[.../bin/php-cgi => how to "digg" deeper and understand what's happening in those process ?]

digging on the web I found this

ps eax |grep php |grep -v grep|awk '{print $1 $8 $15 $19}'

it shows PID - REDIRECT_SCRIPT_URL= ..... usefull stuff to know at least what php "could" have a bad infinite loop on the code....

[.../bin/php-cgi => how to "digg" deeper and understand what's happening in those process ?]

CONTEXT : I have many php scripts executing live...

WHAT I DO :

pgrep -ia php

(to get details of the process containing php...) option -ia show the command...

WHAT I GET :

112233 /opt/cpanel/ea-php80/root/usr/bin/php-cgi
223355 /opt/cpanel/ea-php80/root/usr/bin/php /home/..some.php

QUESTION :
The php processes that shows the executing file are ..../bin/php
The php processes that don't show the command have ..../bin/php-cgi `/opt/cpanel/ea-php80/root/usr/bin/php-cgi` => How can I digg deeper to understand what php is doing in those processes ? they run for long time, can I kill them without causing problems ? what is they comes back, how can I pinpoint if there is a bad line of code executing ?? (note : I checked error_log file => no error are being reported) but thoses php-cgi tends to come back.... why ?

WHAT I TRIED {1} :

pidstat -p 112233

Linux 4.18.0-305.19.1.el8_4.x86_64 (host.example.com) 2021-10-14 _x86_64_ (8 CPU) 05:08:53 PM UID PID %usr %system %guest %wait %CPU CPU Command 05:08:53 PM 1001 1234277 0.00 0.00 0.00 0.00 0.00 6 php-cgi

WHAT I TRIED {2} :

pstree -aps 112233 

systemd,1 
	--switched-root 
	--system 
	--deserialize 17 
		└─httpd,3515834 -k start 
			└─httpd,1233812 -k start 
				└─php-cgi,112233

WHAT I TRIED {3} :

lsof -p 1234277

COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME php-cgi 1234277 stubfeed cwd DIR 252,1 32 62973745 /home/user/public_html/.... 
php-cgi 1234277 stubfeed rtd DIR 252,1 4096 128 / 
php-cgi 1234277 stubfeed txt REG 252,1 6387136 138429576 /opt/cpanel/ea-php80/root/usr/bin/php-cgi 
php-cgi 1234277 stubfeed DEL REG 0,1 110106200 /dev/zero php-cgi 1234277 user mem REG 252,1 315112 578898832 /opt/cpanel/ea-php80/root/usr/lib64/php/modules/mailparse.so 
php-cgi 1234277 user mem

 

[Food for thoughts : mixing output_buffer OFF + sending gzipped compressed html....]

On 8/25/2021 at 5:12 PM, kicken said:

That may or may not work as you expect.  There are lots of potential buffers besides PHP's that can get in the way of trying to stream a response like that.  

If you want to provide incremental updates/progress notifications to a user you really need to design a system that will work with buffering enabled.  Usually this takes the form of executing the task in the background on your server by some means and then outputting a page that uses Javascript to request periodic updates.

My apache is also at output buffer off

I don't use NGINX

and I uninstalled PHP-FPM from my server ==> note : since I did that my websites are 10x to 100x faster.

( I hope no one here will hire a hitman to finish me... when I said "remove PHP-FPM" in stackoverflow I was removed the right to ask new questions....... fans of PHP-FPM are angry at me. )

with all that => echo '1'; sleep(2); echo '2'; sleep(2); echo '3'; sleep(2); echo '4'; sleep(2); .... ==> I see 1, then wait, then 2, then wait, etc. => with PHP-FPM or "fast-cgi" or output_buffer on => i wait very long and then 123

This setup works with centos 7, centos8 and AlmaLinux

 

[Food for thoughts : mixing output_buffer OFF + sending gzipped compressed html....]

On 8/25/2021 at 5:21 PM, requinix said:

WebSockets are also a possibility. They're more complicated to set up, but the advantage is that they're specifically designed to exchange messages like this: you start request to do something, socket emits periodic messages about progress, and it all closes down when complete.

I tried web sockets, but with the secure version wss => was "waiting the end of the script to push the info...."

[Food for thoughts : mixing output_buffer OFF + sending gzipped compressed html....]

Turns out I was deceived by  https://tools.pingdom.com/ ..... it says I don't have gzip.....

but other tools like https://www.webpagetest.org/show I do.

I use Clouflare with brotli and http3 active... the brotli thing does compressed output each echo (output buffer on or off)...

😀

[Food for thoughts : mixing output_buffer OFF + sending gzipped compressed html....]

2 minutes ago, kicken said:

Why?

Content-Encoding

what else can I try other than header('Content-Type:gzip;'); ??

 

**** FOR THE WHY : some php can take long time to execute, so outputting each echo gives "feed backs" to the user.

1 of many example : FFMPEG takes long time to convert, with PHP and can echo stuff while the ffmpeg is being executed to show the compression progress

an other example : some advanced search => with a script that does many queries in different tables => echoing results while continue to search....

an other example : probing travel GDS API => echoing results as they come

 

** Handled by server in PHP ==> in apache or php.ini ?

[Food for thoughts : mixing output_buffer OFF + sending gzipped compressed html....]

CONTEXT (1) : I know that output_buffer off + gzip output don't "mix" well together.

CONTEXT (1) : But I need output_buffer off.

TEST (1) : In php I can gzcompress();

$a = 'this is a sting'.'<br>';
$b = gzcompress($a,6);
echo $b; // but the result is this x�+��,V�D��̼t��";D���

TEST (2) : if I add ( at the top of my php 

header('Content-Type:gzip;');

==> the result is this xœ+ÉÈ,V�¢D…â’̼t›¤";�DáÉ (in the browser)

 

QUESTION : So is there a way to digg deeper and "tell" the receiving browser to ungzip that output ??

[Can't figure out why HTML blocks won't change]

I usually "clean" user input before adding to the database.

the retrieved $_REQUEST['input...'] => I replace all

quotes ' with ’  (right single quotation)

and double quotes " by ” (right double quotation)

when inserting right single or double quotation marks, it fits without the need of escaping....