systemd看门狗机制应用示例systemd看门狗watchdog是一种监控服务健康状态的机制当服务卡死或崩溃时systemd会自动重启该服务。下面我提供一个完整的示例包含服务程序、配置文件和监控脚本。一、示例程序架构watchdog-example/ ├── service-app.c # 主服务程序需要喂狗 ├── Makefile # 编译文件 ├── mywatchdog.service # systemd服务单元文件 └── watchdog-monitor.sh # 监控脚本可选二、完整代码实现2.1 主服务程序service-app.c/* * systemd看门狗示例服务程序 * 编译: gcc -o service-app service-app.c */#includestdio.h#includestdlib.h#includestring.h#includeunistd.h#includesys/socket.h#includesys/un.h#includesignal.h#includeerrno.h#defineSD_NOTIFY_READYREADY1#defineSD_NOTIFY_WATCHDOGWATCHDOG1#defineSD_NOTIFY_STOPPINGSTOPPING1#defineSD_NOTIFY_STATUSSTATUS#defineNOTIFY_SOCKET_ENVNOTIFY_SOCKETstaticintnotify_socket_fd-1;staticchar*notify_socket_pathNULL;staticvolatilesig_atomic_trunning1;// 信号处理函数staticvoidsignal_handler(intsig){if(sigSIGTERM||sigSIGINT){running0;}}// 向systemd发送通知staticintsd_notify(constchar*message){if(notify_socket_fd0||!notify_socket_path)return-1;structsockaddr_unaddr;memset(addr,0,sizeof(addr));addr.sun_familyAF_UNIX;strncpy(addr.sun_path,notify_socket_path,sizeof(addr.sun_path)-1);returnsendto(notify_socket_fd,message,strlen(message),0,(structsockaddr*)addr,sizeof(addr));}// 喂狗操作staticvoidpet_watchdog(void){if(sd_notify(SD_NOTIFY_WATCHDOG)0){fprintf(stderr,喂狗失败: %s\n,strerror(errno));}else{printf(看门狗已喂食...\n);}}// 模拟工作负载随机卡死staticvoiddo_work(void){// 模拟正常工作printf(服务正在处理任务...\n);// 随机模拟卡死状态用于测试看门狗staticintcounter0;counter;if(counter%50){// 每5次循环模拟一次卡死printf(!!! 服务模拟卡死状态不再喂狗 !!!\n);sleep(15);// 长时间睡眠不喂狗}}intmain(intargc,char*argv[]){intwatchdog_usec0;constchar*env;// 设置信号处理signal(SIGTERM,signal_handler);signal(SIGINT,signal_handler);// 1. 获取NOTIFY_SOCKET环境变量由systemd设置notify_socket_pathgetenv(NOTIFY_SOCKET_ENV);if(!notify_socket_path){fprintf(stderr,未检测到NOTIFY_SOCKET可能未通过systemd启动\n);fprintf(stderr,继续运行但不支持看门狗功能\n);}else{printf(检测到systemd通知套接字: %s\n,notify_socket_path);// 创建通知套接字notify_socket_fdsocket(AF_UNIX,SOCK_DGRAM,0);if(notify_socket_fd0){perror(创建通知套接字失败);return1;}}// 2. 获取看门狗超时时间envgetenv(WATCHDOG_USEC);if(env){watchdog_usecatoi(env);printf(看门狗超时时间: %d 微秒 (%.2f 秒)\n,watchdog_usec,watchdog_usec/1000000.0);}// 3. 通知systemd服务已启动完成if(notify_socket_fd0){sd_notify(SD_NOTIFY_READY);printf(已发送READY通知\n);}// 4. 主循环printf(服务开始运行按CtrlC退出\n);printf(每2秒喂狗一次 (看门狗超时: %.2f秒)\n,watchdog_usec?watchdog_usec/1000000.0:0);while(running){// 执行实际工作do_work();// 喂狗 - 必须在看门狗超时前调用if(notify_socket_fd0watchdog_usec0){pet_watchdog();}// 等待2秒后继续小于看门狗超时时间sleep(2);}// 5. 服务停止前通知systemdif(notify_socket_fd0){sd_notify(SD_NOTIFY_STOPPING);sd_notify(STATUS服务正在停止...);printf(已发送STOPPING通知\n);close(notify_socket_fd);}printf(服务正常退出\n);return0;}2.2 Makefile# Makefile CC gcc CFLAGS -Wall -O2 TARGET service-app SRCS service-app.c all: $(TARGET) $(TARGET): $(SRCS) $(CC) $(CFLAGS) -o $ $^ install: $(TARGET) cp $(TARGET) /usr/local/bin/ cp mywatchdog.service /etc/systemd/system/ chmod 644 /etc/systemd/system/mywatchdog.service clean: rm -f $(TARGET) .PHONY: all install clean2.3 systemd服务单元mywatchdog.service[Unit] DescriptionSystemd Watchdog Example Service Afternetwork.target [Service] Typenotify ExecStart/usr/local/bin/service-app Restartalways RestartSec5 # 看门狗配置 WatchdogSec5 # 看门狗超时时间秒 NotifyAccessall # 允许服务发送通知 StartLimitInterval0 # 无启动限制 # 资源限制可选 CPUQuota50% # CPU使用率限制 MemoryMax100M # 内存限制 # 环境变量可选 EnvironmentMY_ENV_VARtest # 安全配置 PrivateTmpyes NoNewPrivilegesyes ProtectSystemstrict ProtectHomeyes [Install] WantedBymulti-user.target2.4 监控脚本watchdog-monitor.sh#!