Reverse proxy service on manager node: “host not found in upstream” but it works if the same service is on worker node

问题

I am running docker swarm with 1 manager and 2 worker nodes (centos VMs). Each worker node has 1 stack that has an nginx and php service that load "This is site1" in the browser.

The site will only be loaded if the worker node has a reverse proxy (but it will always connect to the stack on that worker node which is part of the problem) but If I run the service on manager node that doesn't have the nginx/php stack it won't automatically direct to the stacks on worker nodes even though I have created an overlay "reprox" network that the stack and reverse proxy is connected to.

docker-compose.yml for the stack:

version: '3'

services:
  nginx:
      image: nginx:latest
      deploy:
        placement: 
          constraints: 
            - node.role==worker
        mode: global 
      volumes:
          - "./code:/code"
          - "./site.conf:/etc/nginx/conf.d/site.conf"
      networks:
          - site1
          - reprox
  php:
      image: php:7-fpm
      deploy:
        placement: 
          constraints: 
            - node.role==worker
        mode: global      
      volumes:
          - "./code:/code"
      networks:
        - site1

networks:
  reprox:
    external: true 
  site1:

docker-compose.yml for reverse-proxy (when I run this service on worker nodes it works by typing site1.local in the browser which is defined in my /etc/hosts as the IP address of the node):

version: '3'

services:
  nginx:
      image: nginx:latest
      deploy:
        placement: 
          constraints: 
            - node.role==manager
        mode: global
      ports:
          - "80:80"
      volumes:
          - "./conf.d/:/etc/nginx/conf.d"

      networks:
          - reprox
networks:
  reprox:
    external: true

conf.d/site1.conf configuration for reverse proxy:

    server {
    listen 80;
        server_name site1.local;

        location / {
            proxy_pass http://site1_nginx;
            proxy_set_header Host $host;
            proxy_set_header X-Real-IP $remote_addr;
            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
            proxy_set_header X-Forwarded-Proto $scheme;
        }
    }

also site.conf for the stack:

server {
    listen *:80;
    index index.php index.html;
    server_name site1.local;
    resolver 127.0.0.11;

    error_log  /var/log/nginx/error.log;
    access_log /var/log/nginx/access.log;
    root /code;

    location / {
        try_files $uri $uri/ =404;
    }

    location ~ \.php$ {
        set $upstream php:9000;        
        try_files $uri =404;
        fastcgi_split_path_info ^(.+\.php)(/.+)$;
        fastcgi_pass $upstream;
        fastcgi_index index.php;
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
        fastcgi_param PATH_INFO $fastcgi_path_info;
    }
}

I get this error when running reverse proxy on manager:

nginx: [emerg] host not found in upstream "site1_nginx" in /etc/nginx/conf.d/site1.conf:6

I need this so I can load balance the stacks between nodes, I will also add other stacks that will load different sites, that's why I need reverse-proxy.

I know I can add a separate load balancer on each node that will redirect to reverse proxies on these nodes, but that's too much of an overhead. The reverse-proxy should automatically connect to available stacks in the swarm, that's why I created an overlay network.

来源：https://stackoverflow.com/questions/58393434/reverse-proxy-service-on-manager-node-host-not-found-in-upstream-but-it-works

标签

Docker

nginx

networking

load-balancing

reverse-proxy