PHP bottlenecks no matter what Plain Text
11:23:47 JB hay n-tech 11:24:37 @n-tech yo 11:25:13 JB try our new idea if u want 11:25:50 JB kill nginx/php, then run command ulimit -n 1024000 11:25:54 JB then start it again 11:26:26 ranix you should see log spam if you crossed ulimit 11:26:37 ranix too many open files etc 11:26:55 ranix it's more interesting to know what it is now than to change it 11:28:51 JB yea, is there anything funky in /var/log/messages 11:29:14 JB i can also help if you see "sending cookies" or "neighbour table overflow" in there 11:32:11 @n-tech i've made the changes to both 11:32:20 @n-tech god this is soulcrushing 11:32:27 mona what is? 11:32:36 OdiliTime he's been working day and night on this for weeks 11:32:44 OdiliTime he's got more stamina than I would 11:32:44 mona if it's causing you suffering might be a good time to give up 11:32:50 JB ok which changes did you just make, ulimit -n and which else? 11:33:30 @n-tech i reduced the number of processes back to the same number as the cores again 11:33:32 @n-tech that's it 11:33:33 OdiliTime don't give up, it's solvable. Just need to be broken down 11:33:44 @n-tech page render times are a little down but that's because the cores are opened again 11:33:56 @n-tech the main issue continues to be the mystery lag between request and render 11:34:51 OdiliTime do you think m serving static files slowly is weird? 11:34:52 JB i still think it's because there aren't enough php-fpm children running simultaneously :3 11:35:14 @n-tech jb, i upped that to 80 11:35:16 @n-tech i've had it on demand 11:35:17 @n-tech i've tried 11:35:17 @n-tech every 11:35:18 @n-tech imagianble 11:35:20 @n-tech configuration 11:35:22 JB top -bn 1 | grep "php-fpm" | wc -l 11:35:26 @n-tech dude I've tried 11:35:26 JB ^ that is up to 80 now aye? 11:35:27 @n-tech every 11:35:30 @n-tech yes 11:35:36 JB ok 11:35:37 @n-tech and it slowed it down 11:35:39 @n-tech so i went back to 8 11:35:41 JB oh fug 11:37:25 JB if php is bottlenecking no matter how many children you have available then i wonder if the database is the bottleneck 11:37:25 atc n-tech: getting file by hash is broken afaik 11:37:39 @n-tech okay dude it has to wait 11:37:44 atc And also can't install next due to sql errors 11:37:49 @n-tech update 11:37:51 @n-tech i fixed that 11:37:56 @n-tech but the rest of the shit has to fucking wait 11:38:01 atc Cool thanks 11:38:04 atc Keep it up 11:38:17 OdiliTime the queries are fine 11:38:33 OdiliTime I wonder if it's connections waiting for an open spot 11:38:52 OdiliTime but he said nginx on m.8ch.net doesn't connect to postgres for each connection 11:38:59 OdiliTime er each request 11:39:02 @n-tech jb i'm only getting 18 11:39:04 @n-tech 16* 11:39:08 JB 80 child php-fpm processes each handling requests in 490ms on average should be able to sustain 39 hits per second 11:39:11 @n-tech even though I set it to 80 11:39:21 @n-tech maybe it's an issue with the command 11:39:28 OdiliTime command? 11:39:30 @n-tech find another way to count processes 11:39:32 @n-tech to count 11:39:41 OdiliTime ps auxw|grep -v grep|grep fpm|wc -l 11:40:01 JB when you edited fastcgi_params to add my suggested params, did you ever take them back out 11:40:07 @n-tech odili, 81 11:40:22 @n-tech no, but it makes no difference in time either way 11:40:27 @n-tech this is a problem of night and day 11:40:30 JB the possibility crossed my mind that maybe it is still in that file and it was set to 16 instead of 160 by a typo or something 11:40:37 JB since i suggested 160 11:40:40 JB sorry if im totaly wrong 11:40:41 JB totally* 11:40:42 @n-tech okay, i'll do 160 11:40:49 @n-tech but keep in mind 11:40:54 @n-tech each process takes 512mb of ram 11:40:57 @n-tech and we get less than 20 requests per second 11:41:04 OdiliTime and once it's swaps it toast 11:41:06 @n-tech yes 11:41:13 @n-tech the PM config is 100% fine 11:41:17 @n-tech there is nothing wrong with it at all