We are facing issues with a php-fpm server that spawns too many children. This morning, I set up php-fpm's status page, and got an interesting piece of information : among all the scripts that are executed by the the children, one has a request duration that amounts to 3050111212255
µs which equals roughly to 35 days, which is the uptime of the machine, although we restart php-fpm
recently. Here the output of the status page for this script.
pid: 19998
state: Idle
start time: 29/Aug/2014:14:48:25 +0200
start since: 578
requests: 244
request duration: 3050444744915
request method: POST
request URI: /app.php?_format=json
content length: 96
user: jobuser
script: /home/frontoffice/instances/encoding/current/web/app.php
last request cpu: 0.00
last request memory: 7864320
Do you think this is a bug in php-fpm ?
Look like an old bug not resolved yet:
https://bugs.php.net/bug.php?id=62382
People who have same problem can try to add request_terminate_timeout = 300 to their php-fpm pool configuration