简化的nginx多进程模型demo
//version 1:以下代码仅演示nginx多进程模型
[test@vm c]$ cat mynginx.c
#include <stdio.h> #include <string.h> #include <unistd.h> char **os_argv; char *os_argv_last; init_setproctitle(void) { char *p; size_t size; int i; size = 0; os_argv_last = os_argv[0]; for (i = 0; os_argv[i]; i++) { if (os_argv_last == os_argv[i]) { os_argv_last = os_argv[i] + strlen(os_argv[i]) + 1; } } os_argv_last += strlen(os_argv_last); } void setproctitle(char *title) { char *p; os_argv[1] = NULL; p = strncpy((char *) os_argv[0], (char *) title, strlen(title)); p += strlen(title); if (os_argv_last - (char *) p > 0) { memset(p, ' ', os_argv_last - (char *) p); } } void start_woker_processes() { setproctitle("mynginx:worker process"); for(;;) { sleep(1); printf("worker pid=%d\n",getpid()); } } void start_dispatcher_process() { setproctitle("mynginx:dispatcher process"); for(;;) { sleep(1); printf("\tdispatcher pid=%d\n",getpid()); } } void save_argv(int argc, char *const *argv) { os_argv = (char **) argv; } int main(int argc, char **argv) { int fd; int i; pid_t pid; save_argv(argc, argv); init_setproctitle(); printf("father pid1=%d\n",getpid()); for (i = 0; i <3; i++) { pid = fork(); if (pid == 0) { start_woker_processes(); } } pid = fork(); if (pid == 0) { start_dispatcher_process(); } printf("father pid2=%d\n",getpid()); while(1) sleep(1); return 0; }
make mynginx
./mynginx
ps -ef |grep mynginx 即可看到
test 20553 20463 0 23:54 pts/0 00:00:00 ./mynginx
test 20554 20553 0 23:54 pts/0 00:00:00 mynginx:worker process
test 20555 20553 0 23:54 pts/0 00:00:00 mynginx:worker process
test 20556 20553 0 23:54 pts/0 00:00:00 mynginx:worker process
test 20557 20553 0 23:54 pts/0 00:00:00 mynginx:dispatcher process
root 20574 20560 0 23:54 pts/2 00:00:00 grep mynginx
version 2：
改进：支持父进程重启被异常中断的子进程。
一个master，一个dispatcher，三个worker
[root@vm-10-154-252-59 nginx-1.7.10]# ps -ef | grep getslice liubin 726 3469 0 20:00 pts/4 00:00:00 ./getslice liubin 727 726 0 20:00 pts/4 00:00:00 getslice:worker process liubin 728 726 0 20:00 pts/4 00:00:00 getslice:worker process liubin 740 726 0 20:00 pts/4 00:00:00 getslice:dispatcher process liubin 745 726 0 20:00 pts/4 00:00:00 getslice:worker process
[root@vm-10-154-252-59 nginx-1.7.10]# kill -9 745
[root@vm-10-154-252-59 nginx-1.7.10]# ps -ef | grep getslice
liubin 726 3469 0 20:00 pts/4 00:00:00 ./getslice
liubin 727 726 0 20:00 pts/4 00:00:00 getslice:worker process
liubin 728 726 0 20:00 pts/4 00:00:00 getslice:worker process
liubin 740 726 0 20:00 pts/4 00:00:00 getslice:dispatcher process
liubin 2714 726 0 20:30 pts/4 00:00:00 getslice:worker process
//可以看到，杀掉745后，马上又起来新的2714进程
下面是代码，可以直接编译
#include <errno.h> #include <signal.h> #include <stdio.h> #include <stdlib.h> #include <string.h> #include <sys/prctl.h> #include <sys/types.h> #include <sys/wait.h> #include <unistd.h> #define DefaultConfigFile "/opt/project/getslice/conf/getslice_lunbo.conf" #define PID_FILENAME "/var/run/getslice_lunbo.pid" #define VERSION "1.0.0" #define MAX_PROCESSES 128 #define PROCESS_NORESPAWN -1 #define PROCESS_JUST_SPAWN -2 #define PROCESS_RESPAWN -3 #define PROCESS_JUST_RESPAWN -4 #define PROCESS_DETACHED -5 const char *appname="getslice_lunbo"; char *ConfigFile = NULL; typedef void (*spawn_proc_pt) (void *data); typedef struct { pid_t pid; int status; //int channel[2]; spawn_proc_pt proc; void *data; char *name; unsigned respawn:1; unsigned just_spawn:1; unsigned detached:1; unsigned exiting:1; unsigned exited:1; } process_t; process_t processes[MAX_PROCESSES]; int process_slot; /* getopt args*/ int opt_no_daemon = 0; int opt_debug_stderr = -1; int opt_parse_cfg_only = 0; int opt_send_signal = -1; sig_atomic_t reap; sig_atomic_t terminate; sig_atomic_t quit; int last_process; int exiting; char **os_argv; char *os_argv_last; init_setproctitle(void) { char *p; size_t size; int i; size = 0; os_argv_last = os_argv[0]; for (i = 0; os_argv[i]; i++) { if (os_argv_last == os_argv[i]) { os_argv_last = os_argv[i] + strlen(os_argv[i]) + 1; } } os_argv_last += strlen(os_argv_last); } void setproctitle(char *title) { char *p; os_argv[1] = NULL; p = strncpy((char *) os_argv[0], (char *) title, strlen(title)); p += strlen(title); if (os_argv_last - (char *) p > 0) { memset(p, ' ', os_argv_last - (char *) p); } } void worker_process_init(int worker) { } void worker_process_exit(void) { } void worker_process_cycle(void *data) { int worker = (intptr_t) data; worker_process_init(worker); setproctitle("getslice:worker process"); for ( ;; ) { if (exiting) { fprintf(stderr, "exiting"); worker_process_exit(); } sleep(1); fprintf(stderr, "\tworker cycle pid: %d\n", getpid()); if (terminate) { fprintf(stderr, "exiting"); worker_process_exit(); } if (quit) { quit = 0; fprintf(stderr, "gracefully shutting down"); setproctitle("worker process is shutting down"); if (!exiting) { exiting = 1; } } } } void dispatcher_process_init(int worker) { } void dispatcher_process_exit() { } void dispatcher_process_cycle(void *data) { int worker = (intptr_t) data; dispatcher_process_init(worker); setproctitle("getslice:dispatcher process"); for ( ;; ) { if (exiting) { fprintf(stderr, "exiting\n"); dispatcher_process_exit(); } sleep(1); fprintf(stderr, "\tdispatcher cycle pid: %d\n", getpid()); if (terminate) { fprintf(stderr, "exiting\n"); dispatcher_process_exit(); } if (quit) { quit = 0; fprintf(stderr, "gracefully shutting down\n"); setproctitle("worker process is shutting down"); if (!exiting) { exiting = 1; } } } } void start_worker_processes(int n, int type) { int i; fprintf(stderr, "start worker processes\n"); for (i = 0; i < n; i++) { spawn_process(worker_process_cycle, (void *) (intptr_t) i, "worker process", type); } } void start_dispatcher_process(int type) { fprintf(stderr, "start dispatcher processes\n"); spawn_process(dispatcher_process_cycle, (void *) (intptr_t) 0, "dispatcher process", type); } void save_argv(int argc, char *const *argv) { os_argv = (char **) argv; } void sig_child(int sig) { reap = 1; int status; int i; pid_t pid; do { pid = waitpid(-1, &status, WNOHANG); for (i = 0; i < last_process; i++) { if (processes[i].pid == pid) { processes[i].status = status; processes[i].exited = 1; //process = processes[i].name; break; } } } while (pid > 0); signal(sig, sig_child); } typedef void SIGHDLR(int sig); void signal_set(int sig, SIGHDLR * func, int flags) { struct sigaction sa; sa.sa_handler = func; sa.sa_flags = flags; sigemptyset(&sa.sa_mask); if (sigaction(sig, &sa, NULL) < 0) fprintf(stderr, "sigaction: sig=%d func=%p: %s\n", sig, func, strerror(errno)); } void init_signals(void) { signal_set(SIGCHLD, sig_child, SA_NODEFER | SA_RESTART); } static pid_t readPidFile(void) { FILE *pid_fp = NULL; const char *f = PID_FILENAME; pid_t pid = -1; int i; if (f == NULL) { fprintf(stderr, "%s: error: no pid file name defined\n", appname); exit(1); } pid_fp = fopen(f, "r"); if (pid_fp != NULL) { pid = 0; if (fscanf(pid_fp, "%d", &i) == 1) pid = (pid_t) i; fclose(pid_fp); } else { if (errno != ENOENT) { fprintf(stderr, "%s: error: could not read pid file\n", appname); fprintf(stderr, "\t%s: %s\n", f, strerror(errno)); exit(1); } } return pid; } int checkRunningPid(void) { pid_t pid; pid = readPidFile(); if (pid < 2) return 0; if (kill(pid, 0) < 0) return 0; fprintf(stderr, "getslice_master is already running! process id %ld\n", (long int) pid); return 1; } void writePidFile(void) { FILE *fp; const char *f = PID_FILENAME; fp = fopen(f, "w+"); if (!fp) { fprintf(stderr, "could not write pid file '%s': %s\n", f, strerror(errno)); return; } fprintf(fp, "%d\n", (int) getpid()); fclose(fp); } void usage(void) { fprintf(stderr, "Usage: %s [-?hvVN] [-d level] [-c config-file] [-k signal]\n" " -h Print help message.\n" " -v Show Version and exit.\n" " -N No daemon mode.\n" " -c file Use given config-file instead of\n" " %s\n" " -k reload|rotate|kill|parse\n" " kill is fast shutdown\n" " Parse configuration file, then send signal to \n" " running copy (except -k parse) and exit.