I'm trying to use PHP to trigger a bash script that should never stop running. It's not just that the command needs to run and I don't need to wait for output, it needs to continue running after PHP is finished. This has worked other times (and the question has been asked already), the difference seems to be my bash script has a trap for when it's closed.
Here is my bash script:
#!/bin/bash
set -e
WAIT=5
FILE_LOCK="$1"
echo "Daemon started (PID $$)..."
echo "$$" > "$FILE_LOCK"
trap cleanup 0 1 2 3 6 15
cleanup()
{
echo "Caught signal..."
rm -rf "$FILE_LOCK"
exit 1
}
while true; do
# do things
sleep "$WAIT"
done
And here is my PHP:
$command = '/path/to/script.sh /tmp/script.lock >> /tmp/script.log 2>&1 &';
$lastLine = exec($command, $output, $returnVal);
I see the script run, the lock file get created, then it exits, and the trap removes the lock file. In my /tmp/script.log I see:
Daemon started (PID 55963)...
Caught signal...
What's odd is that this only happens when running the PHP via Apache. From command line it keeps running as expected.
The signal on the trap that's being caught is 0.
I've tried wrapping my command in a bash environment, like $command = '/bin/bash -c "' . addslashes($command) . '"';, also tried adding nohup to the beginning. Nothing seems to be working. Is this possible to do for a never ending script?
Found the problem thanks to #lxg.
My # do things command was giving errors, which was causing the script to exit. For some reason they were suppressed.
When removing set -e from the beginning of my bash script I started seeing the errors output to my log file. Not sure why they didn't show up before.
The issue was in my bash loop it was running PHP commands. Even though my bash user and Apache user are the same, for some reason they had different $PATHs. This meant that when running on command line I was using a PHP7 binary, but when Apache trigged bash commands it was using a PHP5 binary (even though Apache itself is configured to use PHP7). So the application errored out and that is what caused the script to die.
The solution was to explicitly set the PHP binary path in my bash loop.
I was doing this with
BIN_PHP=$(which php)
But on true command line it would return one value (/path/to/php7/bin/php) vs command line initiated by Apache (/path/to/php5/bin/php). Despite Apache being the same as a my command line user, it didn't load the ~/.bashrc which specified my correct PHP path.
I have a Beanstalk MQ where I put task to create APK, and a consumer named AppBuilder.php, which reads messages from Beanstalk MQ, and then exec the command which creates the App (android app).
The AppBuilder.php is run from the crontab. the process is
Crontab runs a health-check.sh shell script
health-check.sh runs AppBuilder.php in background
AppBuilder.php calls exec to create the process
Following is the relevant code snippet(s) from each file:
Root crontab is like so:
* * * * * /opt/cron/health-check/health-check.sh
health-check.sh is like this:
#!/bin/bash
PATH=$PATH:/sbin/
#HEALTH CHECK FOR AppBuilder Daemon
if (( $(ps -ef | grep "[A]ppBuilder" | wc -l) > 0 ))
then
echo "AppBuilder is Running"
else
echo "AppBuilder is Stopped, attempting to restart"
$PHP_CMD /opt/appbuilder/AppBuilder.php &
if pgrep "AppBuilder" > /dev/null
then
echo "AppBuilder is now Running"
else
echo "AppBuilder is still not Running"
fi
fi
AppBuilder.php has following exec command:
exec('sudo sh /var/www/cgi-bin/appbuilder/oneClickApk.sh &', $output, $resultCode);
If I run the AppBuilder.php directly, things work fine. However, from cron, it does not.
I've followed this SO Post, and modified the exec command to the following:
exec('/usr/bin/nohup /usr/bin/sudo /usr/bin/env TERM=xterm /bin/sh /var/www/cgi-bin/appbuilder/oneClickApk.sh &', $output, $resultCode);
However, still things fail. Any clues where this may be going wrong? I've spend a lot of time digging the forums, none helping. Please help!
EDIT 1:
The crontab runs, AppBuilder.php gets initialized, but after exec command, I could not see the oneClickApk.sh in process list
EDIT 2:
I changed the crontab from root to ec2-user, as suggested in comments: Still the process does not run.
Just following your first approach like below. But to fix the issue you have to do the following.
