Once in a while, you may need your Docker host's IP address. For instance, you need to be able to connect to the host network from inside a Docker container to access your app or database running locally on the host. Debugging or reverse proxies running on your host are two additional example use-cases.
I'll show you how to easily make this work simultaneously for macOS, Windows, and Linux - because their docker networking settings differ.
Docker Networking on macOS and Windows vs. Linux
For macOS and Windows the following special DNS name can be used:
The host has a changing IP address (or none if you have no network access). From 18.03 onwards our recommendation is to connect to the special DNS name
host.docker.internal, which resolves to the internal IP address used by the host. This is for development purpose and will not work in a production environment outside of Docker Desktop for Mac/Windows.The gateway is also reachable as
gateway.docker.internal.Source: https://docs.docker.com/docker-for-mac/networking/#use-cases-and-workarounds
Source: https://docs.docker.com/docker-for-windows/networking/#use-cases-and-workarounds
On Docker for Linux, the IP address of the gateway between the Docker host and the bridge network is 172.17.0.1 if you are using default networking.
Do you see the problem already? They are different, so you cannot simply run docker-compose up -d and all operating systems behave the same. But I got you covered, there's an easy approach to make this work.
Setup docker-compose
I've seen some suggestions, like creating a Linux-specific config file docker-compose.override.yml (docs), but the solution a co-worker of mine came up with seems more elegant and less complex to me.
# docker-compose.yml
version: '3.7'
services:
app:
image: your-app:latest
ports:
- "8080:8080"
environment:
DB_UPSTREAM: http://${DOCKER_GATEWAY_HOST:-host.docker.internal}:3000
So, what is happening here? The DB_UPSTREAM should point to the host's IP and port 3000. ${DOCKER_GATEWAY_HOST:-host.docker.internal} is the critical piece here. This expression tells docker-compose to either us the environment variable DOCKER_GATEWAY_HOST or use the fallback host.docker.internal when resolving this value.
On both macOS and Windows it works out of the box now without anything left to do. 🚀
If you are running this stack on Linux you need to have the DOCKER_GATEWAY_HOST environment variable set for the Docker gateway host. Simply put this line into your .bashrc (.bash_profile or .zshrc):
export DOCKER_GATEWAY_HOST=172.17.0.1
Now you can start the stack from macOS, Windows, and Linux without further configuration or overwrites. If you stick to this pattern - as we do - this works for every project of your company.
Great, isn't it? I hope this saves you some time!
References
-
compose-file#extra_hosts: an entry with the ip address and hostname is created in
/etc/hostsinside containers at build-time.
>> Let's connect on Twitter 🐦: https://twitter.com/natterstefan <<
Top comments (17)
Another alternative is to add the entry below to your Linux host's
/etc/hosts:Then you can simple define the environments as:
Is 172.17.0.1 host ip?
If it is means every time host's ip address changes for some reason maybe network switching to WIFI to LAN we need to update this and rerun our compose?
Running WSL (Ubuntu 20.04 LTS), I added this to my .bashrc file.
This runs
/sbin/ip route, searches for the line containing "dev eth0 proto kernel", then grabs the 9th element from that line which corresponds to the IP address.Then, in docker-compose.yml I referenced it here,
In case you are dealing with several Docker bridge networks, it's important to get the gateway IP of the correct network:
I tried this method:
$ export DOCKER_GATEWAY_HOST=172.17.0.1
put a new env docker container:HOST_IP=${DOCKER_GATEWAY_HOST:-host.docker.internal}
When the docker runs, it shows HOST_IP=host.docker.internal
The same is picked up by my python code.
Does this (
host.docker.internal) also needs to be resolved to get the actual host IP?Thank you, you give me new insight. I do a lot research about this too and write on my gist github. problem connect container to host
if you need the host.docker.internal (for example, nginx proxy to your host network), use an entrypoint script (IE: ENTRYPOINT['/entrypoint.sh'])
" install iputils-ping & iproute2 (or like packages)
" check if the host name resolves,
" if not, then get the gateway of the current network (which is the host)
" and set it in the containers host file
!/bin/sh
HOST_DOMAIN="host.docker.internal"
ping -q -c1 $HOST_DOMAIN > /dev/null 2>&1
if [ $? -ne 0 ]; then
HOST_IP=$(ip route | awk 'NR==1 {print $3}')
echo -e "$HOST_IP\t$HOST_DOMAIN" >> /etc/hosts
fi
Cheers ;-)
Nice trick!! I I was searching for a solution like this since I moved to an Ubuntu machine at work. Thanks a ton! 👍
Hi Jeremie,
Thanks for the feedback. I’m happy I could help you. ✌️
to automatize the iP (only get the first)
Thanks for the feedback!
This is broken (at least on MacOS). I don’t get the host’s IP address. I get 192.168.65.3 for host.docker.internal. Am I missing something?
As I understand 'host.docker.internal' inside docker default network will be bonded to your host machine. If you can access with 'host.docker.internal' your host ports then everything is ok
Great tip, way simpler and works out of the box without any need for AWK/SED.
Nice! This could definitely come in handy!
Thanks Niall. It does - resolved some issues for our team.
this could be used it in linux for production?