Bug #48403 [Com]: shell_exec() is much slower than running from shell

From: Date: Sat, 20 Jan 2018 13:09:01 +0000
Subject: Bug #48403 [Com]: shell_exec() is much slower than running from shell
References: 1  Groups: php.bugs 
Request: Send a blank email to php-bugs+get-213634@lists.php.net to get a copy of this message
Edit report at https://bugs.php.net/bug.php?id=48403&edit=1 ID: 48403 Comment by: spam2 at rhsoft dot net Reported by: gregor at hostgis dot com Summary: shell_exec() is much slower than running from shell Status: Not a bug Type: Bug Package: Performance problem Operating System: Linux PHP Version: 5.2.9 Block user comment: N Private report: N New Comment: > Moral of story: Who runs php on apache anyway? Use nginx/lighttpd + php-fpm! laughable attitude! php-fpm brings a ton of other troubles (php-options per directory/htaccess and so on) and as long as a benchmark on a 6 years old 4-core desktop machine with current patchlevel (metldown/spectre) shows 10000 requests/second on our core-cms show me *one* valid reason for rebuild many hundret of vhost-configs with your perfered toy Previous Comments: ------------------------------------------------------------------------ [2018-01-20 13:01:36] oliver at openbrackets dot net TL;DR this problem does not occur in php-fpm Just for the record, I did some testing with Imagick cmd line calls: (the Fill Area Flag "^" is not supported via Imagick extension) convert img.jpg -resize 200x200^ -gravity center -extent 200x200 thumb.jpg 100 operations on a 2MB jpeg image. Results below: bash loop: for i in {1..100}; do convert ... ; done : 4.4s php-cli : for ($i = 0; $i<100; $i++) { exec($cmd); } : 4.4s php-fpm : for ($i = 0; $i<100; $i++) { exec($cmd); } : 4.4s No difference whatsoever for this workload in the 3 environments. - No slow performance of "convert" - shell forking for each exec() call is insignificant for this test case php -v PHP 7.2.1 Moral of story: Who runs php on apache anyway? Use nginx/lighttpd + php-fpm! ------------------------------------------------------------------------ [2014-08-15 13:08:36] info at sistemasgsl dot com 15/08/2014 Same problem. It's a bug. ------------------------------------------------------------------------ [2009-06-01 23:23:00] gregor at hostgis dot com Our workaround is rather crude: I wrote a Perl CGI program, then have our PHP program run generate an URL to it and call file_get_contents() to collect the output. This is rather barbaric, but if PHP's exec() family runs at 1/6 speed, it seems our only option. This issue occurs only in PHP, and only when running as an Apache DSO. The issue does not occur with CGI programs in Perl or Bourne shell, so is not Apache's environment but something in how PHP and Apache work together. As such, the sentiment that "once we call exec() it's not our problem" seems less than productive. I am glad to continue working on it, as we would be most eager to see a real diagnosis and fix for this. ------------------------------------------------------------------------ [2009-05-28 04:34:04] gregor at hostgis dot com I get "current: -1 max: -1" in all cases. # su - www -c "/var/www/cgi-bin/foof" current: -1 max: -1 # curl http://localhost/cgi-bin/foof current: -1 max: -1 # php foof.php current: -1 max: -1 # curl http://localhost/~gregor/foof.php current: -1 max: -1 I agree, it's something about the shell environment which PHP sets up. The same command in a bash CGI doesn't do this, so it's not Apache forking, but PHP forking under Apache. ------------------------------------------------------------------------ [2009-05-27 23:03:07] rasmus@php.net That points to memory thrashing to me. The brk calls are fiddling with the data segment size. Compile this little test: #include <stdio.h> #include <sys/time.h> #include <sys/resource.h> int main() { struct rlimit rlim; getrlimit(RLIMIT_DATA,&rlim); printf("current: %ld max: %ld\n",rlim.rlim_cur,rlim.rlim_max); } And shell-exec that from php-cli and from php-apache and see what you get. Just want to double check that you are indeed running in the same environment in both. You could also check removing all ulimits from the php-apache environment and see if that makes a difference. It still isn't something we can fix here. By the time you have done the fork+exec, PHP is out of the picture. It is down to the OS and the environment the shell exec is running in at that point, so you need to figure out how that environment differs between the two. ------------------------------------------------------------------------ The remainder of the comments for this report are too long. To view the rest of the comments, please view the bug report online at https://bugs.php.net/bug.php?id=48403 -- Edit this bug report at https://bugs.php.net/bug.php?id=48403&edit=1

« previous php.bugs (#213634) next »