Check your cron-tab entry whether properly set or not.
In cron-tab provide full path to your executable
Check whether cron-tab user account has proper setting to execute sudo
Check whether cron-tab account has proper permission to execute your PHP application as well as your shell script. If not add the account into the proper group so that it can run both the PHP and shell script.
exec('sudo sh /var/www/cgi-bin/appbuilder/oneClickApk.sh &', $output, $resultCode);
Use above command its seems OK.
In you shell script put a wait command and check whether it is working fine or not. Just add only below two lines without any if condition first. If working then you need to check your if condition why they are not satisfying as per your requirement.
$PHP_CMD /opt/appbuilder/AppBuilder.php &
while (( $(ps -ef | grep "[h]ealth-check.sh" | wc -l) <= 0 ))
do
sleep 5
done
sleep 30
I need to make a background script that is spawned by PHP command line script that echos to the SSH session. Essentially, I need to do this linux command:
script/path 2>&1 &
If I just run this command in linux, it works great. Output is still displayed to the screen, but I can still use the same session for other commands. However, when I do this in PHP, it doesn't work the same way.
I've tried:
`script/path 2>&1 &`;
exec("script/path 2>&1 &");
system("script/path 2>&1 &")
...And none of these work. I need it to spawn the process, and then kill itself so that I can free up the session, but I still want the output from the child process to print to the screen.
(please comment if this is unclear... I had a hard time putting this into words :P)
I came up with a solution that works in this case.
I created a wrapper bash script that starts up the PHP script, which in turn spawns a child script that has its output redirected to a file, which the bash script wrapper tails.
Here is the bash script I came up with:
php php_script.php "$#"
ps -ef | grep php_script.log | grep -v grep | awk '{print $2}' | xargs kill > /dev/null 2>&1
if [ "$1" != "-stop" ]
then
tail -f php_script.log -n 0 &
fi
(it also cleans up "tail" processes that are still running so that you don't get a gazillion processes when you run this bash script multiple times)
And then in your php child script, you call the external script like this:
exec("php php_script.php >> php_script.log &");
This way the parent PHP script exits without killing the child script, you still get the output from the child script, and your command prompt is still available for other commands.
I have this in post-commit:
#!/bin/sh
REPOS="$1"
REV="$2"
/usr/bin/php /home/name/svn/scripts/post-commit.php $REPOS $REV
But whatever I do, post-commit.php isn't being executed, not even with a chmod a+rw on it. Also there is no output from exec.
What am I missing?
Update: removed exec > ./logs/log.txt from this example since it seems to confuse people.
try:
#!/bin/sh
REPOS="$1"
REV="$2"
#debug:
echo "------------------------------"
date >> /tmp/debug.txt
echo "$#" >> /tmp/debug.txt
id >> /tmp/debug.txt
env >> /tmp/debug.txt
/usr/bin/php /home/name/svn/scripts/post-commit.php "$REPOS" "$REV" > /full/path/to/log.txt 2>&1
Also, verify that your post script works fine when executed by hand.
exec replaces the current shell process, and doesn't start a new one. So after the exec command, your shell stops.
The purpose of your particular exec command eludes me by the way ... So just remove it and you should be fine.
you'd better to exec a 'cd' first, to a directory where you really want the shell to execute.
i'm not sure the path SVN will have when running this, but of course your script have potential privilege problems
I have a PHP script that listens on a queue. Theoretically, it's never supposed to die. Is there something to check if it's still running? Something like Ruby's God ( http://god.rubyforge.org/ ) for PHP?
God is language agnostic but it would be nice to have a solution that works on windows as well.
I had the same issue - wanting to check if a script is running. So I came up with this and I run it as a cron job. It grabs the running processes as an array and cycles though each line and checks for the file name. Seems to work fine. Replace #user# with your script user.
exec("ps -U #user# -u #user# u", $output, $result);
foreach ($output AS $line) if(strpos($line, "test.php")) echo "found";
In linux run ps as follows:
ps -C php -f
You could then do in a php script:
$output = shell_exec('ps -C php -f');
if (strpos($output, "php my_script.php")===false) {
shell_exec('php my_script.php > /dev/null 2>&1 &');
}
The above code lists all php processes running in full, then checks to see if "my_script.php" is in the list of running processes, if not it runs the process and does not wait for the process to terminate to carry on doing what it was doing.
