多进程失败拉起的demo

#include <iostream>
#include <vector>
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <signal.h>
#include <sys/wait.h>
#include <errno.h>

void RunChild(int idx) {
    for (int i = 0; i < 10; i++) {
        printf("worker %d %d
", getpid(), idx);
        sleep(1);
    }
}

int main() {
    int worker_cnt = 10;
    std::vector<int> worker_pids(worker_cnt);
    for (int i = 0; i < worker_cnt; i++) {
        int pid = fork();
        if(pid != 0) {
            //no process < 0 ?
            worker_pids[i] = pid;
            continue;
        }

        RunChild(i);
        exit(0);
    }

    //avoid some unkown action for SIGCHLD
    signal(SIGCHLD, SIG_DFL);
    while (true) {
        //1.get exit child 
        //2.get worker by pid
        //3.fork it
        //4.child run

        //1.get exit child 
        int pid = -1;
        while((pid = wait(NULL)) == -1) {
            if (errno == EINTR) {
                printf("wait pid err %d %s
", errno, strerror(errno));
                continue;
            }
            else {
                break;
            }
        }
        if(pid == -1) {
            printf("wait pid err %d %s
", errno, strerror(errno));
            //sleep( 1 );
            continue;
        }
        printf("wait pid %d
", pid);

        //2.get worker by pid
        int worker_id = -1;
        for (int i = 0; i < worker_cnt; i++) {
            if (worker_pids[i] == pid) {
                worker_id = i;
                break;
            }
        }
        printf("worker pid %d id %d exit
", pid, worker_id);
        if( worker_id == -1 ) {
            printf("master wait pid %d not worker
", pid);
            continue;
        }

        //3.fork it
        usleep(100000);
        pid = fork();
        if (pid != 0) {
            // no process pid < 0 ?
            printf("%s new worker id %d
", __func__, pid);
            worker_pids[worker_id] = pid;
            continue;
        }

        //4.child run
        RunChild(worker_id);
        exit(0);
    }
    return 0;
}

 这里有个问题，父进程退出了，子进程还没退出。解决方案是在RunChild中调用：

void RunChild(int idx) {
    prctl(PR_SET_PDEATHSIG, SIGHUP);
    for (;true;) {
        printf("worker %d %d
", getpid(), idx);
        sleep(1);
    }
}

可以查看下man 2 prctl。

https://stackoverflow.com/questions/284325/how-to-make-child-process-die-after-parent-exits/284443