subgraph "数据存储"
Registry[ServiceRegistry<br/>状态存储]
Metadata[ServiceStateMetadata<br/>元数据]
Config[LifecycleConfig<br/>配置]
HealthHistory[健康历史记录]
end
subgraph "外部接口"
Orchestrator[MCPOrchestrator<br/>编排器]
FastMCP[FastMCP Client<br/>MCP客户端]
Services[MCP Services<br/>外部服务]
end
%% 核心流程
UserAPI --> Context
Context --> LifecycleManager
LifecycleManager --> StateMachine
LifecycleManager --> HealthManager
LifecycleManager --> ReconnectionManager
StateMachine --> InitProcessor
StateMachine --> EventProcessor
HealthManager --> HealthCheck
HealthCheck --> StateMonitor
ToolsUpdate --> ContentManager
StateMonitor --> PerformanceTracker
%% 数据流
LifecycleManager --> Registry
Registry --> Metadata
Config --> StateMachine
HealthManager --> HealthHistory
%% 外部交互
LifecycleManager --> Orchestrator
Orchestrator --> FastMCP
FastMCP --> Services
%% 反馈循环
Services -.->|健康状态| HealthCheck
StateMonitor -.->|状态变化| StateMachine
ReconnectionManager -.->|重连触发| Orchestrator
PerformanceTracker -.->|性能数据| Registry
%% 样式
classDef user fill:#e3f2fd
classDef lifecycle fill:#f3e5f5
classDef processor fill:#e8f5e8
classDef monitor fill:#fff3e0
classDef storage fill:#fce4ec
classDef external fill:#f1f8e9
class UserAPI,Context user
class LifecycleManager,StateMachine,HealthManager,ReconnectionManager lifecycle
class InitProcessor,EventProcessor,ContentManager processor
class HealthCheck,ToolsUpdate,StateMonitor,PerformanceTracker monitor
class Registry,Metadata,Config,HealthHistory storage
class Orchestrator,FastMCP,Services external🔄 7状态生命周期状态机¶
stateDiagram-v2
[*] --> INITIALIZING : 服务注册
INITIALIZING --> HEALTHY : 连接成功<br/>工具获取完成
INITIALIZING --> RECONNECTING : 初始化失败<br/>连接超时
HEALTHY --> WARNING : 偶发失败<br/>响应变慢
HEALTHY --> RECONNECTING : 连续失败<br/>达到重连阈值
HEALTHY --> DISCONNECTING : 手动停止<br/>用户操作
WARNING --> HEALTHY : 恢复正常<br/>响应时间改善
WARNING --> RECONNECTING : 持续失败<br/>达到重连阈值
RECONNECTING --> HEALTHY : 重连成功<br/>服务恢复
RECONNECTING --> UNREACHABLE : 重连失败<br/>超过最大重试次数
UNREACHABLE --> RECONNECTING : 重试重连<br/>定期尝试
UNREACHABLE --> DISCONNECTED : 放弃重连<br/>手动停止
DISCONNECTING --> DISCONNECTED : 断开完成<br/>资源清理
DISCONNECTED --> [*] : 服务删除<br/>完全移除
DISCONNECTED --> INITIALIZING : 服务重启<br/>重新注册
note right of INITIALIZING
• 配置验证完成
• 执行首次连接
• 获取工具列表
• 设置初始状态
end note
note right of HEALTHY
• 连接正常稳定
• 心跳检查成功
• 工具调用可用
• 响应时间正常
end note
note right of WARNING
• 偶发心跳失败
• 响应时间变慢
• 未达到重连阈值
• 仍可提供服务
end note
note right of RECONNECTING
• 连续失败达到阈值
• 正在执行重连
• 服务暂时不可用
• 自动恢复中
end note
note right of UNREACHABLE
• 重连失败
• 进入长周期重试
• 服务完全不可用
• 需要人工干预
end note
note right of DISCONNECTING
• 执行优雅关闭
• 清理连接资源
• 保存状态信息
• 准备完全停止
end note
note right of DISCONNECTED
• 服务完全终止
• 等待手动删除
• 或准备重新启动
• 状态信息保留
end note🧩 核心组件架构¶
ServiceLifecycleManager¶
graph TB
subgraph "ServiceLifecycleManager"
Core[核心管理器]
TaskScheduler[任务调度器]
EventDispatcher[事件分发器]
end
subgraph "管理的组件"
StateMachine[状态机]
HealthManager[健康管理器]
ReconnectionManager[重连管理器]
ContentManager[内容管理器]
end
subgraph "定时任务"
HealthCheckTask[健康检查任务<br/>30秒间隔]
ToolsUpdateTask[工具更新任务<br/>2小时间隔]
CleanupTask[清理任务<br/>24小时间隔]
ReconnectionTask[重连任务<br/>动态间隔]
end
subgraph "事件处理"
StateChangeEvent[状态变化事件]
HealthChangeEvent[健康变化事件]
ReconnectionEvent[重连事件]
ToolsUpdateEvent[工具更新事件]
end
Core --> TaskScheduler
Core --> EventDispatcher
TaskScheduler --> HealthCheckTask
TaskScheduler --> ToolsUpdateTask
TaskScheduler --> CleanupTask
TaskScheduler --> ReconnectionTask
