Creation of child process
The general way to write a child process is:
<?php $pid = pcntl_fork(); if($pid == -1){ //创建失败 die('could not fork'); } else{ if($pid){ //从这里开始写的代码是父进程的 exit("parent!"); } else{ //子进程代码,为防止不停的启用子进程造成系统资源被耗尽的情况,一般子进程代码运行完成后,加入exit来确保子进程正常退出。 exit("child"); } } ?>
If the above code successfully creates a child process, the system will have two processes, one is the parent process and the other is the child process. The ID number of the child process is $pid. When the system reaches $pid = pcntl_fork();, branches are made at this place, and the parent and child processes each start running their own program codes. The running results of the code are parent and child. It's strange, why are the results output in a code where if and else are mutually exclusive? In fact, as mentioned above, when the code is in pcntl_fork, a parent process runs parent and a child process runs child. The parent and child are displayed on the code result. As for the question of who comes first, it depends on the allocation of system resources.
If you need to start multiple processes to process data, you can start sub-processes according to the amount of data, such as 1,000 processes per agreed number. Just use a for loop.
#如果获得的总数小于或等于0,等待60秒,并退出 if ($count <= 0) { sleep(60); exit; } #如果大于1000,计算需要起的进程数 if ($count > 1000) { $cycleSize = ceil($count/1000); } else { $cycleSize = 1; } for ($i=0; $i<$cycleSize; $i++) { $pid = pcntl_fork(); if($pid == -1) { break; } else { if($pid) { #父进程获得子进程的pid,存入数组 $pidArr[] = $pid; } else { //开始发送,子进程执行完自己的任务后,退出。 exit; } } } while(count($pidArr) > 0) { $myId = pcntl_waitpid(-1, $status, WNOHANG); foreach($pidArr as $key => $pid) { if($myId == $pid) unset($pidArr[$key]); } }
Then use crontab to make this PHP program automatically execute at regular intervals.
Of course, the sample code is relatively simple. Specifically, you need to consider how to prevent multiple sub-processes from executing the same piece of data or when the current process has not completed processing the data, crontab starts executing the PHP file again to enable a new process, etc.
PHP multi-process implementation method
Let’s systematically sort out the implementation of PHP multi-process:
1. Direct method
pcntl_fork() creates a process. The return value in the parent process is the pid of the child process. The return value in the child process is 0. -1 indicates that the process creation failed. Very similar to C.
Test script test.php
<?php // example of multiple processes date_default_timezone_set( 'Asia/Chongqing'); echo "parent start, pid ", getmypid(), "\n" ; beep(); for ($i=0; $i<3; ++$i){ $pid = pcntl_fork(); if ($pid == -1){ die ("cannot fork" ); } else if ($pid > 0){ echo "parent continue \n"; for ($k=0; $k<2; ++$k){ beep(); } } else if ($pid == 0){ echo "child start, pid ", getmypid(), "\n" ; for ($j=0; $j<5; ++$j){ beep(); } exit ; } } // *** function beep(){ echo getmypid(), "\t" , date( 'Y-m-d H:i:s', time()), "\n" ; sleep(1); } ?>
Run from command line
#php -f test.php
Output results
parent start, pid 1793 1793 2013-01-14 15:04:17 parent continue 1793 2013-01-14 15:04:18 child start, pid 1794 1794 2013-01-14 15:04:18 1794 2013-01-14 15:04:19 1793 2013-01-14 15:04:19 1794 2013-01-14 15:04:20 parent continue 1793 2013-01-14 15:04:20 child start, pid 1795 1795 2013-01-14 15:04:20 17931794 2013-01-14 15:04:212013-01-14 15:04:21 1795 2013-01-14 15:04:21 1794 2013-01-14 15:04:22 1795 2013-01-14 15:04:22 parent continue 1793 2013-01-14 15:04:22 child start, pid 1796 1796 2013-01-14 15:04:22 1793 2013-01-14 15:04:23 1796 2013-01-14 15:04:23 1795 2013-01-14 15:04:23 1795 2013-01-14 15:04:24 1796 2013-01-14 15:04:24 1796 2013-01-14 15:04:25 1796 2013-01-14 15:04:26
As you can see, 3 child processes are created and run in parallel with the parent process. One of the lines has a slightly different format than the others,
17931794 2013-01-14 15:04:212013-01-14 15:04:21
Because two processes perform write operations at the same time, a conflict occurs.
