routing problem on apache

rubyonrails-talk
1

I've got an application that works fine when run on Brick, but which throws a routing problem when deployed to apache/fastcgi.

Both servers are running rails 1.2.2

The problem happens with simple links. I've got a link_to tag like this:

    <%= link_to "Categories", :controller => "/admin/factory/ category", :action => "list" %>

There is indeed a /admin/factory/category controller, and there is a 'list' action. The link works fine with Brick.

However, clicking on the link on the Apache production server gives this error in the production.log

==> current/log/production.log <==

Processing ApplicationController#index (for 24.46.207.128 at 2007-02-11 15:58:16) [GET]   Session ID: ec28effb4fea482174d1e3ccdb2bba08   Parameters: {}

ActionController::RoutingError (no route found to match "/admin/ factory/category/list" with {:method=>:get}):     /usr/local/lib/ruby/gems/1.8/gems/actionpack-1.13.2/lib/ action_controller/routing.rb:1292:in `recognize_path'     /usr/local/lib/ruby/gems/1.8/gems/actionpack-1.13.2/lib/ action_controller/routing.rb:1282:in `recognize'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/dispatcher.rb: 40:in `dispatch'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 168:in `process_request'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 143:in `process_each_request!'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 109:in `with_signal_handler'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 142:in `process_each_request!'     /usr/local/lib/ruby/site_ruby/1.8/fcgi.rb:600:in `each_cgi'     /usr/local/lib/ruby/site_ruby/1.8/fcgi.rb:597:in `each'     /usr/local/lib/ruby/site_ruby/1.8/fcgi.rb:597:in `each_cgi'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 141:in `process_each_request!'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 55:in `process!'     /usr/local/lib/ruby/gems/1.8/gems/rails-1.2.2/lib/fcgi_handler.rb: 25:in `process!'     /home/mrs/public/dispatch.fcgi:24

However, when I try to exercise the same route using console, also on the production server, there's no problem:

root@secure [/home/mrs/current]# ruby script/console Loading production environment.

rs = ActionController::Routing::Routes

=> #<ActionController::Routing::RouteSet:0xb7c22ad8 @routes=[#<ActionController::Routing::Route:0xb774441c

?> puts rs.routes ANY /fckeditor/check_spelling/ {:action=>"check_spelling", :controller=>"fckeditor"} ANY /fckeditor/command/ {:action=>"command", :controller=>"fckeditor"} ANY /fckeditor/upload/ {:action=>"upload", :controller=>"fckeditor"} ANY / {:action=>"list", :controller=>"admin/factory/item"} ANY /:controller/service.wsdl/ {:action=>"wsdl"} ANY /:controller/:action/:id/ {} => nil

rs.recognize_path "/admin/factory/category/list"

=> {:action=>"list", :controller=>"admin/factory/category"}

I've been struggling with this for a couple days, but can't figure it out. Any suggestions greatly appreciated.

Thanks: John

2

A little more information on this issue...

I woke up thinking... surely this is an orthogonal system, so the choice of webserver shouldn't influence the routing.

I had been testing the sofware on two different machines: a test machine running debian and Brick, and a production machine running RHL and Apache/FastCGI. Both machines had the same Ruby & Rails versions. However, I figured there must be some difference in the infrastructure that causes the routing error on the production machine. And it must be something lower down in the stack than the webserver, right?

To test this hypothosis, I fired up Brick on the production server. To my surprise, the routing worked fine!

So it looks like Apache is handing the Rails router something it can't handle properly.

Another bit of light: After seeing the above, I switched the rails version to 1.1.6 (from 1.2.2). This fixed the routing problem on Apache. Curious, I installed Rails 1.2.1 and tried running the app under that. 1.2.1 also had the routing problem.

So, it looks like there is an incompatibility between my install of Apache/Fastcgi and the Rails router in 1.2.1 and above. The app works fine with Brick on all versions.

Should I just fall back to 1.1.6? I've got another commercial appliation running on the same webserver with 1.1.6 with no problems. Is 1.2.x not ready for production?

Any ideas, much appreciated.

Thanks: John

3

I just got an email from someone else with this problem, asking if I've solved it.

Unfortunately, I was not able to find an answer. I 'solved' it by switching to Mongrel. Not a very satisfying solution, but I was burning up too much time (i.e. money) trying to get Apache/FastCGI working.