CODEWITHSUNDEEP
phplinuxsegmentation-fault
Lethgir
Lethgir

Reputation: 1

PHP-FPM core dump not generated

I am trying to generate core dumps when PHP fails with WARNING: [pool www] child 2756 exited on signal 11 (SIGSEGV) after 153.518521 seconds from start. I am using a custom Docker container based on Alpine hosted on ECS.

I've followed multiple guides but PHP-FPM doesn't generate core dumps (I can generate core dump with php cli or with kill -s SIGSEGV $$). Guides I followed : - https://ma.ttias.be/generate-php-core-dumps-segfaults-php-fpm/ https://www.bggofurther.com/2015/04/generating-core-dumps-for-php-fpm/ - Core dump file is not generated

My config is the following : 

/ # ulimit -c
unlimited

/ # cat /proc/sys/kernel/core_pattern
/tmp/core-dumps/core-%t-%p

/ # cat /proc/sys/kernel/core_uses_pid
0

/ # php-fpm -tt
[global]
 pid = undefined
 error_log = /proc/self/fd/2
 syslog.ident = php-fpm
 syslog.facility = 24
 log_buffering = yes
 log_level = unknown value
 log_limit = 8192
 emergency_restart_interval = 0s
 emergency_restart_threshold = 0
 process_control_timeout = 0s
 process.max = 0
 process.priority = undefined
 daemonize = no
 rlimit_files = 0
 rlimit_core = 0
 events.mechanism = epoll

[www]
 prefix = undefined
 user = www-data
 group = www-data
 listen = 9000
 listen.backlog = 511
 listen.owner = undefined
 listen.group = undefined
 listen.mode = undefined
 listen.allowed_clients = undefined
 process.priority = undefined
 process.dumpable = no
 pm = dynamic
 pm.max_children = 35
 pm.start_servers = 6
 pm.min_spare_servers = 4
 pm.max_spare_servers = 8
 pm.process_idle_timeout = 10
 pm.max_requests = 500
 pm.status_path = /_/php-fpm/status
 ping.path = undefined
 ping.response = undefined
 access.log = /proc/self/fd/2
 access.format = [%{%Y-%m-%dT%H:%M:%S%z}t] %R "%m %{REQUEST_SCHEME}e://%{HTTP_HOST}e%{REQUEST_URI}e" %s %f %{mili}d %{kilo}M %C%%
 slowlog = undefined
 request_slowlog_timeout = 0s
 request_slowlog_trace_depth = 20
 request_terminate_timeout = 0s
 request_terminate_timeout_track_finished = no
 rlimit_files = 0
 rlimit_core = unlimited
 chroot = undefined
 chdir = undefined
 catch_workers_output = yes
 decorate_workers_output = no
 clear_env = no
 security.limit_extensions = .php .phar

configuration file /usr/local/etc/php-fpm.conf test is successful

I tried with /proc/sys/kernel/core_uses_pid to 1 and with process.dumpable = yes but core dumps are still not generated. I suppose there is an issue with PHP-FPM config because when I execute a script in cli (I mean like php generate_sigsegv.php) the core dump is well generated. However, I have no clues why things are going wrong.

For the CLI I use this script : 

<?php

  function segfault ($i = 1) {
    echo "$i\n";
    segfault($i + 1);
  }

  segfault();

?>

For PHP-FPM I am investigating a bug in my app so I have an API to call to generate SIGSEGV.

Upvotes: 0

Views: 4974

Answers (2)

rovast
rovast

Reputation: 31

I met the problem too, and finally, I solved it by enable another setting in fpm pool conf.

process.dumpable = yes
rlimit_core = unlimited

NOTICE: Not only rlimit_core should be set, But Also process.dumpable.

Upvotes: 3

Lethgir
Lethgir

Reputation: 1

The issue was rights on folder where core dumps are generated. Ensure the user used by PHP-FPM has rights to write on the folder. Because PHP in CLI is executed with the user in your container (mainly root), you may have rights but not PHP-FPM.

Upvotes: 0

Related Questions