Type #bug changed to #feature

for one week logs, on each PHP-FPM server, i have 278.519 useless lines:

$> grep "/ping" /var/log/php8/mypool/access.log{,.1} | wc -l
278519

Moreover I have almost 2% of logs with %C%% (user + system CPU, not tested separately) that display absurdity: 

$> grep "0.00%$" /var/log/php8/mypool/access.log{,.1} | wc -l
277059

$> grep -v "0.00%$" /var/log/php8/mypool/access.log{,.1} | wc -l
3974

$> grep -v "0.00%$" /var/log/php8/mypool/access.log{,.1}  | grep "/ping" | head
- -  19/Nov/2020:22:57:09 +0100 "GET /ping" 200 - 0.161 2048 6211.18%
- -  19/Nov/2020:22:57:44 +0100 "GET /ping" 200 - 0.170 2048 5882.35%
- -  19/Nov/2020:23:04:09 +0100 "GET /ping" 200 - 0.171 2048 5847.95%
- -  19/Nov/2020:23:06:03 +0100 "GET /ping" 200 - 0.101 2048 9900.99%
- -  19/Nov/2020:23:12:00 +0100 "GET /ping" 200 - 0.137 2048 7299.27%
- -  19/Nov/2020:23:17:27 +0100 "GET /ping" 200 - 0.130 2048 7692.31%
- -  19/Nov/2020:23:18:13 +0100 "GET /ping" 200 - 0.139 2048 7194.24%
- -  19/Nov/2020:23:19:47 +0100 "GET /ping" 200 - 0.138 2048 7246.38%
- -  19/Nov/2020:23:23:10 +0100 "GET /ping" 200 - 0.157 2048 6369.43%
- -  19/Nov/2020:23:28:55 +0100 "GET /ping" 200 - 0.215 2048 4651.16%

Really not seeing the need for this, a `grep -v "GET /ping"` does the same in a more generic way.

Pretexting is normal to generate don't wanted logs because of, shell parsing exists ?

I prefer to not have /ping logs as FPM can eliminate this entries ; it will save the resouces. And some economies from here and from there, terminates by a lot of resources, so a lot of economies. Think big, how many logs represents unwanted logs on thousands servers ? Your argument "not seen why" does it match with needs ? Let me bring few elements of response.

Apache permit full conditionnal logs :

- https://httpd.apache.org/docs/trunk/en/logs.html#conditional

Haproxy permit some conditionnal logs :

HAproxt - Options dontlog-normal / dontlognull :
- https://cbonte.github.io/haproxy-dconv/2.4/configuration.html#4.2-option%20dontlog-normal
- https://cbonte.github.io/haproxy-dconv/2.4/configuration.html#4.2-option%20dontlognull

Haproxy - Disabling logging of successful connections :
- https://cbonte.github.io/haproxy-dconv/2.4/configuration.html#chapter-8.3.3

For me and I think for everyone else who is working with PHP in a container environment like Docker this change is very important. In my specific use case I have a healthcheck set up for checking the availability of the php-fpm endpoint in a php-fpm + nginx image. So my healthcheck looks like 

#!/bin/sh

export SCRIPT_NAME=/ping
export SCRIPT_FILENAME=/ping
export REQUEST_METHOD=GET

if ! cgi-fcgi -bind -connect localhost:9000; then
	exit 1
fi

This script gets executed every 10 seconds and so my container logs get filled up with messages like

127.0.0.1 -  03/Feb/2021:10:12:01 +0000 "GET /ping" 200

I wouldn't bother about it if I would not rely on the container logs inside my infrastructure logging system. My logs and their size get bigger and bigger just because of a log that is actually not relevant for me because Docker would take action if the healthcheck is failing. So having a option to disable logging requests to the ping endpoint would be a big step ins saving space and effort getting rid of these entries inside the logs. And I think this option would also come in handy for many other people like DevOps (like me) and SysAdmins.

nikic@php.net > The patch has been given. Could you understand the need ?

Anybody from PHP Team can take this improvement seriously ?

We can discuss, however, this is a need from many of us, SysAdmins.

Our file logs are too bigger, useless to expanse logs for that purpose.

Description:
------------
I have PHP-FPM, I want generate access logs but, not /ping logs when :

ping.path = /ping

I am looking for an option to disable /ping access from the logs, because I have some haproxy servers that ping and it produce quantity of logs i don't need.

May a new config entry could prevent it ?

suggestion ( ping.log = no ) [default]

$> tail /var/log/php8/mypool/access.log
- -  27/Nov/2020:02:12:50 +0100 "GET /ping" 200 - 0.241 2048 0.00%
- -  27/Nov/2020:02:12:54 +0100 "GET /ping" 200 - 0.197 2048 0.00%
- -  27/Nov/2020:02:12:57 +0100 "GET /ping" 200 - 0.238 2048 0.00%
- -  27/Nov/2020:02:12:58 +0100 "GET /ping" 200 - 0.309 2048 0.00%
- -  27/Nov/2020:02:13:01 +0100 "GET /ping" 200 - 0.258 2048 0.00%
- -  27/Nov/2020:02:13:04 +0100 "GET /ping" 200 - 0.190 2048 0.00%
- -  27/Nov/2020:02:13:04 +0100 "GET /ping" 200 - 0.166 2048 0.00%

Closing as duplicate of <https://github.com/php/php-src/issues/8174>.

Keeping this open as it contains more info and it was first.

Although after re-reading the issue and this bug, here's not that much extra info. Anyway doesn't really matter if it's open or not - I'm keeping a note of this and update it here once this is fixed.

It would be great if this could be also implemented for the `/status` page, as it may also be used to determine the liveness or readiness of php-fpm (ex: https://github.com/renatomefi/php-fpm-healthcheck )

Looks like I forgot to close this. 

This has been implemented in https://github.com/php/php-src/commit/327bb219867b16e1161da632fba170f46692484b and will be available in PHP 8.2.

> It would be great if this could be also implemented for the `/status` page

This is possible in the implemented solution: See for example
https://github.com/php/php-src/blob/327bb219867b16e1161da632fba170f46692484b/sapi/fpm/tests/config-array.phpt#L16-L17