EventDispatcher --> StateChangeEvent
EventDispatcher --> HealthChangeEvent
EventDispatcher --> ReconnectionEvent
EventDispatcher --> ToolsUpdateEvent
Core --> StateMachine
Core --> HealthManager
Core --> ReconnectionManager
Core --> ContentManager
HealthCheckTask --> HealthManager
ToolsUpdateTask --> ContentManager
ReconnectionTask --> ReconnectionManager
StateMachine --> StateChangeEvent
HealthManager --> HealthChangeEvent
ReconnectionManager --> ReconnectionEvent
ContentManager --> ToolsUpdateEventServiceStateMachine¶
graph TB
subgraph "状态转换引擎"
TransitionEngine[转换引擎]
RuleValidator[规则验证器]
ThresholdManager[阈值管理器]
end
subgraph "转换规则"
SuccessRules[成功转换规则]
FailureRules[失败转换规则]
TimeoutRules[超时转换规则]
ManualRules[手动转换规则]
end
subgraph "状态处理器"
InitializingHandler[初始化处理器]
HealthyHandler[健康处理器]
WarningHandler[警告处理器]
ReconnectingHandler[重连处理器]
UnreachableHandler[不可达处理器]
DisconnectingHandler[断开处理器]
DisconnectedHandler[已断开处理器]
end
TransitionEngine --> RuleValidator
TransitionEngine --> ThresholdManager
RuleValidator --> SuccessRules
RuleValidator --> FailureRules
RuleValidator --> TimeoutRules
RuleValidator --> ManualRules
TransitionEngine --> InitializingHandler
TransitionEngine --> HealthyHandler
TransitionEngine --> WarningHandler
TransitionEngine --> ReconnectingHandler
TransitionEngine --> UnreachableHandler
TransitionEngine --> DisconnectingHandler
TransitionEngine --> DisconnectedHandler
ThresholdManager -.-> WarningHandler
ThresholdManager -.-> ReconnectingHandler
ThresholdManager -.-> UnreachableHandlerHealthManager¶
graph TB
subgraph "健康检查引擎"
CheckEngine[检查引擎]
StatusEvaluator[状态评估器]
TimeoutManager[超时管理器]
end
subgraph "检查策略"
PingCheck[Ping检查]
ToolsCheck[工具检查]
ResponseTimeCheck[响应时间检查]
AvailabilityCheck[可用性检查]
end
subgraph "健康等级"
HealthyLevel[HEALTHY<br/>< 1秒, >95%]
WarningLevel[WARNING<br/>1-3秒, 90-95%]
SlowLevel[SLOW<br/>3-10秒, 80-90%]
UnhealthyLevel[UNHEALTHY<br/>>10秒, <80%]
end
subgraph "数据收集"
ResponseTracker[响应跟踪器]
FailureTracker[失败跟踪器]
PerformanceTracker[性能跟踪器]
HistoryTracker[历史跟踪器]
end
CheckEngine --> StatusEvaluator
CheckEngine --> TimeoutManager
CheckEngine --> PingCheck
CheckEngine --> ToolsCheck
CheckEngine --> ResponseTimeCheck
CheckEngine --> AvailabilityCheck
StatusEvaluator --> HealthyLevel
StatusEvaluator --> WarningLevel
StatusEvaluator --> SlowLevel
StatusEvaluator --> UnhealthyLevel
CheckEngine --> ResponseTracker
CheckEngine --> FailureTracker
CheckEngine --> PerformanceTracker
CheckEngine --> HistoryTracker
ResponseTracker --> StatusEvaluator
FailureTracker --> StatusEvaluator
PerformanceTracker --> StatusEvaluatorSmartReconnectionManager¶
graph TB
subgraph "重连策略引擎"
StrategyEngine[策略引擎]
BackoffCalculator[退避计算器]
PriorityManager[优先级管理器]
end
subgraph "重连队列"
CriticalQueue[关键服务队列<br/>0.5x延迟]
HighQueue[高优先级队列<br/>0.7x延迟]
NormalQueue[普通队列<br/>1.0x延迟]
LowQueue[低优先级队列<br/>1.5x延迟]
end
subgraph "重连策略"
ExponentialBackoff[指数退避<br/>60s → 600s]
MaxAttempts[最大尝试次数<br/>10次]
CircuitBreaker[熔断器<br/>失败保护]
HealthyReset[健康重置<br/>成功后清零]
end
subgraph "监控指标"
AttemptCounter[尝试计数器]
SuccessRate[成功率统计]
AverageDelay[平均延迟]
QueueLength[队列长度]
end
StrategyEngine --> BackoffCalculator
StrategyEngine --> PriorityManager
PriorityManager --> CriticalQueue
PriorityManager --> HighQueue
PriorityManager --> NormalQueue
PriorityManager --> LowQueue
BackoffCalculator --> ExponentialBackoff
BackoffCalculator --> MaxAttempts
