kill -HUP failing (0.6.32)

[Guy Naor]

Hi,

I have an nginx setup as a reverse proxy into a large number of Rails and
Java applications. As it's a very dynamic system, nginx incudes a whole
directory of conf files, one for each application.

nginx works really well, but now that we have about 320 different config
files nginx stopped responding to kill -HUP for reloading the conf files. It
just ignores the request and nothing happens. My test shows that at 247
files it stop responding to HUP. Up to that number it's all ok and HUP works
perfectly. The files are all pretty much the same, with this format:


# Definition of the different compute instances we proxy into
upstream redmine_nginx {
  server 10.252.179.143:5150;
}

# vhost definition for this application
server {
  listen 80;
  client_max_body_size 10M;
  client_body_buffer_size 128k;
  server_name redmine.morphexchange.com ; # Add the redirected domains here
(like: www.gadgets.com that is redirected to gadgets.mex.com)
  set $original_uri $uri;
  access_log /mnt/nginx/logs/redmine/access.log main;
  error_log  /mnt/nginx/logs/redmine/error.log warn;
  location ~* ^.*/(themes|stylesheets|javascripts|images)(/.*$|$) {
    proxy_pass        http://redmine_nginx;
    proxy_set_header  X-Real-IP  $http_x_source_address;
    proxy_set_header Host $host;
  }
  location / {
    proxy_set_header  X-Real-IP  $http_x_source_address;
    proxy_intercept_errors  on;
    proxy_set_header Host $host;
    error_page  400 401 403 404 414 500 501 502 503 504 505 506 507 508 509
510  =  @rails_redmine;
    if ($request_method = POST ) {
      proxy_pass        http://redmine_nginx;
     break;
    }
    if ($is_args = "?" ) {
      proxy_pass        http://redmine_nginx;
     break;
    }

  }
  location @rails_redmine {
    rewrite (.*) $original_uri break;
    proxy_pass        http://redmine_nginx;
    proxy_set_header  X-Real-IP  $http_x_source_address;
    proxy_set_header Host $host;
   }
}

I also tried to increase the server hash size and bucket size but it didn't
help.

Looking at the code, there are a lot of conditions that might result in HUP
failing, but those are not logged. Any idea what might be causing this? The
same config with less files has no problem with the HUP signal.


Regards,

Guy Naor.

-- 
Guy Naor
CTO
Morph Labs, Inc.
guy at mor.ph
www.morphexchange.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://nginx.org/pipermail/nginx/attachments/20081010/7c6b14cb/attachment.html>