nginx container unable to pick the container's /etc/hosts file

[princejha95]

I am running the nginx container and i want that the container should perform the DNS mapping by using the container's /etc/hosts file. But instead of that, the container is using the host's machine /etc/hosts file for DNS mapping.

Can anyone provide a solution for this ? I am using nginx v1.15.5

[KevinWMatthews]

Can you use a bind mount when you start the container? If you're using Docker commands from the console, it would be something like this:

$ docker run --rm --name <container_name> --mount type=bind,source="$(pwd)/hosts",target=/etc/hosts <image_name>

The key part is the --mount argument. Some details:

• type=bind will copy/overwrite a file from your host into the container
• source="$(pwd)/hosts" specifies where the file is on your PC. I'm assuming that your desired hosts file exists in the current directory
• target=/etc/hosts specifies where the file is put in the container

Once your container is running you can check the hosts file using:

$ docker exec -it <container_name> cat /etc/hosts

Of course, different solutions (Kubernetes, docker-compose) will have different syntax but the underlying concept will likely be similar.

If that solution isn't appropriate you could edit the Dockerfile so that it copies your hosts file into the image.

[tlc]

nginx does not check /etc/hosts for proxy_pass. The standard answer for (non-docker) nginx is to run dnsmasq, which will make /etc/hosts lookups part of a dns call. But this not a good solution for the docker image.

Also, the docker-compose.yml element 'extra_hosts' works by writing /etc/hosts. So 'extra_hosts' does not work for docker-nginx.

[silencej]

For docker-compose, using an extra dnsmasq container may be a solution.

[thresheek]

I'm not sure where the idea that /etc/hosts doesnt work for nginx comes from. nginx will use information from there when starting up, so maybe your /etc/hosts in the container isnt populated when nginx starts?

Anyhow, the following works fine here:

$ cat docker-compose.yml
nginx:
  image: 'nginx:1.19.7-alpine'
  volumes:
  - ./configs/proxy.conf:/etc/nginx/conf.d/default.conf:ro
  extra_hosts:
    - "somehost:127.0.0.1"
    - "otherhost:127.0.0.2"
  ports:
    - '80:80'

$ cat configs/proxy.conf
upstream backends {
        server somehost:8888;
        server otherhost:8888;
}

server {
    listen 80;
    location / {
        proxy_connect_timeout 1s;
        proxy_pass http://backends;
    }
}

server {
    listen 8888;
    location / { return 200 '$server_addr\n'; }
}

After docker-compose up -d:

$ curl http://127.0.0.1/asdfasdf  
127.0.0.1
$ curl http://127.0.0.1/asdfasdf
127.0.0.2

$ docker exec -ti myapp_nginx_1 cat /etc/hosts 
127.0.0.1       localhost
::1     localhost ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
127.0.0.1       somehost
127.0.0.2       otherhost
172.17.0.3      5395310df226

[maxim-avramenko]

princejha95 Nginx worker process working in isolated place from container operation system, it means that nginx worker process don't know anything about outside his process. For example worker process can not get environment variables from container where process working, worker process don't know address of DNS server that you set inside container. To fix your issue you need to initiate "resolver" property with ip of docker DNS server, after that you will resolve custom domains from your hosts file. you need just explain to nginx worker process where DNS server live and what is his address, you can use dnsmasq or Docker DNS server, or any other dns server, but you need to explain to worker process where he can find working DNS.