zeaslity/ProjectAGiPrompt

Fork 0

Files

zeaslity 631cce9e1e RMDC系统设计文档整体转换为SKILL

2026-01-21 16:15:49 +08:00

27 KiB

Raw Blame History

RMDC-Watchdog 业务流程图

一、项目注册流程

1.1 完整的注册时序流程图

sequenceDiagram
    autonumber
    participant Admin as 管理员<br/>(RMDC Portal)
    participant PM as rmdc-project-management<br/>(项目管理)
    participant ExHub as rmdc-exchange-hub<br/>(消息网关)
    participant MQTT as MQTT Broker
    participant WD as rmdc-watchdog<br/>(二级授权中心)
    participant Node as watchdog-node<br/>(主机守护)
    participant Agent as watchdog-agent<br/>(业务代理)
    
    rect rgb(200, 220, 255)
        Note over Admin,Agent: ===== 阶段1: 项目创建 (RMDC平台侧) =====
    end
    
    Admin->>PM: 1. 创建新项目
    PM->>PM: 2. 生成项目信息<br/>project_id = namespace_<8位随机数>
    PM->>PM: 3. 生成一级授权密钥<br/>tier_one_secret = generateSecret()
    PM->>PM: 4. 生成二级授权密钥<br/>tier_two_secret = generateSecret()
    PM->>PM: 5. 生成时间偏移值<br/>time_offset_allowed
    PM->>PM: 6. 持久化项目信息<br/>{project_id, namespace,<br/>tier_one_secret, tier_two_secret,<br/>auth_duration, time_offset_allowed}
    PM-->>Admin: 7. 返回项目配置文件<br/>(包含tier_one_secret和tier_two_secret<br/>用于部署Watchdog)
    
    rect rgb(220, 255, 220)
        Note over Admin,Agent: ===== 阶段2: Watchdog启动与MQTT连接 =====
    end
    
    Note over WD: 使用项目配置文件部署<br/>本地已有tier_one_secret和tier_two_secret<br/>(密钥不通过公网传输)
    WD->>MQTT: 8. 连接MQTT Broker
    WD->>WD: 9. 创建MQTT订阅<br/>订阅: wdd/RDMC/command/down/<project_id><br/>订阅: wdd/RDMC/message/down/<project_id>
    
    rect rgb(255, 240, 220)
        Note over Admin,Agent: ===== 阶段3: 项目注册 - 挑战应答机制 =====
    end
    
    WD->>WD: 10. 生成Tier-One TOTP验证码<br/>(8位, 30分钟有效)<br/>使用本地tier_one_secret
    WD->>MQTT: 11. 发布注册Command到<br/>wdd/RDMC/command/up<br/>{type:"register", project_id,<br/>namespace, totp_code, env_info}
    MQTT->>ExHub: 12. 转发注册Command
    
    ExHub->>PM: 13. 验证项目信息合法性<br/>GetProjectInfo(project_id)
    PM-->>ExHub: 14. 返回项目密钥等信息<br/>{tier_one_secret, tier_two_secret}
    
    ExHub->>ExHub: 15. 验证TOTP验证码 (可配置开关)<br/>VerifyTierOneTOTP(totp_code, tier_one_secret)
    
    alt TOTP验证成功 或 验证已关闭
        ExHub->>ExHub: 16. 生成32位随机挑战码<br/>challenge = randomString(32)
        ExHub->>ExHub: 17. 生成服务端TOTP响应码<br/>server_totp = GenerateTierOneTOTP(tier_one_secret)
        ExHub->>ExHub: 18. 记录挑战码到缓存<br/>cache[project_id] = challenge
        ExHub->>MQTT: 19. 发布注册确认Message到<br/>wdd/RDMC/message/down/<project_id><br/>{type:"register_ack", challenge, server_totp}
        ExHub->>MQTT: 20. 发布授权Command到<br/>wdd/RDMC/command/down/<project_id><br/>{type:"auth_response", success:true}<br/>(不传输密钥,密钥已在本地)
    else TOTP验证失败
        ExHub->>MQTT: 发布注册拒绝<br/>{type:"register_reject", reason}
    end
    
