Totally stuck now.

This is what I have. I build and run a simple HTML page in nginx on port 8080 (I eventually want to run a Java tomcat app when I've figured this out).

This is on Windows 10 with Docker 17.09 and nginx 1.13

This is the config:

$ cat Dockerfile
FROM nginx
COPY nginx.conf /etc/nginx
COPY static-html /usr/share/nginx/html

and the config:

$ cat nginx.conf

user  nginx;
worker_processes  1;

error_log  /dev/stdout warn;
pid        /var/run/nginx.pid;

events {
    worker_connections  1024;

http {
    include       /etc/nginx/mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';

    access_log  /dev/stdout  main;

    sendfile        on;
    #tcp_nopush     on;

    keepalive_timeout  65;

    #gzip  on;

    include /etc/nginx/conf.d/*.conf;

Run it like this:

$ docker run -p 8080:80 --name dummy --rm dummy-nginx

Responds nicely to curl:

$ curl -i http://localhost:8080
HTTP/1.1 200 OK
Server: nginx/1.13.7
Date: Thu, 30 Nov 2017 16:46:17 GMT
Content-Type: text/html
Content-Length: 101
Last-Modified: Wed, 29 Nov 2017 16:24:39 GMT
Connection: keep-alive
ETag: "5a1edf47-65"
Accept-Ranges: bytes

  <title>oh yeah</title>
  <p>did this work?</p>

So that all works fine. This is the same stuff for the reverse proxy which is the part that fails:

Reverse proxy Dockerfile

FROM nginx
COPY nginx.conf /etc/nginx/nginx.conf
COPY apt-proxy.conf /etc/apt/apt.conf.d
COPY nginx_signing.key /etc/apt
RUN apt-get update
RUN apt-get -y install apt-utils
RUN apt-get -y install gnupg gnupg2 gnupg1
RUN apt-key add /etc/apt/nginx_signing.key
RUN apt-get -y install curl
RUN apt-get -y install net-tools
EXPOSE 10040
CMD ["nginx"]

(I put those extra tools in there to try to help diagnose this)

Reverse proxy nginx.conf

daemon off;
user  nginx;
worker_processes  1;
pid        /var/run/nginx.pid;
error_log /dev/stdout debug;
events { worker_connections 1024; }

http {
    access_log /dev/stdout;

    upstream myapp {
        server localhost:8080;

    server {
        listen 10040;

        location / {
            proxy_pass $scheme://myapp;
            proxy_redirect default;
            proxy_set_header HOST $host;
            proxy_set_header Referer $http_referer;
            proxy_set_header X-Forwarded-For $remote_addr;
            proxy_set_header X-Forwarded-Proto $scheme;

I heavily suspect the whole problem is just a minor error in there somewhere (banging head against wall).

Command line to start up the reverse proxy

This includes a bit of the logging with the error from nginx, and the access_log item which I also piped to /dev/stdout.

$ docker run -p 10040:10040 --name rproxy --rm we1p202420008.cloud.registry(example).com:11095/rproxy
2017/11/30 16:55:20 [notice] 1#1: using the "epoll" event method
2017/11/30 16:55:20 [notice] 1#1: nginx/1.13.7
2017/11/30 16:55:20 [notice] 1#1: built by gcc 6.3.0 20170516 (Debian 6.3.0-18)
2017/11/30 16:55:20 [notice] 1#1: OS: Linux 4.9.49-moby
2017/11/30 16:55:20 [notice] 1#1: getrlimit(RLIMIT_NOFILE): 1048576:1048576
2017/11/30 16:55:20 [notice] 1#1: start worker processes
2017/11/30 16:55:20 [notice] 1#1: start worker process 5
2017/11/30 16:55:25 [error] 5#5: *1 connect() failed (111: Connection refused) while connecting to upstream, client:, server: , request: "GET / HTTP/1.1", upstream: "", host: "localhost:10040" - - [30/Nov/2017:16:55:25 +0000] "GET / HTTP/1.1" 502 173 "-" "curl/7.56.1"
2017/11/30 16:55:25 [info] 5#5: *1 client closed keepalive connection

That error is logged when I try to access the dummy website thro the r-proxy like this:

$ curl -i http://localhost:10040
HTTP/1.1 502 Bad Gateway
Server: nginx/1.13.7
Date: Thu, 30 Nov 2017 18:05:06 GMT
Content-Type: text/html
Content-Length: 173
Connection: keep-alive

<head><title>502 Bad Gateway</title></head>
<body bgcolor="white">
<center><h1>502 Bad Gateway</h1></center>

Now if there is nothing obvious in the config I've just posted and it really isn't just my stupidity, then there is some baffling stuff which I can't get any further with. For instance, when I run curl inside my reverse proxy instance:

docker exec rproxy curl -i http://localhost:8080

it gives me the corporate firewall's blocked content result:

The website http://localhost/ that you are trying to access is currently categorised as Uncategorized URLs< and has been deemed to be potentially unsafe or unsuitable for browsing.

  • I think the rproxy nginx is trying to connect to 8080 on itself, localhost, and not the hosts "localhost". It would have to proxy_pass to an address that reaches the other container. Check The IP address docker has assigned to the "dummy" container. – mjb2kmn Nov 30 '17 at 19:30
  • @mjb2kmn Agreed, the docker container will have an IP address of its own, the java app will be on the bridged address of the host (if I remember correctly). – Simon Greenwood Nov 30 '17 at 22:28
  • So I should put the dummy app's FQDN in the reverse proxy nginx config? – Adam Dec 1 '17 at 10:08
  • Not really sure I understand, unless this is actually really difficult testing it with curl http://localhost on the command line – Adam Dec 1 '17 at 12:02

Using bridge as the magic key word, I finally started getting the search engine results I needed.

The comment on the question pointed out the issue: nginx was just forwarding requests to localhost:8080 internally within its own container, rather than the host:ports where the other docker containers are found.

I checked around and lots of people are writing the FQDNs of their proxied servers into their nginx.conf, which I didn't want to since I figured it might get overcomplicated when the build populates the configuration files for each environment.

So from Accessing host machine from within docker container and Docker docs I learnt about docker network today and implemented the following to give my dev environment a unique and unchanging host ip.

In my nginx.conf for the reverse proxy, I have proxy_pass http://myapp:8080 where myapp will be the docker container name I give the proxied app.

Then I create the user-defined bridge network:

docker network create -d bridge myapp-net

I don't yet understand why the default docker bridge network won't work.

And then I launch the containers, attaching them to the myapp-net:

docker run -p 8080:8080 --network=myapp-net --name myapp --rm myregistry:11095/myapp

docker run -p 10040:443 --network=myapp-net --name rproxy --rm myregistry:11095/myapp-rproxy

and it works.

I'm working on my Windows workstation and I figured writing the real external host FQDN into nginx.conf would require a new environment definition for each workstation, but with myapp in the conf file instead, it removes the dependency on the host name or IP.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.