Hi, thanks for the replys.

ulimit output: unilimited
To run dstat I will have to wait until php becomes unresponsive, it could be hours or days, I can't reproduce it since it's not releated to the load on the server, I tried stressing it with Apache Benchmark but even with all the cpus at 100% the site was still working fine (only slow), I use munin and every time it happens I don't see any weird behavior apart form the mysql threads that have a spike.

PHP_FCGI_MAX_REQUESTS = 500
I'm using dynamic child with php-fpm (but I've already tried many setting here even without php-fpm)

Here's the first part of the nginx configuration, let me know if you need the whole thing

user www-data www-data;
worker_processes 12;
error_log logs/error.log;
pid /dev/shm/nginx.pid;
worker_rlimit_nofile 10000;

events {
worker_connections 1024;
use epoll;
}

http {
include mime.types;
#include /etc/nginx/proxy.conf;
include /usr/local/nginx/conf/fastcgi_params;

default_type application/octet-stream;

log_format main '$remote_addr - $remote_user [$time_local] "$request" '
'$status $body_bytes_sent "$http_referer" '
'"$http_user_agent" "$http_x_forwarded_for"';

access_log logs/access.log main;
sendfile on;
tcp_nopush on;
server_names_hash_bucket_size 128; # this seems to be required for some vhosts
client_max_body_size 12m;
large_client_header_buffers 8 32k;

# output compression saves bandwidth
gzip on;
gzip_proxied any;
gzip_http_version 1.1;
#gzip_min_length 1100;
gzip_comp_level 2;
#gzip_buffers 4 8k;
gzip_types text/plain text/css application/x-javascript text/xml application/xml application/xml+rss text/javascript application/atom+xml;
#gzip_vary on;
#gzip_disable "MSIE [1-6]\.";