/bin/bash# 监控systemd看门狗状态的脚本RED\033[0;31mGREEN\033[0;32mYELLOW\033[1;33mNC\033[0m# No ColorSERVICE_NAMEmywatchdog.serviceecho-e${YELLOW} systemd看门狗监控工具 ${NC}# 检查服务状态check_service_status(){localstatus$(systemctl is-active $SERVICE_NAME2/dev/null)localsubstate$(systemctl show-pSubState $SERVICE_NAME|cut-d-f2)if[$statusactive];thenecho-e${GREEN}服务状态:$status${NC}elseecho-e${RED}服务状态:$status${NC}fiecho子状态:$substate}# 查看看门狗配置check_watchdog_config(){localwatchdog_usec$(systemctl show-pWatchdogUSec $SERVICE_NAME|cut-d-f2)localwatchdog_timestamp$(systemctl show-pWatchdogTimestamp $SERVICE_NAME|cut-d-f2)echo看门狗超时:$watchdog_usececho最后喂狗时间:$watchdog_timestamp}# 查看服务重启历史check_restart_history(){localstart_count$(systemctl show-pNRestarts $SERVICE_NAME|cut-d-f2)localstart_limit_hits$(systemctl show-pNRestartHits $SERVICE_NAME|cut-d-f2)echo重启次数:$start_countecho启动限制命中:$start_limit_hits}# 查看最近的日志check_recent_logs(){echo-e\n${YELLOW}最近10条服务日志:${NC}journalctl-u$SERVICE_NAME--since5 minutes ago-n10--no-pager}# 模拟服务卡死测试用simulate_hang(){echo-e${YELLOW}模拟服务卡死发送SIGSTOP信号...${NC}localpid$(systemctl show-pMainPID $SERVICE_NAME|cut-d-f2)if[$pid!0];thenkill-SIGSTOP$pidecho服务PID$pid已暂停看门狗应在5秒后触发重启elseecho-e${RED}无法获取服务PID${NC}fi}# 主菜单whiletrue;doecho-e\n${YELLOW}请选择操作:${NC}echo1) 查看服务状态echo2) 查看看门狗配置echo3) 查看重启历史echo4) 查看最近日志echo5) 模拟服务卡死测试echo6) 退出read-p选择 (1-6): choicecase$choicein1)check_service_status;;2)check_watchdog_config;;3)check_restart_history;;4)check_recent_logs;;5)simulate_hang;;6)exit0;;*)echo-e${RED}无效选择${NC};;esacdone三、部署和测试步骤3.1 编译安装# 1. 编译程序makecleanmake# 2. 安装到系统sudomakeinstall# 3. 重新加载systemd配置sudosystemctl daemon-reload3.2 启动测试# 1. 启动服务sudosystemctl start mywatchdog.service# 2. 查看服务状态sudosystemctl status mywatchdog.service# 输出示例:● mywatchdog.service - Systemd Watchdog Example Service Loaded: loaded(/etc/systemd/system/mywatchdog.service;disabled;vendor preset: enabled)Active: active(running)since Wed2026-03-1814:30:25 CST;10s ago Main PID:12345(service-app)Status:服务正在运行中...Tasks:1(limit:9421)Memory:1.2M CPU: 15ms CGroup: /system.slice/mywatchdog.service └─12345 /usr/local/bin/service-app# 3. 查看服务日志sudojournalctl-umywatchdog.service-f3.3 测试看门狗功能# 方法1使用监控脚本sudo./watchdog-monitor.sh# 选择选项5模拟服务卡死# 方法2手动模拟卡死sudokill-SIGSTOP$(pidof service-app)# 等待5秒后观察服务是否自动重启sudosystemctl status mywatchdog.service# 应该看到服务已重启Main PID已改变四、工作原理图解时序图: systemd看门狗工作流程 systemd 服务程序 内核 | | | |-- 启动服务 ---------------| | | | | |-- 设置WATCHDOG_USEC5s ---| | | | | | |-- sd_notify(READY) ---| |-- 确认服务已启动 ----------| | | | | | | 每隔2秒喂狗 | | |-- sd_notify(WDOG) ----| | | | | | (服务卡死停止喂狗) | | | | |-- 5秒内未收到通知 ---------| | | | | |-- 发送SIGABRT信号 ---------| | | | | | | (进程终止) | | | | |-- 根据Restartalways -----| | |-- 重新启动服务 -------------| |五、关键配置说明5.1 服务文件关键参数参数值说明Typenotify-表示服务支持systemd通知协议WatchdogSec55秒看门狗超时时间服务必须在此时间内喂狗Restartalways-任何原因退出都重启包括看门狗触发NotifyAccessall-允许服务发送所有类型的通知5.2 