2. Blocking method
Using the direct method, after the parent process creates the child process, it does not wait for the child process to end, but continues to run. There doesn't seem to be any problem here. If the php script does not end automatically after running, but is resident in memory, it will cause the problem that the child process cannot be recycled. That is, a zombie process. You can wait for the process to end through the pcntl_wai() method, and then recycle the ended process.
Change the test script to:
$pid = pcntl_fork(); if ($pid == -1){ ... } else if ($pid > 0){ echo "parent continue \n"; pcntl_wait($status); for ($k=0; $k<2; ++$k){ beep(); } } else if ($pid == 0){ ... }
Run from command line
#php -f test.php
Output results
parent start, pid 1807 1807 2013-01-14 15:20:05 parent continue child start, pid 1808 1808 2013-01-14 15:20:06 1808 2013-01-14 15:20:07 1808 2013-01-14 15:20:08 1808 2013-01-14 15:20:09 1808 2013-01-14 15:20:10 1807 2013-01-14 15:20:11 1807 2013-01-14 15:20:12 parent continue child start, pid 1809 1809 2013-01-14 15:20:13 1809 2013-01-14 15:20:14 1809 2013-01-14 15:20:15 1809 2013-01-14 15:20:16 1809 2013-01-14 15:20:17 1807 2013-01-14 15:20:18 1807 2013-01-14 15:20:19 child start, pid 1810 1810 2013-01-14 15:20:20 parent continue 1810 2013-01-14 15:20:21 1810 2013-01-14 15:20:22 1810 2013-01-14 15:20:23 1810 2013-01-14 15:20:24 1807 2013-01-14 15:20:25 1807 2013-01-14 15:20:26
The parent process blocks itself in pcntl_wait() and waits for the child process to finish running before continuing.
3. Non-blocking method
The blocking method loses the parallelism of multiple processes. There is also a way to recycle the completed child process and parallelize it. This is the non-blocking way.
Modify script:
<?php // example of multiple processes date_default_timezone_set( 'Asia/Chongqing'); declare (ticks = 1); pcntl_signal(SIGCHLD, "garbage" ); echo "parent start, pid ", getmypid(), "\n" ; beep(); for ($i=0; $i<3; ++$i){ $pid = pcntl_fork(); if ($pid == -1){ die ("cannot fork" ); } else if ($pid > 0){ echo "parent continue \n"; for ($k=0; $k<2; ++$k){ beep(); } } else if ($pid == 0){ echo "child start, pid ", getmypid(), "\n" ; for ($j=0; $j<5; ++$j){ beep(); } exit (0); } } // parent while (1){ // do something else sleep(5); } // *** function garbage($signal){ echo "signel $signal received\n" ; while (($pid = pcntl_waitpid(-1, $status, WNOHANG))> 0){ echo "\t child end pid $pid , status $status\n" ; } } function beep(){ echo getmypid(), "\t" , date( 'Y-m-d H:i:s', time()), "\n" ; sleep(1); } ?>
Run from command line
#php -f test.php &
Output results
parent start, pid 2066 2066 2013-01-14 16:45:34 parent continue 2066 2013-01-14 16:45:35 child start, pid 2067 2067 2013-01-14 16:45:35 20662067 2013-01-14 16:45:362013-01-14 16:45:36 2067 2013-01-14 16:45:37 parent continue 2066 2013-01-14 16:45:37 child start, pid 2068 2068 2013-01-14 16:45:37 2067 2013-01-14 16:45:38 2068 2013-01-14 16:45:38 2066 2013-01-14 16:45:38 parent continue 2066 2013-01-14 16:45:40 child start, pid 2069 2069 2067 2013-01-14 16:45:40 2013-01-14 16:45:40 2068 2013-01-14 16:45:40 2066 2013-01-14 16:45:41 2069 2013-01-14 16:45:41 2068 2013-01-14 16:45:41 signel 17 received child end pid 2067, status 0 2069 2013-01-14 16:45:42 2068 2013-01-14 16:45:42 2069 2013-01-14 16:45:43 signel 17 received child end pid 2068, status 0 2069 2013-01-14 16:45:44 signel 17 received child end pid 2069, status 0
Multiple processes are running in parallel again, and after running for about 10 seconds, use ps -ef | grep php to view the running processes. There is only one process
lqling 2066 1388 0 16:45 pts/1 00:00:00 php -f t5.php
It is the parent process and the child process was recycled.
