09:43:33 @n-tech 8 core web server 09:43:34 @n-tech >nginx 09:43:34 @n-tech 8 worker processes 09:43:34 @n-tech 15000 max connections 09:43:34 @n-tech https://ghostbin.com/paste/xekzp 09:43:34 @n-tech >fpm [...] 09:43:36 @n-tech 8 static processes 09:43:38 @n-tech opcache enabled 09:43:42 @n-tech (tinkered around with ondemand and dynamic as well) 09:43:44 @n-tech php.ini https://ghostbin.com/paste/784ee 09:43:46 @n-tech php-fpm.conf https://ghostbin.com/paste/dyvna 09:43:48 @n-tech www.conf https://ghostbin.com/paste/y9omd 09:43:50 @n-tech @jb [...] 09:45:02 Drybones hey n-tech 09:45:14 Drybones change from socket to localhost on www.conf 09:45:19 OdiliTime Socket 09:45:24 OdiliTime Fpm is on socket 09:45:31 OdiliTime You getting 502s? 09:45:41 @n-tech I've tried both 09:45:46 JB on my server i set pm.max_children to 160 and pm.max_requests to 10000 :x 09:45:50 OdiliTime Well you don't want socket 09:46:02 @n-tech max_requests is not the same as like, max_clients 09:46:08 OdiliTime Jb: geebus how much ram you got? 09:46:14 @n-tech max_requests is the number of requests a process can handle before it restarts itself 09:46:18 @n-tech its default is 0 for never 09:46:24 @n-tech it's to handle memory leaks in 3rd party libs 09:46:30 @n-tech and no 09:46:31 JB 8 GB 09:46:40 @n-tech I've tried both socket and port and it's the same experience on both [...] 09:46:44 OdiliTime Have fun with oomkiller 09:47:01 OdiliTime Well fix that first 09:47:02 @n-tech someone suggested I use truss on a php-fpm process and I'm not sure what I'm supposed to do with this information 09:47:11 OdiliTime And then we can track down other issues 09:47:18 @n-tech fix what first? 09:47:36 OdiliTime Pastebin it 09:47:41 OdiliTime We can read it 09:47:52 OdiliTime Convert the socket to ip 09:47:58 JB call me crazy but i don't think your pm.max_children is high enough 09:48:21 @n-tech I just explained max children is not max clients 09:48:26 @n-tech and even if it were, we do not get 100 requests a second 09:48:31 @n-tech i could set it to 20 and make no difference 09:48:39 @n-tech er 09:48:42 OdiliTime Brb 09:48:43 @n-tech okay look 09:48:58 @n-tech i'll change it but I've been doing this for days, I know exactly what will happen 09:50:14 OdiliTime But looks like you have multiple problems 09:50:17 OdiliTime That's just one 09:50:36 OdiliTime You may have solved the other but failed because of the socket 09:51:14 mona what is the plan for the site? 09:51:22 OdiliTime It won't fix it but it'll get it closer 09:51:55 @n-tech the socket was just a trial, I had it on ports before 09:52:07 @n-tech https://beta.8ch.net/ 09:52:11 @n-tech so it's just restarted 09:52:15 @n-tech it'll be a little faster for a while 09:52:19 @n-tech not like, hhvm instant loading fast 09:52:21 @n-tech but nice 09:52:24 @n-tech it'll degrade 09:52:32 @n-tech It's also on dynamic 09:52:42 @n-tech max children is 80 instead of 8, min is 8 09:52:49 OdiliTime You can plan ram usage better on static 09:53:10 OdiliTime Does that web serve static to? 09:53:23 @n-tech No. 09:54:00 OdiliTime And dedicated web server? Anything else running on it? Memcache? 09:54:01 JB i have never used nginx, does it have its own way of communicating with php? for instance lighttpd overrides fpm settings and sets many of them to much lower/worse values by default 09:54:16 OdiliTime Jb: it uses fastcgi 09:54:36 @n-tech nginx 7fpm redis 09:54:48 @n-tech but it's not using any resources so it could be running a thousand things, it's not using it 09:54:51 JB PHP_FCGI_CHILDREN in fastcgi overrides whatever you have set for pm.max_children and sets it to 1 instead unless you provide your own value to fastcgi 09:54:52 @n-tech this does appear faster 09:54:59 OdiliTime K