Can not scale service to multiple container when binding host port in docker-compose.yml

Question

Micro service is running on one container. I have bound that service to Host port 8082 and launched this micro service using docker-compose . Now I want to scale it to 3 container using docker-compose scale command but it giving me following error-

ERROR: for 2 failed to create endpoint composetest_nginx_2 on.network composetest_default: Bind for 0.0.0.0:8082 failed: port is already allocated

ERROR: for 3 failed to create endpoint composetest_nginx_3 on.network composetest_default: Bind for 0.0.0.0:8082 failed: port is already allocated

What would be the solution for this?

Answer 1

You are not supposed to bind 3 containers' ports to the same port in host.

What you can do is setting up:

• 1 load balancer container (bind container's port XXXX to host's port 8082)

• 3 service containers (expose ports to load balancer)

bind only binds 1 port to host. And expose only exposes ports to the linked container so they can be from multiple containers.

Reference:

https://www.brianchristner.io/how-to-scale-a-docker-container-with-docker-compose/

https://github.com/vegasbrianc/docker-compose-demo/blob/master/docker-compose.yml

Answer 2

I solved it by passing a port range, with the correct syntax :P

ports:
      - "9110-9120:3333"

Answer 3

if you are using Nginx or similar, then you can use docker-gen command to auto-update configurations in Nginx.

See this for more info: https://deployeveryday.com/2016/09/28/composing-docker-environments-scale.html

Answer 4

the way to do it is to remove the source port. for example, change

ports:
    - "5000:5000"

to

ports:
    - "5000"

the downside is you don't know the range of the port number until the service started.

Answer 5

Good option is to give range port rather giving specific port in docker-compose...

ports:
 - 8081+:8081

This way it will allow you to do scaling your service without getting into the issue you mentioned ( Port already allocated ) as Host port will different for each scaled instance & Docker will handle that for you...