pattaradanai/docker-compose-networking

A repository created to understand the Docker-Compose networking.

Docker Networking

Sample Services

This demo shows how to connect services run from different docker-compose files. The structure of each service includes three different folders representing three different docker-compose files under three different folders.

The important issue here is the fact that Docker networks created by docker-compose are created using the name of the folder in which the docker-compose.yml file lives. Default networks are created for the different docker-compose up commands so a shared network has been created and shared between the different servers.

Each server contains the following files:

• docker-compose.yml. The docker-compose file that exposes the services. Each service is exposed outside its corresponding Docker image at port 1500x where x is the number of the server.
• Dockerfile. A file that simply prepares the Docker image.
• server.py. A simple Flask web server that shows the number of the server. It runs it locally in port 500x where x is the number of the server.

Requirements

To deploy this PoC, check that you have docker and docker-compose installed in your computer.

$ docker --version
Docker version 18.09.0, build 4d60db4
$ docker-compose --version
docker-compose version 1.23.1, build b02f1306

Once done, clone the repository using git clone and cd into it.

$ git clone https://github.com/febrezo/docker-compose-networking
$ cd docker-compose-networking
$ tree
.
├── AUTHORS.md
├── COPYING
├── dir1
│   ├── docker-compose.yml
│   ├── Dockerfile
│   └── server.py
├── dir2
│   ├── docker-compose.yml
│   ├── Dockerfile
│   └── server.py
├── dir3
│   ├── docker-compose.yml
│   ├── Dockerfile
│   └── server.py
└── README.md

3 directories, 12 files

Deploying the Environment

After doing cd into the folder, build and start the first server.

$ cd dir1
$ docker-compose build
...
$ docker-compose up
Creating network "dir1_some-net" with driver "bridge"
Creating container1 ... done
Attaching to container1
container1 |  * Serving Flask app "server" (lazy loading)
container1 |  * Environment: production
container1 |    WARNING: Do not use the development server in a production environment.
container1 |    Use a production WSGI server instead.
container1 |  * Debug mode: off
container1 |  * Running on http://0.0.0.0:5001/ (Press CTRL+C to quit)

Afterwards, we are starting a second node.

$ cd ../dir2
$ docker-compose build
...
$ docker-compose up
Creating container2 ... done
Attaching to container2
container2 |  * Serving Flask app "server" (lazy loading)
container2 |  * Environment: production
container2 |    WARNING: Do not use the development server in a production environment.
container2 |    Use a production WSGI server instead.
container2 |  * Debug mode: off
container2 |  * Running on http://0.0.0.0:5002/ (Press CTRL+C to quit)

Finally, we are starting a third node. Note that the service in the dir3/docker-compose.yml is named server1 to force a collusion.

$ cd ../dir3
$ docker-compose build
...
$ docker-compose up
Creating container3 ... done
Attaching to container3
container3 |  * Serving Flask app "server" (lazy loading)
container3 |  * Environment: production
container3 |    WARNING: Do not use the development server in a production environment.
container3 |    Use a production WSGI server instead.
container3 |  * Debug mode: off
container3 |  * Running on http://0.0.0.0:5003/ (Press CTRL+C to quit)

From another terminal, we can check whether the service is reachable OUTSIDE the Docker container.

$ curl localhost:15001
Hello World from node 1!
$ curl localhost:15002
Hello World from node 2!
$ curl localhost:15003
Hello World from node 3!

Docker Networking

After creating the environment, we can check the different networks created using docker network ls. Note that the name of the network created in dir1/docker-compose.yml (in our case, some-net), is preceded by dir1_ which is the name of the folder of the docker-compose file that defined the network.

$ docker network ls
NETWORK ID          NAME                DRIVER              SCOPE
3abf72c9cd6d        bridge              bridge              local
0f1aee6f822a        dir1_some-net       bridge              local
1410fa738d4b        host                host                local
a642878f65d3        none                null                local

The aforementioned dir1_some-net network should be declared on each of the other docker-compose.yml files.

version: '3.5'

services:
  server2:
    ...
    networks:
      - dir1_some-net
networks:
  dir1_some-net:
    external: true

We can now login into either one of the previous machines. Note the names of the machines used (server1, server2 AND server1 again).

$ docker exec -it container1 bash
root@9453a4be6d60:/opt/app# curl server1:5001
Hello World from node 1!
root@9453a4be6d60:/opt/app# curl server2:5002
Hello World from node 2!
root@f2bb4d587a8b:/opt/app# curl server1:5003
Hello World from node 3!

We can get the same behaviour from other nodes:

$ docker exec -it container3 bash
root@428fba281b01:/opt/app# curl server1:5001
Hello World from node 1!
root@428fba281b01:/opt/app# curl server1:5002
Hello World from node 2!
root@428fba281b01:/opt/app# curl server1:5003
Hello World from node 3!

Lessons Learned

Some funny have been detected and are addressed below:

• As said before, the networks created use the name of the folder in which the docker-compose.yml file live. This can lead to some misunderstandings when deploying that may take too much time notice. Be advised.
• If the name of the different services included in different docker-compose.yml files are the same, Docker will transparently try to find alive services on different machines. For instance, if two servers are named myservice but one exposes a service in port 5000 and another one in port 6000, both services will be reachable INSIDE the Docker images using curl myservice:5000 and curl myservice:6000 in both machines. If the port is also de same, each machine will only be able to connect to the service exposed by itself.