    MQTT->>WD: 21. 推送注册确认Message
    MQTT->>WD: 22. 推送授权Command
    
    WD->>WD: 23. 验证服务端TOTP (可配置开关)<br/>VerifyTierOneTOTP(server_totp, tier_one_secret)
    
    alt 服务端TOTP验证成功 或 验证已关闭
        WD->>WD: 24. 确认tier_two_secret有效<br/>(使用本地已有的密钥)
        WD->>WD: 25. 解析challenge挑战码
        WD->>MQTT: 26. 发布注册完成Message到<br/>wdd/RDMC/message/up<br/>{type:"register_complete",<br/>project_id, challenge}
    else 服务端TOTP验证失败
        Note over WD: 注册失败,服务端验证不通过<br/>可能是中间人攻击
    end
    
    MQTT->>ExHub: 27. 转发注册完成Message
    ExHub->>ExHub: 28. 验证challenge匹配<br/>cache[project_id] == received_challenge
    
    alt 挑战验证成功
        ExHub->>ExHub: 29. 更新项目状态为Online<br/>UpdateProjectState(project_id, "online")
        ExHub->>PM: 30. 通知项目上线
        PM->>PM: 31. 更新项目在线状态
        Note over WD: 注册成功,开始正常工作
    else 挑战验证失败
        ExHub->>MQTT: 发布验证失败消息
        Note over WD: 注册失败,等待重试
    end
    
    rect rgb(220, 240, 255)
        Note over Admin,Agent: ===== 阶段4: 心跳维持 =====
    end
    
    loop 每5秒
        WD->>MQTT: 发布心跳Message到<br/>wdd/RDMC/message/up<br/>{type:"heartbeat", metrics}
        MQTT->>ExHub: 转发心跳
        ExHub->>ExHub: 刷新项目在线状态
    end

1.2 注册流程关键设计说明

设计要点	说明
密钥生成时机	tier_one_secret 和 tier_two_secret 均在 rmdc-project-management 创建项目时生成
密钥传输方式	密钥通过项目配置文件离线部署到 Watchdog，不通过公网MQTT传输
双向TOTP验证	Watchdog 发送 TOTP 给 ExHub 验证，ExHub 返回 TOTP 给 Watchdog 验证 (可配置开关)
挑战应答机制	32位随机挑战码确保通信双方身份真实性
安全增强	即使 MQTT 被监听，攻击者也无法伪造有效的 TOTP 验证码

二、授权系统完整流程

2.1 授权系统总览架构

graph TB
    subgraph "RMDC平台 (内网)"
        PM["rmdc-project-management<br/>项目管理"]
        CENTER["rmdc-watchdog-center<br/>一级授权中心"]
        EXHUB["rmdc-exchange-hub<br/>消息网关"]
        DB[(授权数据库)]
    end
    
    subgraph "项目环境 (外网/隔离网络)"
        WATCHDOG["rmdc-watchdog<br/>二级授权中心"]
        AGENT1["watchdog-agent<br/>业务A"]
        AGENT2["watchdog-agent<br/>业务B"]
        NODE1["watchdog-node<br/>主机1"]
        NODE2["watchdog-node<br/>主机2"]
    end
    
    PM --"1.项目创建"--> CENTER
    CENTER --"2.授权配置"--> PM
    CENTER <--"3.授权申请/下发<br/>Tier-One TOTP"--> EXHUB
    EXHUB <==" 4.MQTT公网"==> WATCHDOG
    
    NODE1 --"5.主机信息上报"--> WATCHDOG
    NODE2 --"5.主机信息上报"--> WATCHDOG
    
    AGENT1 <--"6.心跳/授权<br/>Tier-Two TOTP"--> WATCHDOG
    AGENT2 <--"6.心跳/授权<br/>Tier-Two TOTP"--> WATCHDOG
    
    CENTER --> DB
    PM --> DB
    
    style CENTER fill:#ff6b6b
    style WATCHDOG fill:#4ecdc4
    style EXHUB fill:#ffd43b
    style PM fill:#a9e34b

