利用ACE实现看门狗（守护进程）

主要原理：被守护进程B由看门进程A启动（通过B的进程名字B.exe），在启动之后注册进程的退出事件，当B退出时，A捕获退出事件，调用回调函数F，F会再次启动一次B。

Process_Manager_Class.h

#pragma once
//CProcessManager.h
#include <stdio.h>
#include <stdlib.h>
#include <vector>
#include <string>
#include <map>


#include "ace/OS.h"
#include "ace/streams.h"
#include "ace/Log_Msg.h"
#include "ace/Date_Time.h"
#include "ace/Event_Handler.h"
#include "ace/Reactor.h"
#include "ace/Process_Manager.h"

using namespace std;

// 进程监控，进程崩溃后，自动重新启动进程
class CProcessManager
{
public:
    CProcessManager()
    {
        ACE_DEBUG((LM_DEBUG,"CProcessManager()\n"));
    };
    virtual ~CProcessManager()
    {
        ACE_DEBUG((LM_DEBUG,"~CProcessManager()\n"));
    };
    // 从配置文件读取进程列表信息
    virtual int ReadConfigProcessInfo(char *pFileName);
    // 启动进程，并进行进程监控，进程崩溃后，自动重新启动进程
    virtual int ProcessMonitor();

private:
    vector<string> m_vConfigProcessInfo;       // 进程列表信息
};

// 回调事件，程序退出后，执行handle_exit函数
class CProcessExitHandler: public ACE_Event_Handler
{
public:
    CProcessExitHandler()
    {
        ACE_DEBUG((LM_DEBUG,"CProcessExitHandler(%d)\n",this->id_ = ++ increase_));
    };
    virtual ~CProcessExitHandler()
    {
        ACE_DEBUG((LM_DEBUG,"~CProcessExitHandler(%d)\n",this->id_));
    };
    // 程序退出后，执行该函数
    virtual int handle_exit(ACE_Process* process);
private:
    static int increase_;
    int id_;
};

Process_Manager_Class.cpp

//CProcessManager.cpp
#include "Process_Manager_Class.h"
#include <fstream>
using namespace std;
ACE_Process_Manager *g_pPM;             // 进程管理
map<string, pid_t> g_mapProcessInfo;    // 进程ID管理列表

int
CProcessExitHandler::increase_ = 0;
// 读取进程列表配置文件
int CProcessManager::ReadConfigProcessInfo(char *pFileName)
{
    if(NULL == pFileName) 
    {
        ACE_DEBUG((LM_ERROR, ACE_TEXT(": Parameter FileName is null.\n")));
        return -1;
    }
    // 打开进程列表配置文件
    ifstream fin(pFileName);
    if(!fin)
    {
        ACE_DEBUG((LM_ERROR, ACE_TEXT(": file open error.[%s]\n"), pFileName));
        return -2;
    }
    // 循环读取进程列表配置文件
    while(!fin.eof())
    {
        string program;
        getline(fin,program);
        if(0 == program.size() || program[0] == '#')// 去掉回车换行和空白符,跳过注释
        {
            continue;
        }
        // 把读取的进程信息放到[进程列表信息]全局变量
        m_vConfigProcessInfo.push_back(program);
    }
    // [进程列表信息]不应该为0
    if(m_vConfigProcessInfo.size() == 0)
    {
        ACE_DEBUG((LM_ERROR, ACE_TEXT(": Process list is null.[%s]\n"), pFileName));
        return -3;
    }
    return 0;
}

// 进程监控
// 根据配置文件启动进程，并注册进程退出时回调函数
int CProcessManager::ProcessMonitor()
{
    int nRet = 0;
    vector<string>::iterator itrConfig;
    string strStartProcess;

    m_vConfigProcessInfo.clear();

    // 读取进程列表配置文件
    nRet = ReadConfigProcessInfo("ProcessManager.conf");
    if(nRet != 0)
    {
        ACE_DEBUG((LM_ERROR, 
            ACE_TEXT(": ReadConfigProcessInfo[%s] error.\n"), "ProcessManager.conf"));
        return -1;
    }

