Docker Network Recap

Recap on Docker’s Networking, mostly Bridge Network.

Recap on Docker’s Networking, mostly Bridge Network.


More than ever before, Docker is being adopted in both work (thank god) and personal project. It’s painless to spin up a disposable container for testing business logic involves PostgreSQL, WebDAV or anything. And it’s NINJA for deploying services for someone who doesn’t know a thing about computers.

I found myself constantly explaining docker network related things to others, such as how does it know which xxx instance to connect.

What I’ve Recapped

Bridge Network

So far the most commonly used network mode by me, I found it works perfectly for small to medium stack. Bridge can be further divided into default and user-defined networks.


All containers were born bridged. – Arisdocker

That is, if you don’t specify anything, this default bridge driver will be used.

Default bridge driver doesn’t provide functionalities like service discovery and DNS resolution. Which means they can’t talk to each other directly via container name, unless:

  1. they are linked via --link flag, this is being deprecated1 thus not recommended.
  2. the connection is established via IP address.


Unlike default, user-defined bridge network allows containers on the same to interpolate.

  1. Create a docker network named `my-net`
  2. docker network create my-net
  3. Create containers that connects to `my-net`
  4. docker run -dit --name=alpine1 \
    --net=my-net \
    alpine ash
    docker run -dit --name=alpine2 \
    --net=my-net --net-alias=a2 \
    alpine ash
  5. Create another container that doesn't connect to `my-net`
  6. docker run -dit --name=alpine3 \
    alpine ash

The above containers:

  1. alpine1 can reach alpine2 by ping alphine2 or ping a2.
  2. alpine2 can reach alpine1 by ping alphine1.
  3. alpine3 cannot reach anyone.

On top of that, containers within the same user-defined bridge network open up all ports to each other.

Host Network

Host network is as straightforward as the name suggests, all containers with a host network will have no network isolation, thus can be directly accessed via host_ip:container_port. Although no service discovery and DNS resolutionwill be provided, containers can use public ip address to communicate with each other.

docker run -dit --name=postgres \
--net=host \

--net=host allows the above postgres instance to be accessible via host_ip:5432.

Everything else (files, processes) will be isolated just like normal.