Nginx代理（jwilder / nginx-proxy）由对等方重置连接（502错误的网关）

Question

I have simple docker-compose.yml where I would like to be able to use nginx as a proxy to the containers. For now I have two containers admin and api which later on I want to make talking to each other.

Right now with configuration presented below, when I try to access api.host.dev I'm getting this:

nginx-proxy     | nginx.1    | 2017/04/19 15:18:35 [error] 26#26: *1 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 192.168.60.1, server: api.host.dev, request: "GET / HTTP/1.1", upstream: "http://172.18.0.4:9000/", host: "api.host.dev"
nginx-proxy     | nginx.1    | api.host.dev 192.168.60.1 - - [19/Apr/2017:15:18:35 +0000] "GET / HTTP/1.1" 502 576 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.87 Safari/537.36"

Right now I'm kind of out of ideas. Here is all configuration:

version: '2'
services:
    nginx-proxy:
       image: jwilder/nginx-proxy
       container_name: nginx-proxy
       ports:
        - "80:80"
        - "443:443"
       volumes:
        - /var/run/docker.sock:/tmp/docker.sock:ro

    admin:
        container_name: admin
        image: php:7.1-fpm
        restart: on-failure
        volumes:
            - ../admin:/var/www/admin
        working_dir: /var/www
        env_file:
            - ./variables/dev-admin.env

    api:
        container_name: api
        image: php:7.1-fpm
        restart: on-failure
        volumes:
            - ../api:/var/www/api
        working_dir: /var/www
        env_file:
            - ./variables/dev-api.env

Content of *.env files:

dev-api.env:

APP_ENV=DEV
VIRTUAL_HOST=api.host.dev
VIRTUAL_PORT=9000

dev-admin.env:

APP_ENV=DEV
VIRTUAL_HOST=admin.host.dev
VIRTUAL_PORT=9000

Content of /etc/nginx/conf.d/default.conf :

# admin.host.dev
upstream admin.host.dev {
                                ## Can be connect with "env_default" network
                        # admin
                        server 172.18.0.3:9000;
}
server {
        server_name admin.host.dev;
        listen 80 ;
        access_log /var/log/nginx/access.log vhost;
        location / {
                proxy_pass http://admin.host.dev;
        }
}
# api.host.dev
upstream api.host.dev {
                                ## Can be connect with "env_default" network
                        # api
                        server 172.18.0.4:9000;
}
server {
        server_name api.host.dev;
        listen 80 ;
        access_log /var/log/nginx/access.log vhost;
        location / {
                proxy_pass http://api.host.dev;
        }
}

Full output of docker-compose up :

sudo docker-compose up --remove-orphans
Recreating admin
Recreating nginx-proxy
Recreating api
Attaching to admin, api, nginx-proxy
admin    | [19-Apr-2017 15:18:24] NOTICE: fpm is running, pid 1
admin    | [19-Apr-2017 15:18:24] NOTICE: ready to handle connections
api      | [19-Apr-2017 15:18:24] NOTICE: fpm is running, pid 1
api      | [19-Apr-2017 15:18:24] NOTICE: ready to handle connections
nginx-proxy     | forego     | starting dockergen.1 on port 5000
nginx-proxy     | forego     | starting nginx.1 on port 5100
nginx-proxy     | dockergen.1 | 2017/04/19 15:18:25 Generated '/etc/nginx/conf.d/default.conf' from 3 containers
nginx-proxy     | dockergen.1 | 2017/04/19 15:18:25 Running 'nginx -s reload'
nginx-proxy     | dockergen.1 | 2017/04/19 15:18:25 Watching docker events
nginx-proxy     | dockergen.1 | 2017/04/19 15:18:25 Contents of /etc/nginx/conf.d/default.conf did not change. Skipping notification 'nginx -s reload'
nginx-proxy     | nginx.1    | 2017/04/19 15:18:35 [error] 26#26: *1 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 192.168.60.1, server: api.host.dev, request: "GET / HTTP/1.1", upstream: "http://172.18.0.4:9000/", host: "api.host.dev"
nginx-proxy     | nginx.1    | api.host.dev 192.168.60.1 - - [19/Apr/2017:15:18:35 +0000] "GET / HTTP/1.1" 502 576 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.87 Safari/537.36"
nginx-proxy     | nginx.1    | 2017/04/19 15:18:45 [error] 26#26: *3 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 192.168.60.1, server: admin.host.dev, request: "GET / HTTP/1.1", upstream: "http://172.18.0.3:9000/", host: "admin.host.dev"
nginx-proxy     | nginx.1    | admin.host.dev 192.168.60.1 - - [19/Apr/2017:15:18:45 +0000] "GET / HTTP/1.1" 502 576 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.87 Safari/537.36"
^[[A^[[Anginx-proxy     | nginx.1    | 2017/04/19 15:24:47 [error] 26#26: *5 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 192.168.60.1, server: api.host.dev, request: "GET / HTTP/1.1", upstream: "http://172.18.0.4:9000/", host: "api.host.dev"
nginx-proxy     | nginx.1    | api.host.dev 192.168.60.1 - - [19/Apr/2017:15:24:47 +0000] "GET / HTTP/1.1" 502 576 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.87 Safari/537.36"

Answer 1

When using compose, each service is exposed to the other containers using its service name, as if it were a DNS hostname. So you want to change references to eg admin.host.dev to just admin . For example, use this:

# admin.host.dev
upstream admin.host.dev {
                                ## Can be connect with "env_default" network
                        # admin
                        server admin:9000;
}

Notice in the server statement it now uses the hostname admin . This is automatically resolved to the container IP of your admin container.

(But note I didn't change the upstream's name - that is an internal name for nginx, and you don't necessarily need to change it.)

You would want to change the server name of the other upstream as well.