2.2 授权申请与下发流程

sequenceDiagram
    autonumber
    participant Node as rmdc-watchdog-node<br/>(主机)
    participant Watchdog as rmdc-watchdog<br/>(二级授权中心)
    participant MQTT as Exchange-Hub<br/>(MQTT)
    participant Center as rmdc-watchdog-center<br/>(一级授权中心)
    participant DB as 授权数据库
    
    rect rgb(200, 220, 255)
        Note over Node,DB: ===== 阶段1: 主机信息收集 =====
    end
    
    Note over Node: 项目启动,以DaemonSet运行
    Node->>Watchdog: 1. 上报主机硬件信息<br/>{MachineID, CPU, Memory, Serial, IP}
    Watchdog->>Watchdog: 2. 加密主机信息<br/>EncryptHostInfo(hostInfo, tierOneSecret)
    Watchdog->>Watchdog: 3. 生成授权申请文件<br/>GenerateAuthorizationFile()
    
    Note over Watchdog: 授权文件包含:<br/>- EncryptedHostMap<br/>- TOTPCode (8位,30分钟有效)<br/>- EncryptedNamespace
    
    rect rgb(220, 255, 220)
        Note over Node,DB: ===== 阶段2: 授权申请 =====
    end
    
    Watchdog->>MQTT: 4. 发布授权申请Command到<br/>wdd/RDMC/command/up<br/>{type:"auth_request", AuthorizationFile}
    MQTT->>Center: 5. 转发授权申请
    
    Center->>Center: 6. 解密项目命名空间<br/>Decrypt(EncryptedNamespace)
    Center->>DB: 7. 获取项目信息<br/>GetProjectInfo(namespace)
    DB-->>Center: 返回项目密钥等信息
    
    Center->>Center: 8. 验证TOTP验证码<br/>VerifyTierOneTOTPCode()
    Center->>Center: 9. 验证主机信息完整性<br/>DecryptHostInfo() 逐个验证
    
    rect rgb(255, 240, 220)
        Note over Node,DB: ===== 阶段3: 授权下发 =====
    end
    
    alt 验证成功
        Center->>Center: 10. 生成新TOTP验证码
        Center->>Center: 11. 构造授权码<br/>{authorized_hosts, expire_time, tier_two_secret}
        Center->>DB: 12. 持久化授权记录
        Center->>MQTT: 13. 发布授权码到<br/>wdd/RDMC/command/down/{project_id}<br/>{type:"auth_response", AuthorizationCode}
    else 验证失败
        Center->>MQTT: 发布授权拒绝消息<br/>{type:"auth_reject", reason}
    end
    
    MQTT->>Watchdog: 14. 推送授权码
    
    Watchdog->>Watchdog: 15. 验证返回的TOTP
    Watchdog->>Watchdog: 16. 解密并验证命名空间
    Watchdog->>Watchdog: 17. 解密每个主机信息
    Watchdog->>Watchdog: 18. 计算时间偏移<br/>timeOffset = now - firstAuthTime
    Watchdog->>Watchdog: 19. 持久化保存授权信息<br/>saveAuthorizationInfo()
    
    Note over Watchdog: 授权存储包含:<br/>- EncryptedAuthorizationCode<br/>- FirstAuthTime<br/>- TimeOffset<br/>- AuthorizedHostMap<br/>- TierTwoTOTPSecret

2.3 Agent授权心跳流程

sequenceDiagram
    autonumber
    participant Agent as rmdc-watchdog-agent<br/>(业务启动器)
    participant Watchdog as rmdc-watchdog<br/>(二级授权中心)
    participant Business as 业务进程<br/>(Java/Python)
    
    rect rgb(200, 255, 200)
        Note over Agent,Business: ===== 首次连接 - 获取密钥 =====
    end
    
    Agent->>Agent: 1. 收集主机信息<br/>GetAllInfo()
    Agent->>Watchdog: 2. 发送心跳请求<br/>{HostInfo, EnvInfo, Timestamp, TOTPCode=""}
    
    Watchdog->>Watchdog: 3. 验证时间戳有效性<br/>|now - timestamp| < 5分钟
    Watchdog->>Watchdog: 4. 添加主机到集合<br/>AddHostInfo()
    Watchdog-->>Agent: 5. 返回响应<br/>{Authorized:false, TierTwoSecret:secret}
    
