Apache 1 fastcgi config correct?


#1

I followed this tutorial on the rails wiki about how to get apache with
fastcgi working on a vps machine:

http://wiki.rubyonrails.com/rails/pages/HowtoInstallAndRunRubyOnRailsOnCpanel

In the tutorial it said to add this to the httpd.conf file to get
fastcgi
working.

FastCgiIpcDir /tmp/fcgi_ipc/ AddHandler fastcgi-script .fcgi

However in the rails book, it says to do:

FastCgiIpcDir /tmp/fcgi_ipc FastCgiServer /path/to/app/public/dispatch.fcgi -initial-env RAILS_ENV=production -processes 15 -idle-timeout 60

Right now i have my app up and running on the vps machine. It seems to
be
pretty responsive considering it is in development mode.

  • How can i tell if the fastcgi setup is correct?

I can see the obvious differences in the two versions of how to change
httpd.conf. However, i cant understand why they are different.

  • In the first example, how many fastcgi servers are running?
  • How would i change to production environment?

I’m not sure if i should use the version from the book because the
tutorial
was closer to how my vps is setup (cpanel on redhat).

Please help…


#2

I am also getting similar errors on fedora. Although, I have set the

/tmp/fcgi_ipc as owned by apache and with 755, the error_log still
cannot seem to create the dynamic servers.

not sure if there is some trick to get fastcgi running as the apache
user without a problem.

[Mon Mar 13 17:03:08 2006] [crit] (13)Permission denied: FastCGI:
can’t create (dynamic) server
“/var/www/html/test.rails/public/dispatch.fcgi”: bind() failed
[/tmp/fastcgi/dynamic/a452aa13feeda365028a2d43d1b2ce13]

In my case the difference between FastCgiServer and the AddHandler. It
seems
from the rails book that this difference is if you specify the number of
fastcgi processors that will startup or if apache will manage it for
you.

From the book, it is better to specify the number manually like you have.

However, if you are having the same permission errors as me, it will not
make any difference and I am at a loss what is needed to get it working.


#3

The problem for me turned out to be SELinux getting in the way. I was
able
to disable it and it started to work - hope that helps you.

http://www.crypt.gen.nz/selinux/faq.html#GA.7


#4

Manish S. wrote:

FastCgiIpcDir /tmp/fcgi_ipc/ AddHandler fastcgi-script .fcgi
FastCgiIpcDir /tmp/fcgi_ipc FastCgiServer /path/to/app/public/dispatch.fcgi -initial-env RAILS_ENV=production -processes 15 -idle-timeout 60
  • In the first example, how many fastcgi servers are running?
  • How would i change to production environment?

Hi Manish,

Here’s what (I think) I’ve gleaned from messing about with this stuff.

In the first example you leave it entirely up to Apache to start and
stop servers as it deems necessary (dynamic servers). In the second
example you have started exactly 15 processes and they run for the
lifetime of the apache server (static servers).

There’s also FastCGIServerExternal where you start the fcgi servers
yourself and just tell apache how to attach to them. (useful for the
spinner/spawner/reaper model with swit…I mean capistrano).

If you use ‘ps’ with the first example you’d see a variable number of
fcgi processes depending on load, with the second you’d see a fixed
number. The first is probably more efficient in terms of processor
resource and memory, the second in response times (apache doesn’t need
to decide about, and actually start any processes while serving a
request).

HTH

Alan


#5

Alan,

thats good stuff. However i cant seem to find any fastcgi processes
running. I used top and ps, but nothing looks like a fastcgi process.
What
should i be lookng for?

Also, if i stay with the first example, how can i set the RAILS_ENV to
production mode?

Thanks,
Manish


#6

Manish S. wrote:

Alan,

thats good stuff. However i cant seem to find any fastcgi processes
running. I used top and ps, but nothing looks like a fastcgi process.
What
should i be lookng for?

Also, if i stay with the first example, how can i set the RAILS_ENV to
production mode?

Thanks,
Manish

Manish,

Have you modifed the public/.htaccess to use fcgi instead of cgi ?

If not, it’ll all still work but apache would create a new instance
every request.

There’s a block of four lines near the bottom of the file and the last
one should have dispatch.fcgi instead of .cgi

The line you’d see in ps would be something like ‘ruby
/path/to/my/dispatch.fcgi’

A.


#7

Manish S. wrote:

if i stay with the first example, how can i set the RAILS_ENV to
production mode?

Also see the top 4 lines of config/environment.rb for a way of forcing
production, not sure if it’s the best way. I’ve been using Apache2.0
and External Servers on a virtual box I have root access to, so may not
be very useful for Apache1.3, dynamic servers and cPanel.

A.


#8

Alan F. wrote:

Manish S. wrote:

if i stay with the first example, how can i set the RAILS_ENV to
production mode?

Also see the top 4 lines of config/environment.rb for a way of forcing
production, not sure if it’s the best way. I’ve been using Apache2.0
and External Servers on a virtual box I have root access to, so may not
be very useful for Apache1.3, dynamic servers and cPanel.

A.

…or possibly…

“DefaultInitEnv RAILS_ENV production” in httpf.conf

…or possibly …

“SetEnv RAILS_ENV production” in public/.htaccess


#9

alan, i checked my .htaccess and i did switch it to use dispatch.fcgi.

I ran ps -A and saw 3 instances of dispatch.fcgi. So i guess it is
working. Because they are dynamic, does apache take care of restarting
them?

Now i just have to find the right way to set rails to be in production
environment.

Thanks for all your help.


#10

I have a feeling i have a bunch of zombie fcgi servers running. I see 8
when i run ps -A which should be wrong since i’m the only one using the
app
in the vps machine.

How can i tell if a fcgi proc is a zombie? Do i need to kill them all?
Also, whats the proper way to fully kill them?

Lastly, i think this is causing a problem for me in switching to
production
mode. I changed my httpd.conf file to be just like the Agile Rails
book. I
tried this twice

First Time: i made the change, and restarted httpd -> result:
application
error (rails) shown in the browser

Second Time: i made the change, kill -USR1 for each dispatch,fcgi
shown in ps -A, restarted httpd -> seemed to work, but after navigating
around a little i got Application Error (rails).

I’m not sure what to do.