PHP8.2.5-FPM automatically terminated/crashed

[cohekoma]

Description

Description

I've been facing this issue for a while and I couldn't find any luck searching on the internet so I'm open a bug report here. My server is installed with PHP 8.2.5 and using PHP-FPM to serve a WordPress site. During operation, there are two types of FPM issues that I usually see:

1. Master process is killed (checked with ps -ef | grep php-fpm) while all the child processes still running, which then leads to the 502 bad gateway response from Nginx on some pages that are rendered from PHP file. When this happens, I could restart PHP-FPM with php-fpm restart because I will receive ERROR: unable to bind listening socket for address '127.0.0.1:9000': Address already in use (98). The only way is to kill all child processes then start FPM again.

2. Whole PHP-FPM processes including master and children are all terminated. In this case, I could restart using php-fpm restart.

In both case, when I check FPM status, they both give me the php-fpm is dead but pid file exists. The problem is, when PHP-FPM master process is killed (or when PHP-FPM processes are all dead, which means there's also no master process), no error message gets printed to my php_error log file. I couldn't even find any related error message using journalctl.

The issue happens once in a while (for example after 1 or 2 months of operating my site), and my site doesn't really have a high traffic.

PHP Version

PHP 8.2.5

Operating System

No response

[nielsdos]

Note that 8.2.5 is a release from 1.5y ago. (Or did you mean 8.2.25 ?)
Can you please retry on 8.2.25 ?
If possible, also provide a backtrace of the crash.

[cohekoma]

@nielsdos
Hello, yes that's right, it's 8.2.5, the old version, not 8.2.25 (I don't think there's a version 8.2.25 for php either). In case FPM abruptly terminated all processes, how can I do a back trace? Because right now, when all processes are terminated then I couldn't look up for the master process ID.

[cohekoma]

To elaborate, I have been able to capture this with journalctl. Based on this, it seems like I have encountered the segfault issue which then led to the result that my master process exited, leaving my child processes orphaned (and eventually they were all termimated).

[cmb69]

how can I do a back trace?

See https://bugs.php.net/bugs-generating-backtrace.php.

[bukka]

@dinhhoangbaochan As noted by @cmb69, you need to set some things up to get the full core dump. Please follow the step in the link . Without the core dump, it's hard to find out why FPM is crashing.

[github-actions]

No feedback was provided. The issue is being suspended because we assume that you are no longer experiencing the problem. If this is not the case and you are able to provide the information that was requested earlier, please do so. Thank you.