    Agent->>Agent: 6. 保存TOTP密钥<br/>tierTwoTotpSecret = secret
    
    rect rgb(220, 255, 220)
        Note over Agent,Business: ===== 后续心跳 - 授权验证 =====
    end
    
    loop 心跳循环 (成功后2小时,失败后1小时)
        Agent->>Agent: 7. 生成TOTP验证码<br/>GenerateTierTwoTOTPCode(secret)<br/>6位,30秒有效
        Agent->>Watchdog: 8. 发送心跳请求<br/>{HostInfo, Timestamp, TOTPCode}
        
        Watchdog->>Watchdog: 9. 验证TOTP验证码<br/>VerifyTierTwoTOTPCode()
        
        alt TOTP验证成功
            Watchdog->>Watchdog: 10. 检查主机授权状态<br/>IsHostAuthorized(hostInfo)
            Watchdog->>Watchdog: 11. 生成响应TOTP
            Watchdog-->>Agent: 12. 返回{Authorized:true/false, TOTPCode}
            
            Agent->>Agent: 13. 验证服务端TOTP<br/>双向验证
            
            alt 授权成功
                Agent->>Agent: 14. failCount = 1<br/>等待2小时
            else 授权失败
                Agent->>Agent: 15. failCount++<br/>等待1小时
            end
        else TOTP验证失败
            Watchdog-->>Agent: 返回错误:无效的TOTP验证码
            Agent->>Agent: failCount++
        end
        
        alt failCount >= 12
            Agent->>Business: 16. 发送SIGTERM信号
            Note over Business: 业务进程终止<br/>(死手系统触发)
        end
    end

2.4 授权撤销流程

sequenceDiagram
    autonumber
    participant Admin as 管理员<br/>(RMDC Portal)
    participant PM as project-management
    participant Center as rmdc-watchdog-center<br/>(一级授权中心)
    participant MQTT as Exchange-Hub<br/>(MQTT)
    participant Watchdog as rmdc-watchdog<br/>(二级授权中心)
    participant Agent as rmdc-watchdog-agent
    participant Business as 业务进程
    
    rect rgb(255, 220, 220)
        Note over Admin,Business: ===== 授权撤销流程 =====
    end
    
    Admin->>PM: 1. 发起撤销授权请求<br/>{project_id, reason}
    PM->>Center: 2. 请求撤销项目授权
    Center->>Center: 3. 更新项目授权状态<br/>status = revoked
    Center->>Center: 4. 生成撤销Command<br/>{type:"auth_revoke", project_id, totp}
    
    Center->>MQTT: 5. 发布撤销Command到<br/>wdd/RDMC/command/down/{project_id}
    MQTT->>Watchdog: 6. 推送撤销指令
    
    Watchdog->>Watchdog: 7. 验证撤销指令TOTP
    Watchdog->>Watchdog: 8. 清除本地授权存储<br/>deleteAuthorizationInfo()
    Watchdog->>Watchdog: 9. 设置授权状态为未授权<br/>initialized = false
    
    Note over Agent: 下次心跳时...
    Agent->>Watchdog: 10. 发送心跳请求
    Watchdog-->>Agent: 11. 返回{Authorized:false}
    
    Agent->>Agent: 12. failCount++
    Note over Agent: 连续失败12次后...
    
    Agent->>Business: 13. 发送SIGTERM信号<br/>触发业务进程终止
    Note over Business: 业务自毁<br/>死手系统生效
    
    Watchdog->>MQTT: 14. 上报撤销完成消息<br/>{type:"auth_revoke_ack"}
    MQTT->>Center: 15. 转发撤销确认
    Center->>PM: 16. 更新项目状态
    PM->>Admin: 17. 通知撤销成功

三、K8S指令执行流程

3.1 K8S指令业务流程图

sequenceDiagram
    autonumber
    participant User as 用户<br/>(RMDC Portal)
    participant Api as rmdc-core<br/>(API Gateway)
    participant Operator as octopus-operator<br/>(执行中心)
    participant ExHub as rmdc-exchange-hub<br/>(消息网关)
    participant MQTT as MQTT Broker
    participant WD as rmdc-watchdog<br/>(K8s Operator)
    participant K8s as Kubernetes API
    
    rect rgb(200, 220, 255)
        Note over User,K8s: ===== 阶段1: 指令发起 =====
    end
    
