Web Hosting Talk : Web Hosting Main Forums : VPS Hosting : VPS Tutorials : Centmin - a low memory CentOS install script

[Dregond Rahl]

Also, just found that my.conf is world writable by default, i'm pretty sure that can't be right, maybe change its permissions after MySQL is setup ?

[BTCentral - Ben]

Originally Posted by Dregond Rahl

Also, just found that my.conf is world writable by default, i'm pretty sure that can't be right, maybe change its permissions after MySQL is setup ?

It should not be, just checked on a server that we used the script on:

PHP Code:

[root@montreal ~]# ls -la /etc/ | grep my.cnf
-rw-r--r--  1 root root     454 Sep 16 11:13 my.cnf 


Edit: Just noticed you said my.conf - it should be my.cnf, not my.conf - are you sure you did not make a typo when editing things and create the file yourself?

[VPSForge-Ray]

Thanks for sharing this script Ill try it out later on my test server

[Dregond Rahl]

Originally Posted by BTCentral - Ben

It should not be, just checked on a server that we used the script on:

PHP Code:

[root@montreal ~]# ls -la /etc/ | grep my.cnf
-rw-r--r--  1 root root     454 Sep 16 11:13 my.cnf 


Edit: Just noticed you said my.conf - it should be my.cnf, not my.conf - are you sure you did not make a typo when editing things and create the file yourself?

Actually I made a typo here, sorry. But yes that file had 777 permission, may have been after I installed phpmyadmin. When I was restarting mysqld it gave a warning that its world writable so I checked it. I'll confirm it next time I run the script.

[barry[CoffeeSprout]]

Originally Posted by Dregond Rahl

I think it would, depending on what is being run, sites running Joomla and WordPress usually work well with Nginx and APC/xCache handling more requests and lowering memory usage per request. I know for sure that APC and xCache properly configured can improve page load times too, that's why with PHP6 they wanted to have APC integrated by default.

There are 100's of benchmarks proving it, but I don't think there has been any recent benchmarks with Nginx + PHP-FPM + xCache/APC on any low end system. I would do it myself if i knew how. haha T.T

Also speaking of WordPress, it seems it won't update, or get plugins unless it has FTP access, so currently trying to configure vsftpd which seems to love to eat memory.

That's generally because the webserver can't write to wordpress_install/wp-content/
Each php-fpm configuration needs to run as the user that owns the wordpress install.

edit: or you could of course give more privileges like 777, but I really recommend you run this as the proper user because it's way more secure in a multiuser setup.

[BTCentral - Ben]

Both PHP-FPM and Nginx run under a user/group created by the script - funnily enough, called nginx.

If the webserver can not write to the directory (as barry suggested - and it sounds like that to me too from what you described) then simply do a:

PHP Code:

chown -R nginx:nginx /home/nginx 


This will set the correct user/group on all the files/folders under it - and hopefully that should sort it out for you.

More updates coming soon, I worked on it a bit a few days back - but have not had the chance to test the changes. Will definitely spend some time on it over the weekend.

[Dregond Rahl]

Originally Posted by BTCentral - Ben

Both PHP-FPM and Nginx run under a user/group created by the script - funnily enough, called nginx.

If the webserver can not write to the directory (as barry suggested - and it sounds like that to me too) then simply do a:

PHP Code:

chown -R nginx:nginx /home/nginx 


Hopefully that should sort it out for you.

More updates coming soon, I worked on it a bit a few days back - but have not had the chance to test the changes. Will definitely spend some time on it over the weekend.

It still refuses, its something to do with WP3.0 and being more safer by allowing updates to happen by FTP/SFTP/FTPS i decided ill go with SFTP much easier.

I also did a benchmark of the Nginx setup using

PHP Code:

ab -kc 1000 -n 10000 http://www.domain.info/test.php 


Strangely I found many requests failed:
http://pastebin.com/0kqQq8jd

I tried static images too 'nginx-logo.png' and it had high failed requests too. I looked into the logs and found this error:

PHP Code:

2010/09/24 10:36:33 [crit] 28301#0: *3180 open() "/home/nginx/domains/domain.info/public/nginx-logo.png" failed (24: Too many open files), client: 199.19.224.168, server: www.domain.info, request: "GET /nginx-logo.png HTTP/1.0", host: "www.domain.info" 


I'm not sure what the problem is.

[BTCentral - Ben]

What version of Nginx are you using? Stable or Devel? Might be a bug (with Nginx).

[Dregond Rahl]

Originally Posted by BTCentral - Ben

What version of Nginx are you using? Stable or Devel? Might be a bug.

Stable:

[root@kuro ~]# nginx -v
nginx version: nginx/0.7.67

[BTCentral - Ben]

And are you testing it locally (as in running ab via SSH on your VPS), or on another machine? If the latter it may be connectivity issues between the two.

Just updated the kernel on my test VPS and am waiting for it to reboot, I'll try to do the same and see if I have any failures.

[Dregond Rahl]

Originally Posted by BTCentral - Ben

And are you testing it locally (as in running ab via SSH on your VPS), or on another machine? If the latter it may be connectivity issues between the two.

Just updated the kernel on my test VPS and am waiting for it to reboot, I'll try to do the same and see if I have any failures.

SSH(Putty) on my VPS from BuyVM

Edit: doesn't happen with .html files however.

[BTCentral - Ben]

Just tried it myself.

PHP Code:

Failed requests:        962
   (Connect: 0, Length: 962, Exceptions: 0) 


I found out why though:

PHP Code:

2010/09/24 12:13:59 [alert] 1539#0: *20413 socket() failed (24: Too many open files) while connecting to upstream, client: 127.0.0.1, server: localhost, request: "GET /test.php HTTP/1.0", upstream: "fastcgi://127.0.0.1:9000", host: "127.0.0.1" 


Note the: (24: Too many open files)

No failures for nginx-logo.png however, perhaps that was a memory constraint?

Edit: Ahha, the magic of Google provided a potential solution .

[Dregond Rahl]

Originally Posted by BTCentral - Ben

Just tried it myself.

PHP Code:

Failed requests:        962
   (Connect: 0, Length: 962, Exceptions: 0) 


I found out why though:

PHP Code:

2010/09/24 12:13:59 [alert] 1539#0: *20413 socket() failed (24: Too many open files) while connecting to upstream, client: 127.0.0.1, server: localhost, request: "GET /test.php HTTP/1.0", upstream: "fastcgi://127.0.0.1:9000", host: "127.0.0.1" 


Note the: (24: Too many open files)

No failures for nginx-logo.png however, perhaps that was a memory constraint?

Edit: Ahha, the magic of Google provided a potential solution .

Thanks! i'm trying out the solution right now, will update on if it works.

[Dregond Rahl]

Well so far:

PHP Code:

[root@kuro ~]# su - nginx
This account is currently not available. 


So i modified the /etc/security/limits.conf to:

PHP Code:

nginx       soft    nofile   10000
nginx       hard    nofile  30000 


When i rebooted the images were being served fine, then i tired PHP and the socket() error came, after that i tried image again and it started to give errors too. I'm going to reboot again and see if it will help.

[BTCentral - Ben]

Originally Posted by Dregond Rahl

PHP Code:

[root@kuro ~]# su - nginx 


su - nginx will not do anything, reason being that the nginx user is setup with the "/sbin/nologin" shell, therefore you can not use it to login .