    // Instantiate a process manager with space for 100 processes.
    g_pPM = new ACE_Process_Manager(ACE_Process_Manager::DEFAULT_SIZE, ACE_Reactor::instance());
    CProcessExitHandler procExitHandler;
    // 循环配置进程列表，启动进程
    for(itrConfig = m_vConfigProcessInfo.begin(); itrConfig != m_vConfigProcessInfo.end(); ++itrConfig)
    {
        strStartProcess = *itrConfig;
        // 启动进程
        ACE_Process_Options options;
        options.command_line(strStartProcess.c_str());
        pid_t pid = g_pPM->spawn(options);
        // 启动进程失败
        if (pid == ACE_INVALID_PID)
        {
            ACE_DEBUG((LM_ERROR, 
                ACE_TEXT(": start a child process failed(%d)%s\n"), 
                pid, strStartProcess.c_str()));
            return -1;
        }
        ACE_DEBUG((LM_INFO, 
            ACE_TEXT(": start a child process success(%d)%s\n"), 
            pid, strStartProcess.c_str()));
        // 注册回调(进程退出时，调用该回调)
        g_pPM->register_handler(&procExitHandler, pid);
        // 添加启动成功的进程到进程ID管理列表
        g_mapProcessInfo[strStartProcess] = pid;
        ACE_OS::sleep(1);
    }

    // Run the reactor event loop waiting for events to occur.
    ACE_Reactor::instance()->run_reactor_event_loop();
    //ACE_Reactor::instance()->end_reactor_event_loop();
    return 0;
}

// 进程退出回调函数
// 程序退出后，执行该函数，重新启动进程
int CProcessExitHandler::handle_exit(ACE_Process* process)
{
    //等待一段时间再重新启动新进程
    ACE_OS::sleep(3);
    map<string, pid_t>::iterator itrProcess;

    ACE_DEBUG((LM_INFO,
        ACE_TEXT(": Process %d exited with exit code %d\n"),
        process->getpid (), process->return_value()));

    // 循环进程ID管理列表，根据进程ID找到退出的进程，重新启动进程
    for(itrProcess = g_mapProcessInfo.begin(); itrProcess != g_mapProcessInfo.end(); ++itrProcess)
    {
        // 根据进程ID，查找进程名
        if(itrProcess->second == process->getpid())
        {
            // 重新启动进程
            ACE_Process_Options options;
            options.command_line(itrProcess->first.c_str());
            pid_t pid = g_pPM->spawn(options);
            // 重新启动进程失败
            if (pid == ACE_INVALID_PID)
            {
                ACE_DEBUG((LM_ERROR, 
                    ACE_TEXT(": restart a child process error(%d)%s\n"), 
                    pid, itrProcess->first.c_str()));
                return -1;
            }
            ACE_DEBUG((LM_INFO, 
                ACE_TEXT(": restart a child process success(%d)%s\n"), 
                pid, itrProcess->first.c_str()));
            // 注册回调(进程退出时，调用该回调),当前对象没有代码负责释放，
            //对象自始至终只有一个副本，不会创建更多的CProcessExitHandler
            g_pPM->register_handler(this, pid);
            // 添加启动成功的进程到进程ID管理列表
            itrProcess->second = pid;
            break;
        }
    }
    return 0;
}

main.cpp

#ifdef _DEBUG
#pragma comment (lib,"ACEd.lib")
#else
#pragma comment (lib,"ACE.lib")
#endif


#pragma comment( linker, "/subsystem:\"windows\" /entry:\"mainCRTStartup\"")


#include "Process_Manager_Class.h"
#include <string>
int main(int argc, char *argv[])
{
    // Running as a child for test.
    if (argc > 1)
    {
        ACE_OS::sleep(10);
        return 0;
    }
    // set output log file
    ACE_OS::setprogname("ProcessManager");
    ACE_Date_Time tvTime(ACE_OS::gettimeofday());
    string szLogFileName;
    szLogFileName =  "log.log";

    ACE_OSTREAM_TYPE *pLogOutput = new ofstream(szLogFileName.c_str());
    ACE_LOG_MSG->msg_ostream(pLogOutput, true);
    ACE_LOG_MSG->set_flags(ACE_Log_Msg::OSTREAM);
    // Prepends timestamp and message priority to each message
    //ACE_LOG_MSG->set_flags(ACE_Log_Msg::VERBOSE_LITE);
    //ACE_LOG_MSG->clr_flags(ACE_Log_Msg::STDERR);

    ACE_DEBUG((LM_INFO, ACE_TEXT(": ---ProcessManager START---.\n")));
    int nRet = 0;
    CProcessManager *pProcMng = new CProcessManager();
    // 启动进程监控
    nRet = pProcMng->ProcessMonitor();
    if(nRet != 0)
    {
        ACE_DEBUG((LM_ERROR, ACE_TEXT(": ---ProcessMonitor Error---\n")));
    }
    delete pProcMng;
    ACE_DEBUG((LM_INFO, ACE_TEXT(": ---ProcessManager STOP---.\n")));
    return 0;
}