Child process exit status
pcntl_waitpid(-1, $status, WNOHANG) $status
Returns the end status of the child process
Multi-threading under windows
The Windows system does not support the pcntl function. Fortunately, there is the curl_multi_exec() tool, which uses internal multi-threads to access multiple links, and each link can be used as a task.
Write script test1.php
<?php date_default_timezone_set( 'Asia/Chongqing'); $tasks = array( 'http://localhost/feedbowl/t2.php?job=task1', 'http://localhost/feedbowl/t2.php?job=task2', 'http://localhost/feedbowl/t2.php?job=task3' ); $mh = curl_multi_init(); foreach ($tasks as $i => $task){ $ch[$i] = curl_init(); curl_setopt($ch[$i], CURLOPT_URL, $task); curl_setopt($ch[$i], CURLOPT_RETURNTRANSFER, 1); curl_multi_add_handle($mh, $ch[$i]); } do {$mrc = curl_multi_exec($mh,$active); } while ($mrc == CURLM_CALL_MULTI_PERFORM); while ($active && $mrc == CURLM_OK) { if (curl_multi_select($mh) != -1) { do {$mrc = curl_multi_exec($mh, $active); } while ($mrc == CURLM_CALL_MULTI_PERFORM); } } // completed, checkout result foreach ($tasks as $j => $task){ if (curl_error($ch[$j])){ echo "task ${j} [$task ] error " , curl_error($ch[$j]), "\r\n" ; } else { echo "task ${j} [$task ] get: \r\n" , curl_multi_getcontent($ch[$j]), "\r\n" ; } } ?>
Write the script test2.php
<?php date_default_timezone_set( 'Asia/Chongqing'); echo "child start, pid ", getmypid(), "\r\n" ; for ($i=0; $i<5; ++$i){ beep(); } exit (0); // *** function beep(){ echo getmypid(), "\t" , date('Y-m-d H:i:s' , time()), "\r\n"; sleep(1); } ?>
Run from command line
#php -f test1.php &
Output results
task 0 [http://localhost/feedbowl/t2.php?job=task1] get: child start, pid 5804 5804 2013-01-15 20:22:35 5804 2013-01-15 20:22:36 5804 2013-01-15 20:22:37 5804 2013-01-15 20:22:38 5804 2013-01-15 20:22:39 task 1 [http://localhost/feedbowl/t2.php?job=task2] get: child start, pid 5804 5804 2013-01-15 20:22:35 5804 2013-01-15 20:22:36 5804 2013-01-15 20:22:37 5804 2013-01-15 20:22:38 5804 2013-01-15 20:22:39 task 2 [http://localhost/feedbowl/t2.php?job=task3] get: child start, pid 5804 5804 2013-01-15 20:22:35 5804 2013-01-15 20:22:36 5804 2013-01-15 20:22:37 5804 2013-01-15 20:22:38 5804 2013-01-15 20:22:39
From the printed time, we can see that multiple tasks are running almost at the same time.