Resolved The timeout specified has expired: AH01075: Error dispatching request to : - without URL in my logs - Error 502 - since php 8.2.13 - randomly

[phildolyon]

Server operating system version

CentOS Linux 7.9.2009

Plesk version and microupdate number

Plesk Obsidian Version 18.0.56 Update #4, last updated on Nov 8, 2023 03:43 AM

I upgrade my PHP to 8.2.13 and I have an issue since couple of days on one of my website ( i have 5 web sites on my plesk )
Before updating php, my website was never down for 4 years

Sometimes (randomly, 2 or 3 times per week ) , my web site is down with error 502.

on my logs there are 2 types of error :
Apache Error : (70007)The timeout specified has expired: AH01075: Error dispatching request to : [ Without any url ]
nginx Error : 23113#0: *1718489 upstream timed out (110: Connection timed out) while reading response header from upstream

I try to restart my PHP-FPM and nothing happens , my web site is still down
I try to restart my Apache server and my web site is up
But 1 or 2 days later, my web site is down again

I think that the Nginx error is caused by my apache server.
But I don't understand why my Apache server is down ? ( only on one site, the others work well )
I have not any big script that can stop my apache server, and i don't need to change the timeout.
I try to downgrade my PHP to 7.4.33 but the issue is still there.
I have a cron every hour , with a little script that use an get_file_content , Is it possible that this is what blocks the php? ( i know there are so many strange issues with get_file_contents )

Please help me

 

[Peter Debik]

Maybe the number of "max_children" in PHP FPM settings (on the PHP configuration page of the domain) is too low?

 

[phildolyon]

it's 5, I don't get a lot of traffic on this website. I don't think that is the problem but i try ! i take all idea

 

[phildolyon]

It was down yesterday ! it's not the solution.
In my logs, 30s before it's down, there is this apache error
AH01068: Got bogus version 100, referer:
And I find this : A website hosted in Plesk contains the error in its log: AH01068: Got bogus version - Support Cases from Plesk Knowledge Base
I don't no if this bug can block Apache but this bug is not fix on CentOS 7.X

I will try to use Nginx only instead Apache + Nginx

 

[Peter Debik]

It could be an infinite rewrite loop that causes more and more instances to spawn. In that case more and more capacity is used to process PHP scripts. Nested "includes" maybe?

 

[WebHostingAce]

In addition to Peters reply, have you tried?

# journalctl -xe

and

# journalctl -u httpd.service

might help to find out the reason for Apache stopped.

 

[phildolyon]

7 days without issue.
I use Nginx instead Apache and i hope it's the solution ! ( it's not really a solution but it's work )
I don't think it was an error in my code, because there would be the same problem with nginx or Apache, no ?

I will use #journalctl to try to find out the reason for Apache Stopped. Thank's

 

[WebHostingAce]

Glad to hear that! ) Yes, I think #journalctl will help. I had a similar issue recently caused by Actomic Mod_Security. The Apache stopped working when the AUM updating the ruleset but your issue might be totally different.

 

[phildolyon]

One month without issue , but bad news, last week my server crashed twice with nginx.
I try a new config for my PHP FPM, i change the pm.max_requests, before unlimited, now at 100
to be continued

 

[Peter Debik]

max_requests can be high, but should not be unlimited. The value ensures that memory leaks (or other issues) cannot grow big enough to crash the service. So maybe a value of 10000 could be good choice.

 

[phildolyon]

max_requests can be high, but should not be unlimited. The value ensures that memory leaks (or other issues) cannot grow big enough to crash the service. So maybe a value of 10000 could be good choice.

thank's ! i tryed 100 randomly i change to 10000

 

[phildolyon]

bad news ! It's down again !
I found this bug ( since PHP 8 ) PHP :: Bug #80931 :: file_get_contents() hangs with HTTP/1.1 if server doesn't close connection with file_get_contents and i use it in my code
i change all file_get_contents with a curl function.
if it's not that, i fold !

 

[phildolyon]

YEEESSSSS i found the solution !
Since PHP 8 i have an issue with File_get_contents() , Curl or fopen() . When i use an absolute URL with my own web site ( file_get_contents('https://www.mywebsite.xxx/xxx.txt') FPM doesn't close connection. If 10 child processes be created at the same time my FPM explode and all my website is down. ( no problem with 8 or 9 but 10 BOOM )
If i use an absolute URL with another website, there is no problem, and if i use an relative URL with my own website, there is no problem too.
I used "Hey" a load testing software ( GitHub - rakyll/hey: HTTP load generator, ApacheBench (ab) replacement ) to help me to find the solution.

I don't know if the problem come from PHP 8.0 or my Plesk optimisation but it's very strange !