\n", appname, DefaultConfigFile); exit(1); } static void show_version(void) { fprintf(stderr, "%s version: %s\n", appname,VERSION); exit(1); } void mainParseOptions(int argc, char *argv[]) { extern char *optarg; int c; while ((c = getopt(argc, argv, "hvNc:k:?")) != -1) { switch (c) { case 'h': usage(); break; case 'v': show_version(); break; case 'N': opt_no_daemon = 1; break; case 'c': ConfigFile = strdup(optarg); break; case 'k': if ((int) strlen(optarg) < 1) usage(); if (!strncmp(optarg, "reload", strlen(optarg))) opt_send_signal = SIGHUP; else if (!strncmp(optarg, "rotate", strlen(optarg))) opt_send_signal = SIGUSR1; else if (!strncmp(optarg, "shutdown", strlen(optarg))) opt_send_signal = SIGTERM; else if (!strncmp(optarg, "kill", strlen(optarg))) opt_send_signal = SIGKILL; else if (!strncmp(optarg, "parse", strlen(optarg))) opt_parse_cfg_only = 1; /* parse cfg file only */ else usage(); break; case '?': default: usage(); break; } } } void enableCoredumps(void) { /* Set Linux DUMPABLE flag */ if (prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) != 0) fprintf(stderr, "prctl: %s\n", strerror(errno)); /* Make sure coredumps are not limited */ struct rlimit rlim; if (getrlimit(RLIMIT_CORE, &rlim) == 0) { rlim.rlim_cur = rlim.rlim_max; if (setrlimit(RLIMIT_CORE, &rlim) == 0) { fprintf(stderr, "Enable Core Dumps OK!\n"); return; } } fprintf(stderr, "Enable Core Dump failed: %s\n",strerror(errno)); } int reap_children(void) { int i, n; int live; live = 0; for (i = 0; i < last_process; i++) { //child[0] 26718 e:0 t:0 d:0 r:1 j:0 fprintf(stderr,"child[%d] %d e:%d t:%d d:%d r:%d j:%d\n", i, processes[i].pid, processes[i].exiting, processes[i].exited, processes[i].detached, processes[i].respawn, processes[i].just_spawn); if (processes[i].pid == -1) { continue; } if (processes[i].exited) { if (!processes[i].detached) { for (n = 0; n < last_process; n++) { if (processes[n].exited || processes[n].pid == -1) { continue; } fprintf(stderr,"detached:%d\n", processes[n].pid); } } if (processes[i].respawn && !processes[i].exiting && !terminate && !quit) { if (spawn_process(processes[i].proc, processes[i].data, processes[i].name, i) == -1) { fprintf(stderr, "could not respawn %s\n", processes[i].name); continue; } live = 1; continue; } if (i == last_process - 1) { last_process--; } else { processes[i].pid = -1; } } else if (processes[i].exiting || !processes[i].detached) { live = 1; } } return live; } pid_t spawn_process(spawn_proc_pt proc, void *data, char *name, int respawn) { long on; pid_t pid; int s; if (respawn >= 0) { s = respawn; } else { for (s = 0; s < last_process; s++) { if (processes[s].pid == -1) { break; } } if (s == MAX_PROCESSES) { fprintf(stderr, "no more than %d processes can be spawned", MAX_PROCESSES); return -1; } } process_slot = s; pid = fork(); switch (pid) { case -1: fprintf(stderr, "fork() failed while spawning \"%s\" :%s", name, errno); return -1; case 0: pid = getpid(); proc(data); break; default: break; } fprintf(stderr, "start %s %d\n", name, pid); processes[s].pid = pid; processes[s].exited = 0; if (respawn >= 0) { return pid; } processes[s].proc = proc; processes[s].data = data; processes[s].name = name; processes[s].exiting = 0; switch (respawn) { case PROCESS_NORESPAWN: processes[s].respawn = 0; processes[s].just_spawn = 0; processes[s].detached = 0; break; case PROCESS_JUST_SPAWN: processes[s].respawn = 0; processes[s].just_spawn = 1; processes[s].detached = 0; break; case PROCESS_RESPAWN: processes[s].respawn = 1; processes[s].just_spawn = 0; processes[s].detached = 0; break; case PROCESS_JUST_RESPAWN: processes[s].respawn = 1; processes[s].just_spawn = 1; processes[s].detached = 0; break; case PROCESS_DETACHED: processes[s].respawn = 0; processes[s].just_spawn = 0; processes[s].detached = 1; break; } if (s == last_process) { last_process++; } return pid; } int main(int argc, char **argv) { int fd; int i; pid_t pid; sigset_t set; mainParseOptions(argc, argv); if (-1 == opt_send_signal) if (checkRunningPid()) exit(1); enableCoredumps(); writePidFile(); save_argv(argc, argv); init_setproctitle(); init_signals(); sigemptyset(&set); printf("father pid1=%d\n",getpid()); int worker_processes = 3; start_worker_processes(worker_processes, PROCESS_RESPAWN); start_dispatcher_process(PROCESS_RESPAWN); printf("father pid2=%d\n",getpid()); int live = 1; for (;;) { printf("father before suspend\n"); sigsuspend(&set); printf("father after suspend\n"); if (reap) { reap = 0; fprintf(stderr, "reap children\n"); live = reap_children(); } } return 0; }
