server_name and listen behaviour

[Ruslan Ermilov]

On Wed, Feb 13, 2013 at 08:56:56AM +0400, ivan babrou wrote:
> On 12 February 2013 23:14, Ruslan Ermilov <ru at nginx.com> wrote:
> 
> > On Tue, Feb 12, 2013 at 09:10:17PM +0400, ivan babrou wrote:
> > >    Hi, I have a question. It's better to describe with example
> > >    I point one.local and two.local to 127.0.0.1 and create following
> > servers
> > >    in config for nginx:
> > >    server {
> > >      listen 80;
> > >      server_name one.local;
> > >      location / {
> > >        return 404;
> > >      }
> > >    }
> > >    server {
> > >      listen two.local:80;
> > >      server_name two.local;
> > >      location / {
> > >        return 403;
> > >      }
> > >    }
> > >    If I request one.local then 403 is returned. If i change listen for
> > >    one.local to one.local:80 then 404 correctly returned, but only after
> > >    restart, reload doesn't help (that's probably bug too).
> > >    I expect to get 404 if i request one.local even if I listen on all
> > >    addresses. Am I right?
> >
> > 127.0.0.1:80 and *:80 are two different socket addresses,
> > and when you connect to 127.0.0.1:80, the first one will
> > be used (two.local in your case).  This is not specific
> > to nginx, it's how sockets work.
> >
> > nginx by default optimizes such a config by creating only
> > one listening socket [1], and pessimizes by using getsockaddr()
> > to differentiate between multiple sockaddrs.  This optimization
> > can be turned off by using "bind" parameter (or some other
> > parameters, see [1] for details).  Nevertheless, 127.0.0.1:80
> > and *:80 are still two different addresses, and 127.0.0.1:80
> > is more specific than *:80.
> 
> 
> See below, I only see one listening address in netstat.
> 
> As to "reload not working", I cannot reproduce this:
[...]
> I have listen *:80 before reload, then change it to listen one.local:80 and
> reaload. nginx/1.2.6
> 
> Before reload:
> 
> callisto ~ # netstat -ntl | fgrep ':80'
> tcp        0      0 0.0.0.0:80              0.0.0.0:*               LISTEN
> 
> After reload:
> 
> callisto ~ # netstat -ntl | fgrep ':80'
> tcp        0      0 0.0.0.0:80              0.0.0.0:*               LISTEN
> 
> After restart:
> 
> callisto ~ # netstat -ntl | fgrep ':80'
> tcp        0      0 127.0.0.1:80            0.0.0.0:*               LISTEN
> 
> Did you have one.local and two local when you check this?

Just for the record.  Here's how it works on BSD-like systems:

: $ ps 36455
:   PID   TT  STAT      TIME COMMAND
: 36455 s004  S+     0:00.00 nginx: master process objs/nginx -p . -c x.conf
: 
: $ cat x.conf
: [...]
: http {
:     server {
:         listen *:8000;
:         server_name one;
:         return 200 "$server_name\n";
:     }
:     server {
:         listen 127.0.0.1:8000;
:         server_name two;
:         return 200 "$server_name\n";
:     }
: }
: 
: $ echo 'GET http://one/' | nc 127.0.0.1 8000
: two
: 
: $ echo 'GET http://two/' | nc 127.0.0.1 8000
: two
: 
: $ netstat -anL | grep 8000
: 0/0/128        *.8000
: 
: $ vim x.conf
: [...]
: 
: $ cat x.conf
: [...]
: http {
:     server {
:         listen 127.0.0.1:8000;
:         server_name one;
:         return 200 "$server_name\n";
:     }
:     server {
:         listen 127.0.0.1:8000;
:         server_name two;
:         return 200 "$server_name\n";
:     }
: }
: 
: $ kill -HUP 36455
: 
: $ netstat -anL | grep 8000
: 0/0/128        127.0.0.1.8000
: $ echo 'GET http://one/' | nc 127.0.0.1 8000
: one
: $ echo 'GET http://two/' | nc 127.0.0.1 8000
: two

But as Igor already mentioned, trying this on Linux fails at
the "reload" step because Linux cannot create two sockets for
*:8000 and 127.0.0.1:8000:

2013/02/13 08:15:57 [notice] 17532#0: signal 1 (SIGHUP) received, reconfiguring
2013/02/13 08:15:57 [notice] 17532#0: reconfiguring
2013/02/13 08:15:57 [emerg] 17532#0: bind() to 127.0.0.1:8000 failed (98: Address already in use)
2013/02/13 08:15:57 [notice] 17532#0: try again to bind() after 500ms
2013/02/13 08:15:57 [emerg] 17532#0: bind() to 127.0.0.1:8000 failed (98: Address already in use)
2013/02/13 08:15:57 [notice] 17532#0: try again to bind() after 500ms
2013/02/13 08:15:57 [emerg] 17532#0: bind() to 127.0.0.1:8000 failed (98: Address already in use)
2013/02/13 08:15:57 [notice] 17532#0: try again to bind() after 500ms
2013/02/13 08:15:57 [emerg] 17532#0: bind() to 127.0.0.1:8000 failed (98: Address already in use)
2013/02/13 08:15:57 [notice] 17532#0: try again to bind() after 500ms
2013/02/13 08:15:57 [emerg] 17532#0: bind() to 127.0.0.1:8000 failed (98: Address already in use)
2013/02/13 08:15:57 [notice] 17532#0: try again to bind() after 500ms
2013/02/13 08:15:57 [emerg] 17532#0: still could not bind()

So nginx rolls back to previous configuration [1].

[1] http://nginx.org/libxslt/en/docs/control.html#reconfiguration