I'd like that Varnish completely ignored certain URLs on my websites.
I.e., www.site1.com/fileuploader and www.site2.com/fileuploader
I have 4 websites on the same VCL, and all those websites are using the same URL structure so the only thing that changes is the domain itself, the URLs I want Varnish to ignore are the same.
I've tried:
sub vcl_recv {
if (req.url ~ "^/fileuploader/\?"
) {
return(pipe);
}
}
But it doesn't work. When trying to download a PDF through it I'm presented with a Varnish 503 Unavailable error (the same error I have if I didn't change the config). If I go through nginx directly using its port, it downloads fine.
Here's the varnishlog for this incident:
16 ObjHeader - Server: nginx/1.2.7
16 ObjHeader - Date: Wed, 17 Apr 2013 11:03:40 GMT
16 ObjHeader - Content-Type: application/pdf
16 ObjHeader - Content-Length: 1078550
16 ObjHeader - X-Powered-By: PHP/5.3.23-1~dotdeb.0
16 ObjHeader - Expires: Thu, 19 Nov 1981 08:52:00 GMT
16 ObjHeader - Set-Cookie: frontend=74go8sgckma5qrobnqc36pcjd4; expires=Wed, 17-Apr-2013 12:03:40 GMT; path=/; domain=www.site1.com; httponly
16 ObjHeader - Pragma: public
16 ObjHeader - Cache-Control: must-revalidate, post-check=0, pre-check=0
16 ObjHeader - Content-Disposition: attachment; filename=File-1338990998.pdf
16 ObjHeader - Content-Encoding: gzip
16 ObjHeader - Vary: Accept-Encoding
30 FetchError - Resource temporarily unavailable
16 FetchError - straight insufficient bytes
Best Answer
If you want to pass it absolutely untouched by Varnish, then the best option is
return(pipe);
, which basically tells Varnish to act as a dumb TCP proxy for the duration of the connection.But the fact that you got a
503
response seems to imply that something else is wrong - can you provide the output fromvarnishlog
when a request for that resource is made, if pipe mode doesn't do the trick?