My intention is this.
My client.html calls a php script check.php via ajax. I want check.php to check if another script task.php is already being run. If it is, I do not
Any other attempts to run it will end as soon as the lock() function is invoked.
//try to set a global exclusive lock on the file invoking this function and die if not successful
function lock(){
$file = isset($_SERVER['SCRIPT_FILENAME'])?
realpath($_SERVER['SCRIPT_FILENAME']):
(isset($_SERVER['PHP_SELF'])?realpath($_SERVER['PHP_SELF']):false);
if($file && file_exists($file)){
//global handle stays alive for the duration if this script running
global $$file;
if(!isset($$file)){$$file = fopen($file,'r');}
if(!flock($$file, LOCK_EX|LOCK_NB)){
echo 'This script is already running.'."\n";
die;
}
}
}
Run this in one shell and try to run it in another while it is waiting for input.
lock();
//this will pause execution until an you press enter
echo '...continue? [enter]';
$handle = fopen("php://stdin","r");
$line = fgets($handle);
fclose($handle);
I would use an flock() based mechanism to make sure that task.php
runs only once.
Use a code like this:
<?php
$fd = fopen('lock.file', 'w+');
// try to get an exclusive lock. LOCK_NB let the operation not blocking
// if a process instance is already running. In this case, the else
// block will being entered.
if(flock($fd, LOCK_EX | LOCK_NB )) {
// run your code
sleep(10);
// ...
flock($fd, LOCK_UN);
} else {
echo 'already running';
}
fclose($fd);
Also note that flock()
is, as the PHP documentation points out, portable across all supported operating systems.
!$
gives you the pid of the last executed program in bash. Like this:
command &
pid=$!
echo pid
Note that you will have to make sure your php code runs on a system with bash support. (Not windows)
Update (after comment of opener).
flock()
will work on all operating systems (As I mentioned). The problem I see in your code when working with windows is the !$
(As I mentioned ;) ..
To obtain the pid of the task.php you should use proc_open() to start task.php. I've prepared two example scripts:
task.php
$fd = fopen('lock.file', 'w+');
// try to get an exclusive lock. LOCK_NB let the operation not blocking
// if a process instance is already running. In this case, the else
// block will being entered.
if(flock($fd, LOCK_EX | LOCK_NB )) {
// your task's code comes here
sleep(10);
// ...
flock($fd, LOCK_UN);
echo 'success';
$exitcode = 0;
} else {
echo 'already running';
// return 2 to let check.php know about that
// task.php is already running
$exitcode = 2;
}
fclose($fd);
exit($exitcode);
check.php
$cmd = 'php task.php';
$descriptorspec = array(
0 => array('pipe', 'r'), // STDIN
1 => array('pipe', 'w'), // STDOUT
2 => array('pipe', 'w') // STDERR
);
$pipes = array(); // will be set by proc_open()
// start task.php
$process = proc_open($cmd, $descriptorspec, $pipes);
if(!is_resource($process)) {
die('failed to start task.php');
}
// get output (stdout and stderr)
$output = stream_get_contents($pipes[1]);
$errors = stream_get_contents($pipes[2]);
do {
// get the pid of the child process and it's exit code
$status = proc_get_status($process);
} while($status['running'] !== FALSE);
// close the process
proc_close($process);
// get pid and exitcode
$pid = $status['pid'];
$exitcode = $status['exitcode'];
// handle exit code
switch($exitcode) {
case 0:
echo 'Task.php has been executed with PID: ' . $pid
. '. The output was: ' . $output;
break;
case 1:
echo 'Task.php has been executed with errors: ' . $output;
break;
case 2:
echo 'Cannot execute task.php. Another instance is running';
break;
default:
echo 'Unknown error: ' . $stdout;
}
You asked me why my flock() solution is the best. It's just because the other answer will not reliably make sure that task.php
runs once. This is because the race condition I've mentioned in the comments below that answer.
I think your are really overdoing it with all the processes and background checks. If you run a PHP
script without a session
, then you are already essentially running it in the background. Because it will not block any other request from the user. So make sure you don't call session_start();
Then the next step would be to run it even when the user cancels the request, which is a basic function in PHP
. ignore_user_abort
Last check is to make sure it's only runs once, which can be easily done with creating a file, since PHP
doesnt have an easy application scope.
Combined:
<?php
// Ignore user aborts and allow the script
// to run forever
ignore_user_abort(true);
set_time_limit(0);
$checkfile = "./runningtask.tmp";
//LOCK_EX basicaly uses flock() to prevents racecondition in regards to a regular file open.
if(file_put_contents($checkfile, "running", LOCK_EX)===false) {
exit();
}
function Cleanup() {
global $checkfile;
unlink($checkfile);
}
/*
actual code for task.php
*/
//run cleanup when your done, make sure you also call it if you exit the code anywhere else
Cleanup();
?>
In your javascript you can now call the task.php directly and cancel the request when the connection to the server has been established.
<script>
function Request(url){
if (window.XMLHttpRequest) { // Mozilla, Safari, ...
httpRequest = new XMLHttpRequest();
} else if (window.ActiveXObject) { // IE
httpRequest = new ActiveXObject("Microsoft.XMLHTTP");
} else{
return false;
}
httpRequest.onreadystatechange = function(){
if (httpRequest.readyState == 1) {
//task started, exit
httpRequest.abort();
}
};
httpRequest.open('GET', url, true);
httpRequest.send(null);
}
//call Request("task.php"); whenever you want.
</script>
Bonus points: You can have the actual code for task.php write occasional updates to $checkfile
to have a sense of what is going on. Then you can have another ajax file read the content and show the status to the user.
Too bad I didn't see this before it was accepted..
I have written a class to do just this. ( using file locking ) and PID, process ID checking, on both windows and Linux.
https://github.com/ArtisticPhoenix/MISC/blob/master/ProcLock.php
You can realize it, using lock file:
if(is_file(__DIR__.'/work.lock'))
{
die('Script already run.');
}
else
{
file_put_contents(__DIR__.'/work.lock', '');
// YOUR CODE
unlink(__DIR__.'/work.lock');
}
Lets make the whole process from B to D simple
Step B-D:
$rslt =array(); // output from first exec
$output = array(); // output of task.php execution
//Check if any process by the name 'task.php' is running
exec("ps -auxf | grep 'task.php' | grep -v 'grep'",$rslt);
if(count($rslt)==0) // if none,
exec('php task.php',$output); // run the task,
Explanation:
ps -auxf --> gets all running processes with details
grep 'task.php' --> filter the process by 'task.php' keyword
grep -v 'grep' --> filters the grep process out
NB:
Its also advisable to put the same check in task.php file.
If task.php is executed directly via httpd (webserver), it will only be displayed as a httpd process and cannot be identified by 'ps' command
It wouldn't work under load-balanced environment. [Edited: 17Jul17]