nginx loadbalancer Too many open files

Question

I've a loadbalancer and I get this kind errors:

2017/09/12 11:18:38 [crit] 22348#22348: accept4() failed (24: Too many open files)
2017/09/12 11:18:38 [alert] 22348#22348: *4288962 socket() failed (24: Too many open files) while connecting to upstream, client: x.x.x.x, server: example.com, request: "GET /xxx.jpg HTTP/1.1", upstream: "http://y.y.y.y:80/xxx.jpg", host: "example.com", referrer: "https://example.com/some-page"
2017/09/12 11:18:38 [crit] 22348#22348: *4288962 open() "/usr/local/nginx/html/50x.html" failed (24: Too many open files), client: x.x.x.x, server: example.com, request: "GET /xxx.jpg HTTP/1.1", upstream: "http://y.y.y.y:80/xxx.jpg", host: "example.com", referrer: "https://example.com/some-page"

nginx version: nginx/1.10.1

Os: Debian GNU/Linux 8 (jessie)

The interesting thing is not always get error. Mostly I get a 30-50 lines of errors then nothing in 5-10 minutes. And then once the errors are coming again...

Here is my nginx.conf:

user                    www-data;
pid                     /usr/local/nginx/nginx.pid;
worker_processes        auto;

error_log               /var/log/nginx/error.log;

events {
    worker_connections  30000;
}

http {
    include             mime.types;
    default_type        application/octet-stream;

    fastcgi_buffers 16 16k;
    fastcgi_buffer_size 32k;

    proxy_buffer_size   128k;
    proxy_buffers   4 256k;
    proxy_busy_buffers_size   256k;
    client_max_body_size 500m;
    rewrite_log on;

    log_format          main    '$remote_addr - "$proxy_add_x_forwarded_for" -  [$time_local] "$request" '
                                '$status $body_bytes_sent "$http_referer" '
                                '"$http_user_agent" "$backend" '
                                'rt=$request_time uct="$upstream_connect_time" '
                                'uht="$upstream_header_time" urt="$upstream_response_time"';

    access_log          /var/log/nginx/access.log main;

    sendfile            on;
    tcp_nopush          on;
    tcp_nodelay         on;
    keepalive_timeout   65;

    geoip_country    /etc/nginx/geodb/GeoIP.dat;
    geoip_city       /etc/nginx/geodb/GeoLiteCity.dat;

    include             /etc/nginx/loadbalancer/loadbalancer.conf;

}

And also some info:

$ ulimit -Hn
65536
$ ulimit -Sn
65536
$ sysctl fs.file-nr
fs.file-nr = 2848   0   70000

I don't know if it is worth but this loadbalancer is behind cloudflare.

Answer 1

I've added the following line to the nginx.conf:

worker_rlimit_nofile    20000;

Now it works, I don't get any error since the modification.

I hope it will help someone if have the same problem.