11 messages in ru.sysoev.nginxRe: worker_processes
FromSent OnAttachments
Šime RamovJan 29, 2007 9:07 am 
Aleksandar LazicJan 29, 2007 1:21 pm 
Šime RamovJan 29, 2007 2:32 pm 
Aleksandar LazicJan 29, 2007 6:05 pm 
Šime RamovJan 30, 2007 12:48 pm 
Aleksandar LazicJan 30, 2007 12:57 pm 
Igor SysoevJan 30, 2007 1:00 pm 
Šime RamovJan 30, 2007 2:34 pm 
Scott YangJan 30, 2007 2:49 pm 
Cliff WellsJan 30, 2007 3:11 pm 
Aleksandar LazicJan 30, 2007 11:48 pm 
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:Re: worker_processesActions...
From:Aleksandar Lazic (al-n@public.gmane.org)
Date:Jan 29, 2007 1:21:04 pm
List:ru.sysoev.nginx

Hi,

On Mon 29.01.2007 18:07, Šime Ramov wrote:

I installed nginx from Debian package and it came configured with worker_processes 1; and worker_connections 1024;

Ok.

Now, I was wondering (since on the english wiki it states that default value for worker_processes is 5), to which value to set this up in some common situations?

It is wrong, sorry :-/

The default is 1, the box is only a config example, I have corrected it on the wiki, thanks for the hint ;-)

For example, if I run one Django site via FastCGI, what value to put in worker_processes? And if I run two Django sites? Three Django sites? One Django site and a couple of static sites?

Hm this is a very difficult question, from my point of view.

I have use mostly this formula:

worker_processes = (max_assumed_concurrent_client * max_fcgi_instances) per site + 2 for the static one

Here for example:

(4 * 2) + 2 = 10 => worker_processes

But you should measure if you setup is the *best* for you application, because if you have *some*/*many* long running answer processes in the app-layer then you could need more worker_processes.

As I have said: it isn't a easy question because there are many, many opinions about this issue ;-)

But as far as I know the follwing are fact, for know:

1.) the proxy/fcgi connection to the backend aren't persistant 2.) with more worker_processes you need more resources 3.) If you setup more worker_processes then *physical* cpus you have some waits, maybe not a issue, but a fact, please correct me if I'am wrong! 4.) You *must* measure you applicaton a *easy* explanation is here: http://mongrel.rubyforge.org/docs/how_many_mongrels.html, yust think on django instead of mongrels ;-)

Hth

Aleks