服务程序关键点函数/宏作用sd_notify(READY)通知systemd服务启动完成sd_notify(WATCHDOG)喂狗操作重置看门狗定时器sd_notify(STOPPING)通知systemd服务正在停止getenv(WATCHDOG_USEC)获取看门狗超时时间微秒六、监控集成在你的Linux维测系统中可以添加对systemd看门狗服务的监控// 在你的监控框架中添加看门狗健康检查typedefstruct{charservice_name[64];intwatchdog_timeout;time_tlast_pet_time;intrestart_count;inthealth_status;}WatchdogMonitor;intcheck_service_watchdog(WatchdogMonitor*wd){charcmd[256];charoutput[1024];// 获取服务状态snprintf(cmd,sizeof(cmd),systemctl show -p WatchdogUSec,NRestarts,ActiveState %s,wd-service_name);FILE*fppopen(cmd,r);if(!fp)return-1;while(fgets(output,sizeof(output),fp)){if(strncmp(output,WatchdogUSec,13)0){wd-watchdog_timeoutatoi(output13)/1000000;}elseif(strncmp(output,NRestarts,10)0){wd-restart_countatoi(output10);}}pclose(fp);// 如果重启次数过多触发告警if(wd-restart_count5){trigger_alert(服务频繁重启可能存在问题,wd-service_name);}return0;}七、总结这个示例展示了服务程序如何实现看门狗通知协议systemd服务配置如何启用看门狗功能喂狗机制的实现方法故障模拟和自动恢复的完整流程监控集成方式通过这个机制你的服务可以在启动时通知systemd服务已就绪定期喂狗防止被误杀在卡死时自动重启提供状态信息给系统管理员这对于构建高可靠性的Linux维测系统非常重要。自测试场景rootubuntu /h/y/r/r/r/w/s/demo2[SIGINT]# systemctl status mywatchdog.service● mywatchdog.service - Systemd Watchdog Example Service Loaded: loaded(/etc/systemd/system/mywatchdog.service;disabled;vendor preset: enabled)Active: active(running)since Wed2026-03-1816:00:24 CST;7min ago Main PID:662806(service-app)Tasks:1(limit:4537)Memory:220.0K CPU: 23ms CGroup: /system.slice/mywatchdog.service └─662806 /usr/local/bin/service-app3月1816:00:24 ubuntu systemd[1]: Starting Systemd Watchdog Example Service...3月1816:00:24 ubuntu systemd[1]: Started Systemd Watchdog Example Service. rootubuntu /h/y/r/r/r/w/s/demo2# kill 662806rootubuntu /h/y/r/r/r/w/s/demo2# journalctl -u mywatchdog.service -f3月1816:10:22 ubuntu service-app[662806]:!!!服务模拟卡死状态不再喂狗!!!3月1816:10:22 ubuntu service-app[662806]: 服务正在处理任务...3月1816:10:22 ubuntu service-app[662806]: 服务正在处理任务...3月1816:10:22 ubuntu service-app[662806]: 服务正在处理任务...3月1816:10:22 ubuntu service-app[662806]: 服务正在处理任务...3月1816:10:22 ubuntu service-app[662806]: 服务正在处理任务...3月1816:10:22 ubuntu service-app[662806]:!!!服务模拟卡死状态不再喂狗!!!3月1816:10:22 ubuntu service-app[662806]: 已发送STOPPING通知3月1816:10:22 ubuntu service-app[662806]: 服务正常退出3月1816:10:22 ubuntu systemd[1]: mywatchdog.service: Deactivated successfully.3月1816:10:27 ubuntu systemd[1]: mywatchdog.service: Scheduled restart job, restart counter is at1.3月1816:10:27 ubuntu systemd[1]: Stopped Systemd Watchdog Example Service.3月1816:10:27 ubuntu systemd[1]: Starting Systemd Watchdog Example Service...3月1816:10:27 ubuntu systemd[1]: Started Systemd Watchdog Example Service. ^C⏎ rootubuntu /h/y/r/r/r/w/s/demo2[SIGINT]# systemctl status mywatchdog.service● mywatchdog.service - Systemd Watchdog Example Service Loaded: loaded(/etc/systemd/system/mywatchdog.service;disabled;vendor preset: enabled)Active: active(running)since Wed2026-03-1816:10:27 CST;35s ago Main PID:663083(service-app)Tasks:1(limit:4537)Memory:220.0K CPU: 18ms CGroup: /system.slice/mywatchdog.service └─663083 /usr/local/bin/service-app3月1816:10:27 ubuntu systemd[1]: Starting Systemd Watchdog Example Service...3月1816:10:27 ubuntu systemd[1]: Started Systemd Watchdog Example Service.