1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
26:
27:
28:
29:
30:
31:
32:
33:
34:
35:
36:
37:
38:
39:
40:
41:
42:
43:
44:
45:
46:
47:
48:
49:
50:
51:
52:
53:
54:
55:
56:
57:
58:
59:
60:
61:
62:
63:
64:
65:
66:
67:
68:
69:
70:
71:
72:
73:

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