    User->>Api: 1. 发起K8S操作请求<br/>{project_id, action, resource, name}
    Api->>Api: 2. 验证用户权限<br/>CheckPermission(user, project, action)
    Api->>Operator: 3. 调用执行中心API
    
    Operator->>Operator: 4. 构造K8S执行指令<br/>K8sExecCommand{<br/>  command_id, namespace,<br/>  resource, name, action,<br/>  command, timeout<br/>}
    Operator->>ExHub: 5. 调用指令下发API<br/>POST /api/command/send
    
    ExHub->>ExHub: 6. 生成唯一CommandID<br/>记录指令到数据库<br/>状态: Pending
    ExHub->>MQTT: 7. 发布K8S执行Command到<br/>wdd/RDMC/command/down/{project_id}<br/>{type:"k8s_exec", payload}
    
    rect rgb(220, 255, 220)
        Note over User,K8s: ===== 阶段2: 指令执行 =====
    end
    
    MQTT->>WD: 8. 推送K8S执行指令
    WD->>WD: 9. 解析指令<br/>路由到K8sHandler
    WD->>WD: 10. 记录开始时间<br/>status = running
    
    alt action == "logs"
        WD->>K8s: 11a. 调用K8S API<br/>GetPodLogs(namespace, name, container)
    else action == "exec"
        WD->>K8s: 11b. 调用K8S API<br/>ExecCommand(pod, container, command)
    else action == "scale"
        WD->>K8s: 11c. 调用K8S API<br/>ScaleDeployment(name, replicas)
    else action == "restart"
        WD->>K8s: 11d. 调用K8S API<br/>RolloutRestart(deployment)
    else action == "delete"
        WD->>K8s: 11e. 调用K8S API<br/>DeleteResource(resource, name)
    end
    
    K8s-->>WD: 12. 返回执行结果
    
    rect rgb(255, 240, 220)
        Note over User,K8s: ===== 阶段3: 结果返回 =====
    end
    
    WD->>WD: 13. 构造执行结果<br/>ExecResult{command_id, status,<br/>exit_code, output, error, duration}
    WD->>MQTT: 14. 发布结果Message到<br/>wdd/RDMC/message/up<br/>{type:"exec_result", payload}
    
    MQTT->>ExHub: 15. 转发执行结果
    ExHub->>ExHub: 16. 更新指令状态<br/>记录执行时长
    ExHub->>Operator: 17. 推送结果给执行中心
    Operator->>Api: 18. 返回执行结果
    Api->>User: 19. 展示执行结果

3.2 K8S支持的操作类型

Action	说明	目标资源	参数
`logs`	获取日志	Pod	container, tail_lines, follow
`exec`	执行命令	Pod	container, command[], timeout
`scale`	扩缩容	Deployment/StatefulSet	scale_count
`restart`	滚动重启	Deployment/StatefulSet	-
`delete`	删除资源	Pod/Deployment/Service等	-
`get`	获取信息	任意资源	output_format
`apply`	应用配置	任意资源	yaml_content

四、主机指令执行流程

4.1 主机指令业务流程图

sequenceDiagram
    autonumber
    participant User as 用户<br/>(RMDC Portal)
    participant Api as rmdc-core<br/>(API Gateway)
    participant Operator as octopus-operator<br/>(执行中心)
    participant ExHub as rmdc-exchange-hub<br/>(消息网关)
    participant MQTT as MQTT Broker
    participant WD as rmdc-watchdog<br/>(二级授权中心)
    participant Node as watchdog-node<br/>(主机守护)
    
    rect rgb(200, 220, 255)
        Note over User,Node: ===== 阶段1: 指令发起 =====
    end
    
    User->>Api: 1. 发起主机操作请求<br/>{project_id, host_id, action, script}
    Api->>Api: 2. 验证用户权限<br/>CheckPermission(user, project, "host_exec")
    Api->>Operator: 3. 调用执行中心API
    
    Operator->>Operator: 4. 构造主机执行指令<br/>HostExecCommand{<br/>  command_id, host_id,<br/>  action, script, args, timeout<br/>}
    Operator->>ExHub: 5. 调用指令下发API<br/>POST /api/command/send
    
    ExHub->>ExHub: 6. 生成唯一CommandID<br/>记录指令到数据库<br/>状态: Pending
    ExHub->>MQTT: 7. 发布主机执行Command到<br/>wdd/RDMC/command/down/{project_id}<br/>{type:"host_exec", payload}
    
    rect rgb(220, 255, 220)
        Note over User,Node: ===== 阶段2: 指令转发与执行 =====
    end
    
    MQTT->>WD: 8. 推送主机执行指令
    WD->>WD: 9. 解析指令<br/>路由到HostHandler
    WD->>WD: 10. 验证目标主机在线<br/>CheckHostOnline(host_id)
    
    alt 主机在线
        WD->>Node: 11. 转发执行指令到目标Node<br/>HTTP POST /api/exec<br/>(内网通信,TOTP验证)
        Node->>Node: 12. 验证请求合法性<br/>VerifyTOTP()
        Node->>Node: 13. 执行命令/脚本<br/>ExecuteScript(script, args)
        Node-->>WD: 14. 返回执行结果<br/>{exit_code, stdout, stderr}
    else 主机离线
        WD->>WD: 构造错误结果<br/>error = "目标主机离线"
    end
    
    rect rgb(255, 240, 220)
        Note over User,Node: ===== 阶段3: 结果返回 =====
    end
    
    WD->>WD: 15. 构造执行结果<br/>ExecResult{command_id, status,<br/>exit_code, output, error, duration}
    WD->>MQTT: 16. 发布结果Message到<br/>wdd/RDMC/message/up<br/>{type:"exec_result", payload}
    
    MQTT->>ExHub: 17. 转发执行结果
    ExHub->>ExHub: 18. 更新指令状态<br/>记录执行时长
    ExHub->>Operator: 19. 推送结果给执行中心
    Operator->>Api: 20. 返回执行结果
    Api->>User: 21. 展示执行结果

4.2 Watchdog到Node的通信架构

graph TB
    subgraph "rmdc-watchdog"
        WD_Router[消息路由器]
        WD_HostH[HostHandler]
        WD_NodeClient[NodeClient<br/>HTTP Client]
    end
    
    subgraph "watchdog-node (DaemonSet)"
        Node_Server[HTTP Server<br/>:8081]
        Node_Auth[TOTP验证中间件]
        Node_Exec[命令执行器]
        Node_Info[信息收集器]
    end
    
    WD_Router --> WD_HostH
    WD_HostH --> WD_NodeClient
    WD_NodeClient ==HTTP/TOTP==> Node_Server
    Node_Server --> Node_Auth
    Node_Auth --> Node_Exec
    Node_Auth --> Node_Info
    
    style WD_NodeClient fill:#4ecdc4,stroke:#087f5b
    style Node_Server fill:#a9e34b,stroke:#5c940d

4.3 主机支持的操作类型

Action	说明	参数
`exec`	执行Shell命令	script, args[], timeout
`info`	获取主机信息	info_type (cpu/memory/disk/network)
`service`	服务管理	service_name, operation (start/stop/restart)
`file`	文件操作	path, operation (read/write/delete)
`dltu`	镜像操作	operation (download/load/tag/upload), params

五、同步指令设计方案

5.1 同步指令可行性分析

适用场景分析

场景	当前模式	同步需求	实现建议
日志查看	异步	高 - 需要实时看到日志输出	长轮询或推送
主机命令执行	异步	中 - 交互式命令需要实时	超时等待机制
K8S资源查询	异步	中 - 查询操作期望快速返回	超时等待机制
业务更新	异步	低 - 长时间操作适合异步	保持异步 + 通知
监控数据	异步	低 - 定期上报即可	保持异步

技术挑战

网络延迟: 跨公网MQTT通信存在不确定延迟
超时处理: 需要合理设置超时时间避免长时间阻塞
连接保持: 长时间等待需要保持连接不断开
资源占用: 同步等待会占用服务器连接资源

5.2 基于现有架构的同步指令实现方案

方案设计 (基于Go Channel + HTTP长轮询)

