Duplicating or forking incoming requests

I’m trying to figure out how to accomplish something with nginx and it
seems to have me baffled.

We use nginx as an endpoint to log incoming events. The default
response is a 200, empty body and an entry in the access log in a very
specific format. For legacy reasons, this will not change any time
soon.

Now, in addition to just simply logging this request, we want to forward
that request to an internal server to also do something with it (and
again, for legacy reasons I can’t just make that server do the logging
instead).

I can get this working fine, except for failure modes I specifically do
not want the original nginx logging affected at all if it can’t proxy
the request upstream, and in point of fact specifically want to return a
200 in that case also. So basically the desired behavior looks like
this:

nginx running
proxy running
log a successful incoming GET request to nginx access log
forward request to proxy
return 200

nginx running
proxy not running (or returns an error)
log a successful incoming GET request to nginx access log
forward request to proxy
proxy returns an error
nginx ignores the error and returns a 200

I haven’t been able to get to that point, but instead have been only
able to get various combinations of “nginx doesn’t log anything”, “nginx
logs everything only when the entire chain is up” and “nginx logs
incoming requests fine when chain is up but logs to the error log when
the proxy is down”.

Here’s the configuration snippets I’m currently using:

location = /events {
access_log events_access.log events_format;
expires 1s;
try_files @proxy @proxy;
# try_files @proxy /empty.html;
# try_files @proxy =200;
}

location @proxy {
proxy_pass http://127.0.0.1:8887;
proxy_intercept_errors on;
access_log events_access.log events_format;
error_page 502 =200 /empty.html;
proxy_set_header X-Real-IP $remote_addr;
}

Basically, what I want is that if it logs every incoming request
normally. If it can forward the request to the upstream proxy, it does
so after logging it, and if it can’t, it simply logs it and returns a
200.

Is this possible and if so how?

Thanks in advance!

– Gary F.

Hello!

On Fri, Jul 12, 2013 at 12:47:02PM -0700, Gary Foster wrote:

[…]

Here’s the configuration snippets I’m currently using:

location = /events {
access_log events_access.log events_format;
expires 1s;
try_files @proxy @proxy;
# try_files @proxy /empty.html;
# try_files @proxy =200;

You are trying to use try_files incorrectly. Please read docs at
http://nginx.org/r/try_files.

Basically, what I want is that if it logs every incoming request
normally. If it can forward the request to the upstream proxy,
it does so after logging it, and if it can’t, it simply logs it
and returns a 200.

Is this possible and if so how?

I would recommend something like this:

location = /events {
    access_log events_access.log events_format;
    error_page 502 504 = /empty;
    proxy_pass ...;
}

location = /empty {
    access_log events_access.log events_format;
    return 200 "";
}

Note the above configuration snippet doesn’t try to intercept
errors returned by upstream servers, but only handles cases when
nginx can’t reach them and/or an invalid response is returned. If
upstream servers are expected to return various errors,
proxy_intercept_errors should be used, as well as additional codes
in error_page.


Maxim D.
http://nginx.org/en/donation.html

Thanks, that did the trick exactly!

Now that I have something that works, I’m off to the docs to figure out
where my fundamental misunderstandings were and correct them. Very much
appreciated!

– Gary F.