SIGSEGV when running PHP 8.3/8.4 (with docker)

[thueske]

Description

Hey,

First of all, sorry if the issue is wrong here. Please let me know if this is better placed in the Docker-PHP repository.

I use the following image:
https://github.com/docker-library/wordpress/blob/470084224c4a2b4f0c2e19da2af9ab96298739e4/latest/php8.3/fpm-alpine/Dockerfile

For a few weeks now, I have been getting the following errors regularly:

app-1 | [18-Feb-2025 18:50:51] NOTICE: fpm is running, pid 1
app-1 | [18-Feb-2025 18:50:51] NOTICE: ready to handle connections
app-1 | [21-Feb-2025 03:40:39] WARNING: [pool www] child 4048 exited on signal 11 (SIGSEGV - core dumped) after 6183.845205 seconds from start
app-1 | [21-Feb-2025 03:40:39] NOTICE: [pool www] child 4176 started
app-1 | [21-Feb-2025 03:41:40] WARNING: [pool www] child 4145 exited on signal 11 (SIGSEGV - core dumped) after 1380.445253 seconds from start
app-1 | [21-Feb-2025 03:41:40] NOTICE: [pool www] child 4177 started
app-1 | [21-Feb-2025 03:42:40] WARNING: [pool www] child 3920 exited on signal 11 (SIGSEGV - core dumped) after 12647.115671 seconds from start
app-1 | [21-Feb-2025 03:42:40] NOTICE: [pool www] child 4178 started

Unfortunately, this happens ‘from time to time’ (approx. every 2-3 days) on all my WordPress websites with the same setup (approx. 30 of them). The nginx (other container) can then no longer establish a connection to the FPM and responds with 502. The only thing that helps is to restart the FPM container.

I have also tried the image https://github.com/serversideup/docker-php, which is basically just the PHP base image. I also have the problem there with both 8.3 and 8.4.

As I unfortunately don't know how I should continue debugging, I wanted to ask if you have any tips for me.

Thank you very much!

PHP Version

Latest PHP 8.3

Operating System

Alpine

[devnexen]

Hi @thueske would it be possible to do these two things :

• Sharing what you can in term of fpm configuration.
• Generating a backtrace, e.g. like in this fashion.

[thueske]

Hi @thueske would it be possible to do these two things :

• Sharing what you can in term of fpm configuration.
• Generating a backtrace, e.g. like in this fashion.

Would love to do that. Are there any debug docker images that make this easy to do?

[devnexen]

I am not necessarily a docker expert but what is possible is to get a shell in your container and attach gdb to a php process id.

gdb -p <php FPM main process id>
> set follow-fork-mode child
or
gdb -p <php FPM one of the child process id>
...
<eventually segmentation fault occurs>
...
> bt all
<shares with us the output>

Also core dumps could be generated eventually, depending on the /proc/sys/kernel/core_pattern content.
We can do something like this echo "core.%p" | sudo tee /proc/sys/kernel/core_pattern then ulimit -c unlimited that would generate a coredump with core.<process id> which you can use with gdb -c <core dump> <fpm executable, usually php-fpm>.