根据项目现有的Go技术栈,建议采用HTTP长轮询 + 结果缓存方案:

sequenceDiagram
    autonumber
    participant User as 用户
    participant Api as API Gateway
    participant ExHub as Exchange-Hub
    participant Cache as ResultCache<br/>(sync.Map)
    participant MQTT as MQTT Broker
    participant WD as Watchdog
    
    rect rgb(200, 220, 255)
        Note over User,WD: ===== 同步指令发送 =====
    end
    
    User->>Api: 1. POST /api/command/sync<br/>{project_id, type, payload, timeout:30s}
    Api->>ExHub: 2. 调用同步指令API
    ExHub->>ExHub: 3. 生成CommandID
    ExHub->>Cache: 4. 创建等待通道<br/>waitChan[cmdId] = make(chan Result)
    ExHub->>MQTT: 5. 发布指令
    MQTT->>WD: 6. 推送到Watchdog
    
    rect rgb(220, 255, 220)
        Note over User,WD: ===== 执行与等待 =====
    end
    
    WD->>WD: 7. 执行指令
    WD->>MQTT: 8. 返回结果
    MQTT->>ExHub: 9. 接收结果
    ExHub->>Cache: 10. 发送结果到通道<br/>waitChan[cmdId] <- result
    
    rect rgb(255, 240, 220)
        Note over User,WD: ===== 结果返回 =====
    end
    
    ExHub->>ExHub: 11. select等待结果或超时
    ExHub->>Api: 12. 返回执行结果
    Api->>User: 13. 返回结果给用户

Exchange-Hub同步指令管理器设计

// SyncCommandManager 同步指令管理器
type SyncCommandManager struct {
    waitChannels sync.Map // map[commandID]chan *ExecResult
    timeout      time.Duration
}

// SendAndWait 发送指令并等待结果
func (m *SyncCommandManager) SendAndWait(projectID string, cmd *CommandMessage, timeout time.Duration) (*ExecResult, error) {
    // 1. 创建等待通道
    waitChan := make(chan *ExecResult, 1)
    m.waitChannels.Store(cmd.MessageID, waitChan)
    defer m.waitChannels.Delete(cmd.MessageID)
    
    // 2. 发送指令
    mqttService := GetMQTTService()
    if err := mqttService.PublishCommand(projectID, cmd); err != nil {
        return nil, err
    }
    
    // 3. 等待结果或超时
    select {
    case result := <-waitChan:
        return result, nil
    case <-time.After(timeout):
        return nil, errors.New("command execution timeout")
    }
}

// OnResult 接收结果回调
func (m *SyncCommandManager) OnResult(result *ExecResult) {
    if ch, ok := m.waitChannels.Load(result.CommandID); ok {
        ch.(chan *ExecResult) <- result
    }
}

5.3 日志实时查看方案

对于日志实时查看场景,建议采用流式日志推送方案:

sequenceDiagram
    autonumber
    participant User as 用户
    participant Frontend as 前端
    participant Api as API Gateway
    participant ExHub as Exchange-Hub
    participant MQTT as MQTT Broker
    participant WD as Watchdog
    participant K8s as K8S API
    
    User->>Frontend: 1. 点击查看实时日志
    Frontend->>Api: 2. POST /api/logs/stream<br/>{project_id, pod, container, follow:true}
    Api->>ExHub: 3. 创建日志流会话<br/>sessionId = uuid()
    ExHub->>MQTT: 4. 发布日志查询指令<br/>{type:"log_query", follow:true, session_id}
    MQTT->>WD: 5. 推送指令
    WD->>K8s: 6. 调用K8S Logs API (follow=true)
    
    loop 日志流
        K8s-->>WD: 7. 返回日志行
        WD->>MQTT: 8. 发布日志Message<br/>{type:"log_result", session_id, lines, is_complete:false}
        MQTT->>ExHub: 9. 转发日志
        ExHub->>ExHub: 10. 根据session_id路由
        ExHub->>Api: 11. 推送到API会话
        Api->>Frontend: 12. SSE/轮询返回日志
        Frontend->>User: 13. 实时显示日志
    end
    