Just append a second command after the script. When/if it stops, the second command is invoked. Eg.:
php daemon.php 2>&1 | mail -s "Daemon stopped" you#example.org
Edit:
Technically, this invokes the mailer right away, but only completes the command when the php script ends. Doing this captures the output of the php-script and includes in the mail body, which can be useful for debugging what caused the script to halt.
Simple bash script
#!/bin/bash
while [true]; do
if ! pidof -x script.php;
then
php script.php &
fi
done
Not for windows, but...
I've got a couple of long-running PHP scripts, that have a shell script wrapping it. You can optionally return a value from the script that will be checked in the shell-script to exit, restart immediately, or sleep for a few seconds -and then restart.
Here's a simple one that just keeps running the PHP script till it's manually stopped.
#!/bin/bash
clear
date
php -f cli-SCRIPT.php
echo "wait a little while ..."; sleep 10
exec $0
The "exec $0" restarts the script, without creating a sub-process that will have to unravel later (and take up resources in the meantime). This bash script wraps a mail-sender, so it's not a problem if it exits and pauses for a moment.
Here is what I did to combat a similar issue. This helps in the event anyone else has a parameterized php script that you want cron to execute frequently, but only want one execution to run at any time. Add this to the top of your php script, or create a common method.
$runningScripts = shell_exec('ps -ef |grep '.strtolower($parameter).' |grep '.dirname(__FILE__).' |grep '.basename(__FILE__).' |grep -v grep |wc -l');
if($runningScripts > 1){
die();
}
You can write in your crontab something like this:
0 3 * * * /usr/bin/php -f /home/test/test.php my_special_cron
Your test.php file should look like this:
<?php
php_sapi_name() == 'cli' || exit;
if($argv[1]) {
substr_count(shell_exec('ps -ax'), $argv[1]) < 3 || exit;
}
// your code here
That way you will have only one active instace of the cron job with my-special-cron as process key. So you can add more jobs within the same php file.
test.php system_send_emails sendEmails
test.php system_create_orders orderExport
Inspired from Justin Levene's answer and improved it as ps -C doesn't work in Mac, which I need in my case. So you can use this in a php script (maybe just before you need daemon alive), tested in both Mac OS X 10.11.4 & Ubuntu 14.04:
$daemonPath = "FULL_PATH_TO_DAEMON";
$runningPhpProcessesOfDaemon = (int) shell_exec("ps aux | grep -c '[p]hp ".$daemonPath."'");
if ($runningPhpProcessesOfDaemon === 0) {
shell_exec('php ' . $daemonPath . ' > /dev/null 2>&1 &');
}
Small but useful detail: Why grep -c '[p]hp ...' instead of grep -c 'php ...'?
Because while counting processes grep -c 'php ...' will be counted as a process that fits in our pattern. So using a regex for first letter of php makes our command different from pattern we search.
One possible solution is to have it listen on a port using the socket functions. You can check that the socket is still listening with a simple script. Even a monitoring service like pingdom could monitor its status. If it dies, the socket is no longer listening.
Plenty of solutions.. Good luck.
If you have your hands on the script, you can just ask him to set a time value every X times in db, and then let a cron job check if that value is up to date.
troelskn wrote:
Just append a second command after the script. When/if it stops, the second command is invoked. Eg.:
php daemon.php | mail -s "Daemon stopped" you#example.org
This will call mail each time a line is printed in daemon.php (which should be never, but still.)
Instead, use the double ampersand operator to separate the commands, i.e.
php daemon.php & mail -s "Daemon stopped" you#example.org
If you're having trouble checking for the PHP script directly, you can make a trivial wrapper and check for that. I'm not sufficiently familiar with Windows scripting to put how it's done here, but in Bash, it'd look like...
wrapper_for_test_php.sh
#!/bin/bash
php test.php
Then you'd just check for the wrapper like you'd check for any other bash script: pidof -x wrapper_for_test_php.sh
I have used cmder for windows and based on this script I came up with this one that I managed to deploy on linux later.
#!/bin/bash
clear
date
while true
do
php -f processEmails.php
echo "wait a little while for 5 secobds...";
sleep 5
done