I limit_req in nginx and other pseudo-defenses, and imitate a bot attack on a php script with the microtime() function:

 ab -kc 10000 -t 10 http://mysite.com/script.php

The attack begins, pools load the CPU, load average grows. It takes 10 seconds, ab gives statistics, and the pools continue to load the CPU for another 3-4 minutes.

In the config is:

 pm = static pm.max_children = 4 pm.max_requests = 1024

In the processes I see that the pools are being killed, new ones are being created and right there they start loading cp again. Why? After all, the stress test has already ended and new GET requests for php script do not come. Explain, I do not understand :(

  • What is the percentage of incorrectly completed requests? - etki

0