    User->>Frontend: 14. 停止查看
    Frontend->>Api: 15. DELETE /api/logs/stream/{session_id}
    Api->>ExHub: 16. 关闭日志流会话
    ExHub->>MQTT: 17. 发布停止指令

前端实现建议

// 使用Server-Sent Events (SSE) 接收实时日志
async function streamLogs(projectId: string, pod: string, container: string) {
  const response = await fetch(`/api/logs/stream`, {
    method: 'POST',
    body: JSON.stringify({ project_id: projectId, pod, container, follow: true })
  });
  
  const sessionId = (await response.json()).session_id;
  
  // 使用SSE或长轮询接收日志
  const eventSource = new EventSource(`/api/logs/stream/${sessionId}`);
  
  eventSource.onmessage = (event) => {
    const logData = JSON.parse(event.data);
    appendLogToUI(logData.lines);
  };
  
  return {
    stop: () => {
      eventSource.close();
      fetch(`/api/logs/stream/${sessionId}`, { method: 'DELETE' });
    }
  };
}

六、数据结构与安全机制汇总

6.1 MQTT Topic与消息类型对照表

方向	Topic	消息类型	说明
上行	`wdd/RDMC/command/up`	`register`	项目注册
上行	`wdd/RDMC/command/up`	`auth_request`	授权申请
上行	`wdd/RDMC/message/up`	`register_complete`	注册完成确认
上行	`wdd/RDMC/message/up`	`heartbeat`	心跳数据
上行	`wdd/RDMC/message/up`	`monitor`	监控数据上报
上行	`wdd/RDMC/message/up`	`exec_result`	指令执行结果
上行	`wdd/RDMC/message/up`	`log_result`	日志查询结果
上行	`wdd/RDMC/message/up`	`alert`	告警信息
下行	`wdd/RDMC/command/down/{id}`	`auth_response`	授权响应
下行	`wdd/RDMC/command/down/{id}`	`auth_revoke`	授权撤销
下行	`wdd/RDMC/command/down/{id}`	`log_query`	日志查询指令
下行	`wdd/RDMC/command/down/{id}`	`host_exec`	主机执行指令
下行	`wdd/RDMC/command/down/{id}`	`k8s_exec`	K8S执行指令
下行	`wdd/RDMC/command/down/{id}`	`update`	业务更新指令
下行	`wdd/RDMC/message/down/{id}`	`register_ack`	注册确认消息

6.2 安全机制汇总

场景	安全机制	参数
Center ↔ Watchdog	Tier-One TOTP + AES-GCM	8位码, 30分钟有效期, SHA256
Watchdog ↔ Agent	Tier-Two TOTP	6位码, 30秒有效期, SHA1
Watchdog ↔ Node	Tier-Two TOTP复用	内网HTTP + TOTP认证
HTTP备用接口	复用Tier-Two TOTP密钥	需要TOTP认证
主机信息	硬件指纹绑定	MachineID+CPU+Memory+Serial
死手系统	心跳失败自毁	连续12次失败触发
消息传输	TLS加密	MQTT over TLS
敏感数据	AES-256-GCM加密	授权码、密钥等

27 KiB Raw Blame History Unescape Escape

RMDC-Watchdog 业务流程图

一、项目注册流程

1.1 完整的注册时序流程图

1.2 注册流程关键设计说明

二、授权系统完整流程

2.1 授权系统总览架构

2.2 授权申请与下发流程

2.3 Agent授权心跳流程

2.4 授权撤销流程

三、K8S指令执行流程

3.1 K8S指令业务流程图

3.2 K8S支持的操作类型

四、主机指令执行流程

4.1 主机指令业务流程图

4.2 Watchdog到Node的通信架构

4.3 主机支持的操作类型

五、同步指令设计方案

5.1 同步指令可行性分析

适用场景分析

技术挑战

5.2 基于现有架构的同步指令实现方案

方案设计 (基于Go Channel + HTTP长轮询)

Exchange-Hub同步指令管理器设计

5.3 日志实时查看方案

前端实现建议

六、数据结构与安全机制汇总

6.1 MQTT Topic与消息类型对照表

6.2 安全机制汇总

27 KiB

Raw Blame History