Workers CPU leak [epoll_wait,epoll_ctl]

Hello, I have strange issuses with nginx workers. For some time after start Nginx I notice that some process of workers cause high load to CPU ( principally sys CPU). At first I've got syscall traces from one of such process: futex(0x157d914, FUTEX_WAKE_OP_PRIVATE, 1, 1, 0x157d910,

Re: Workers CPU leak [epoll_wait,epoll_ctl]

P.S: we are using Gentoo with 4.4.1 kernel and CPU X3330 @ 2.66GHz GenuineIntel GNU/Linux Posted at Nginx Forum: https://forum.nginx.org/read.php?2,264764,264766#msg-264766 ___ nginx mailing list nginx@nginx.org

Re: Workers CPU leak [epoll_wait,epoll_ctl]

Thank you. We don't make any changes with files or overwrite them while sendfile processing it. Only create temp file and then mv it. Maybe Is it the same bug concerned with treads aio like in this messeges, https://forum.nginx.org/read.php?21,264701,265016#msg-265016 and it would be fixed in

Re: Workers CPU leak [epoll_wait,epoll_ctl]

Debug log regarding to hanged PID 7479 http://dev.vizl.org/debug.log.txt Posted at Nginx Forum: https://forum.nginx.org/read.php?2,264764,265188#msg-265188 ___ nginx mailing list nginx@nginx.org http://mailman.nginx.org/mailman/listinfo/nginx

Re: Workers CPU leak [epoll_wait,epoll_ctl]

Thank you. Waiting for 1.9.13 branch. Posted at Nginx Forum: https://forum.nginx.org/read.php?2,264764,265536#msg-265536 ___ nginx mailing list nginx@nginx.org http://mailman.nginx.org/mailman/listinfo/nginx

Re: Workers CPU leak [epoll_wait,epoll_ctl]

We found, that we are running 'truncate -s 0' to file before removing them. Can it potentially cause the mentioned above problems ? Posted at Nginx Forum: https://forum.nginx.org/read.php?2,264764,265516#msg-265516 ___ nginx mailing list

Re: Workers CPU leak [epoll_wait,epoll_ctl]

Sorry for long answer, but we have doing some tests, and notice that probles is appear when thread_pool enabled. thread_pool default threads=128 max_queue=1024 We need to use thread_pool, and can't permenent disable it unfortunately Posted at Nginx Forum:

Re: Workers CPU leak [epoll_wait,epoll_ctl]

user www; worker_processes 16; thread_pool default threads=128 max_queue=1024; worker_rlimit_nofile 65536; ###timer_resolution 100ms; #error_log /home/logs/error_log.nginx error; error_log /home/logs/error_log.nginx.debug debug; events { worker_connections 3; use epoll; } http {

Re: nginx caching proxy

And how about result cache file ? Will it be only 1 object in cache when 2 client GET 1 file SIMULTANEOUSLY or 2 different objects in Nginx proxy_cache_path ? Posted at Nginx Forum: https://forum.nginx.org/read.php?2,281621,281686#msg-281686 ___ nginx

Upstream keepalive and fastcgi_next_upstream

Hello I notice strange behavior with keepalive and upstream module. I've backend PHP-FPM with setting configuration request_terminate_timeout = 5s And nginx config: upstream phpfpm { server unix:/tmp/php-fpm-7.sock max_fails=0 fail_timeout=1s; keepalive 8; } location ^\*.php$ { include