BackoffCalculator --> CircuitBreaker
BackoffCalculator --> HealthyReset
StrategyEngine --> AttemptCounter
StrategyEngine --> SuccessRate
StrategyEngine --> AverageDelay
StrategyEngine --> QueueLength📊 数据流架构¶
健康检查数据流¶
sequenceDiagram
participant Timer as 定时器
participant HealthManager as 健康管理器
participant FastMCP as FastMCP客户端
participant Service as MCP服务
participant StateMachine as 状态机
participant Registry as 注册表
Timer->>HealthManager: 触发健康检查
HealthManager->>FastMCP: ping_service()
FastMCP->>Service: MCP Ping
Service-->>FastMCP: Pong + 响应时间
FastMCP-->>HealthManager: 检查结果
HealthManager->>HealthManager: 评估健康状态
alt 状态需要变化
HealthManager->>StateMachine: 触发状态转换
StateMachine->>Registry: 更新服务状态
StateMachine->>HealthManager: 状态变化确认
end
HealthManager->>Registry: 更新健康元数据
Registry-->>HealthManager: 更新完成重连流程数据流¶
sequenceDiagram
participant StateMachine as 状态机
participant ReconnectionManager as 重连管理器
participant Orchestrator as 编排器
participant FastMCP as FastMCP客户端
participant Service as MCP服务
participant Registry as 注册表
StateMachine->>ReconnectionManager: 添加重连任务
ReconnectionManager->>ReconnectionManager: 计算重连延迟
loop 重连循环
ReconnectionManager->>Orchestrator: 触发重连
Orchestrator->>FastMCP: 断开旧连接
Orchestrator->>FastMCP: 创建新连接
FastMCP->>Service: 建立连接
alt 连接成功
Service-->>FastMCP: 连接确认
FastMCP->>Service: 获取工具列表
Service-->>FastMCP: 工具列表
FastMCP-->>Orchestrator: 重连成功
Orchestrator-->>ReconnectionManager: 成功通知
ReconnectionManager->>StateMachine: 触发成功转换
StateMachine->>Registry: 更新为HEALTHY
else 连接失败
Service-->>FastMCP: 连接失败
FastMCP-->>Orchestrator: 重连失败
Orchestrator-->>ReconnectionManager: 失败通知
ReconnectionManager->>ReconnectionManager: 增加失败计数
ReconnectionManager->>ReconnectionManager: 计算下次重连时间
end
end🔧 配置架构¶
生命周期配置层次¶
graph TB
subgraph "全局配置"
GlobalConfig[全局生命周期配置]
DefaultThresholds[默认阈值配置]
DefaultTimeouts[默认超时配置]
end
subgraph "服务类型配置"
CriticalConfig[关键服务配置]
NormalConfig[普通服务配置]
BackgroundConfig[后台服务配置]
end
subgraph "运行时配置"
DynamicThresholds[动态阈值调整]
AdaptiveTimeouts[自适应超时]
LoadBasedConfig[负载基础配置]
end
subgraph "服务特定配置"
ServiceOverrides[服务特定覆盖]
CustomHealthChecks[自定义健康检查]
SpecialHandling[特殊处理规则]
end
GlobalConfig --> DefaultThresholds
GlobalConfig --> DefaultTimeouts
GlobalConfig --> CriticalConfig
GlobalConfig --> NormalConfig
GlobalConfig --> BackgroundConfig
CriticalConfig --> DynamicThresholds
NormalConfig --> AdaptiveTimeouts
BackgroundConfig --> LoadBasedConfig
DynamicThresholds --> ServiceOverrides
AdaptiveTimeouts --> CustomHealthChecks
LoadBasedConfig --> SpecialHandling📈 性能优化架构¶
并发处理架构¶
graph TB
subgraph "任务调度器"
MainScheduler[主调度器]
HealthScheduler[健康检查调度器]
ReconnectionScheduler[重连调度器]
CleanupScheduler[清理调度器]
end
subgraph "工作线程池"
HealthWorkers[健康检查工作线程<br/>并发执行]
ReconnectionWorkers[重连工作线程<br/>优先级队列]
CleanupWorkers[清理工作线程<br/>后台执行]
end
subgraph "缓存层"
StateCache[状态缓存<br/>快速访问]
HealthCache[健康缓存<br/>减少检查]
MetadataCache[元数据缓存<br/>性能优化]
end
subgraph "批处理优化"
BatchHealthCheck[批量健康检查]
BatchStateUpdate[批量状态更新]
BatchNotification[批量通知]
end
MainScheduler --> HealthScheduler
MainScheduler --> ReconnectionScheduler
MainScheduler --> CleanupScheduler
HealthScheduler --> HealthWorkers
ReconnectionScheduler --> ReconnectionWorkers
CleanupScheduler --> CleanupWorkers
HealthWorkers --> StateCache
ReconnectionWorkers --> HealthCache
CleanupWorkers --> MetadataCache
HealthWorkers --> BatchHealthCheck
HealthWorkers --> BatchStateUpdate
HealthWorkers --> BatchNotification