I have two pylons processes running on my devbox, one is my app in
production mode, one in development mode – I have
dev-server.mydomain.com and www.mydomain.com both aliased to the devbox.
Rather than have the nginx config duplicated once for each domain, I
have one config with an if statement in it. Thus, in my config file is
this:
if ($http_host ~ “(dev-server)”) {set $proxy_port 5000;} # dev mode =
port 5000
if ($http_host !~ “(dev-server)”) {set $proxy_port 5080;} # live mode =
port 5080
location / {proxy_pass http://127.0.0.1:$proxy_port;}
So far, so good.
But now, I want to make use of SSIs, so I stick “ssi on;” at the top of
this server {} block – trying it, it doesn’t work: when I request
/ssi-test which simply includes /include1 (“hello”) and /include2
(“world”), nginx actually just keeps fetching /ssi-test recursively and
inserting it into itself until I kill the backend process.
So why did I mention my proxy setup? Well, it seems that this bug only
appears when proxy_pass has a variable in it. If I proxy to a static
server, ie “proxy_pass http://127.0.0.1:5000;” (no variable), SSIs work
perfectly – but this means I have to edit the config file and reload
nginx to flick between dev and production views 
I have also tried removing the ifs and just having
set $proxy_port 5000;
location / {proxy_pass http://127.0.0.1:$proxy_port;}
but this doesn’t work either, which suggests that it is the variable
rather than the if statements that are the trigger.
Posted at Nginx Forum: