发表新帖

关于 swoole 使用 systemctl 重启问题

# swoole 重启

版本：swoole 4.5.10

配置文件 service
```ini
[Service]
Type=simple
LimitNOFILE=65535
ExecStart=/usr/bin/php /home/www/content_service/bin/hyperf.php start
ExecStop=/usr/bin/sh /home/www/content_service/bin/stop.sh
ExecReload=/bin/kill -USR1 $MAINPID
Restart=always
RestartSec=5s
TimeoutStopSec=120s
```

stop.sh 伪代码

```sh
pid=$(cat ./runtime/hyperf.pid)
echo $pid
/bin/kill -USR1 $pid
```

/bin/kill -USR1 master_pid

```sh
[root@prod-ali-hd2-content-service-2 ~]# systemctl status content.service
● content.service - content service Http Server
   Loaded: loaded (/home/thefair/www/content_service/bin/content.service; enabled; vendor preset: disabled)
   Active: active (running) since 三 2021-01-20 17:11:19 CST; 2min 42s ago
  Process: 30375 ExecStop=/usr/bin/sh /home/thefair/www/content_service/bin/stop.sh (code=exited, status=0/SUCCESS)
 Main PID: 30474 (php)
   CGroup: /system.slice/content.service
           ├─30474 content_service.Master
           ├─30481 content_service.Manager
           ├─30486 content_service.TaskWorker.4
           ├─30487 content_service.TaskWorker.5
           ├─30488 content_service.Worker.0
           ├─30489 content_service.Worker.1
           ├─30490 content_service.Worker.2
           └─30491 content_service.Worker.3

1月 20 17:11:19 prod-ali-hd2-content-service-2 systemd[1]: Started content service Http Server.
[root@prod-ali-hd2-content-service-2 ~]# systemctl restart content.service
```

## 调试

`strace -p 30481 (manage pid)`

```sh
[root@prod-ali-hd2-content-service-2 ~]# strace -p 30481
strace: Process 30481 attached
wait4(-1,

0x7ffffb4c5268, 0, NULL)      = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
--- SIGUSR1 {si_signo=SIGUSR1, si_code=SI_USER, si_pid=30474, si_uid=0} ---
close(6)                                = 0
open("/home/thefair/logs/www/content_service/service.error.txt", O_RDWR|O_CREAT|O_APPEND, 0666) = 5
--- SIGTERM {si_signo=SIGTERM, si_code=SI_USER, si_pid=1, si_uid=0} ---
--- SIGCONT {si_signo=SIGCONT, si_code=SI_USER, si_pid=1, si_uid=0} ---
--- SIGIO {si_signo=SIGIO, si_code=SI_USER, si_pid=30488, si_uid=0} ---
rt_sigreturn({mask=[USR1 TERM]})        = 0
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=30487, si_uid=0, si_status=0, si_utime=0, si_stime=1} ---
--- SIGIO {si_signo=SIGIO, si_code=SI_USER, si_pid=30490, si_uid=0} ---
rt_sigreturn({mask=[USR1 TERM CHLD]})   = 0
rt_sigreturn({mask=[USR1 TERM]})        = 0
rt_sigreturn({mask=[USR1]})             = 5
rt_sigreturn({mask=[]})                 = -1 EINTR (Interrupted system call)
clock_gettime(CLOCK_REALTIME, {1611134155, 277667141}) = 0
stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=528, ...}) = 0
flock(5, LOCK_EX)                       = 0
write(5, "[2021-01-20 17:15:55 $30481.0]\tI"..., 72) = 72
flock(5, LOCK_UN)                       = 0
rt_sigaction(SIGALRM, {0x7fea01075ff0, [], SA_RESTORER, 0x7fea1758b5d0}, {0x7fea01075ff0, [], SA_RESTORER, 0x7fea1758b5d0}, 8) = 0
setitimer(ITIMER_REAL, {it_interval={6, 0}, it_value={6, 0}}, NULL) = 0
kill(30488, SIGTERM)                    = 0
kill(30489, SIGTERM)                    = 0
kill(30490, SIGTERM)                    = 0
kill(30491, SIGTERM)                    = 0
kill(30486, SIGTERM)                    = 0
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30486
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30820
kill(30487, SIGTERM)                    = 0
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30487
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30821
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30489
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30822
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30490
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30823
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30491
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30824
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 30488
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=30488, si_uid=0, si_status=0, si_utime=1, si_stime=2} ---
rt_sigreturn({mask=[]})                 = 30488
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30839
wait4(-1, 0x7ffffb4c5268, 0, NULL)      = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
--- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
rt_sigreturn({mask=[]})                 = -1 EINTR (Interrupted system call)
setitimer(ITIMER_REAL, {it_interval={0, 1000}, it_value={0, 1000}}, NULL) = 0
wait4(-1, 0x7ffffb4c5268, 0, NULL)      = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
--- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
rt_sigreturn({mask=[]})                 = -1 EINTR (Interrupted system call)
kill(30491, SIG_0)                      = -1 ESRCH (No such process)
kill(30490, SIG_0)                      = -1 ESRCH (No such process)
kill(30489, SIG_0)                      = -1 ESRCH (No such process)
kill(30488, SIG_0)                      = -1 ESRCH (No such process)
kill(30487, SIG_0)                      = -1 ESRCH (No such process)
kill(30486, SIG_0)                      = -1 ESRCH (No such process)
setitimer(ITIMER_REAL, {it_interval={0, 0}, it_value={0, 0}}, NULL) = 0
wait4(-1,  <unfinished ...>
+++ killed by SIGKILL +++
```

## 无法重启

在调度 manage 进程时，又fork了一个 worker 30839 进程，这是一个新的进程ID，导致无法重启，等到 TimeoutStopSec 120s 之后，systemctl kill -9 强制杀死

```sh
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30839
```

发布于5年前 · 4 次浏览 · 来自提问

明志

# swoole 重启

版本：swoole 4.5.10

stop.sh 伪代码

```sh
pid=$(cat ./runtime/hyperf.pid)
echo $pid
/bin/kill -USR1 $pid
```

/bin/kill -USR1 master_pid

1月 20 17:11:19 prod-ali-hd2-content-service-2 systemd[1]: Started content service Http Server.
[root@prod-ali-hd2-content-service-2 ~]# systemctl restart content.service
```

## 调试

`strace -p 30481 (manage pid)`

```sh
[root@prod-ali-hd2-content-service-2 ~]# strace -p 30481
strace: Process 30481 attached
wait4(-1,

## 无法重启

在调度 manage 进程时，又fork了一个 worker 30839 进程，这是一个新的进程ID，导致无法重启，等到 TimeoutStopSec 120s 之后，systemctl kill -9 强制杀死

```sh
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fea1ae47b50) = 30839
```

登录后参与评论

2021-01-21

明志

2021-01-21

序冢回复明志

2021-01-22

Rango 回复明志

2021-01-21

序冢