Compose: docker-compose up -d doesn't expose ports when defined with build directive

[akadoya]

docker-compose up -d is supposed to expose the ports and supposedly be able to publish the ports according to the yml, however, it is not working for the services build from build: configuration.

docker-compose.yml

version: '3.1'

services:
  nginx:
    build:
      context: "."
    ports:
      - "9999:80"
    network_mode: 'host'
  service_cassandra:
    image: 'cassandra:3.0'
    ports:
      - "19042:9042"

Dockerfile for nginx

FROM nginx
COPY nginx.conf /etc/nginx/nginx.conf
EXPOSE 80

CMD ["nginx", "-g", "daemon off;"]

These files above end up like this:

$ docker-compose up -d --build
Building nginx
Step 1/4 : FROM nginx
 ---> 46102226f2fd
Step 2/4 : COPY nginx.conf /etc/nginx/nginx.conf
 ---> Using cache
 ---> f36baa7f9388
Step 3/4 : EXPOSE 80
 ---> Running in 8b4a5621a0b3
 ---> 75fbceea828b
Removing intermediate container 8b4a5621a0b3
Step 4/4 : CMD nginx -g daemon off;
 ---> Running in 50f1efaf665d
 ---> e1cd7d703925
Removing intermediate container 50f1efaf665d
Successfully built e1cd7d703925
Successfully tagged tmp_nginx:latest
Recreating tmp_nginx_1
tmp_service_cassandra_1 is up-to-date
$ docker-compose ps
         Name                        Command               State                                Ports
----------------------------------------------------------------------------------------------------------------------------------
tmp_nginx_1               nginx -g daemon off;             Up
tmp_service_cassandra_1   /docker-entrypoint.sh cass ...   Up      7000/tcp, 7001/tcp, 7199/tcp, 0.0.0.0:19042->9042/tcp, 9160/tcp
$ netstat -lna |grep 19042
tcp6       0      0  ::1.19042                                     *.*                                           LISTEN
tcp4       0      0  *.19042                *.*                    LISTEN
$ netstat -lna |grep 9999
(no result)

$ docker ps
CONTAINER ID        IMAGE               COMMAND                  CREATED              STATUS              PORTS                                                        NAMES
43d7167e3671        tmp_nginx           "nginx -g 'daemon ..."   About a minute ago   Up About a minute                                                                tmp_nginx_1
a49374c38de3        cassandra:3.0       "/docker-entrypoin..."   6 minutes ago        Up 6 minutes        7000-7001/tcp, 7199/tcp, 9160/tcp, 0.0.0.0:19042->9042/tcp   tmp_service_cassandra_1

so..ports configuration seems working for containers that is built from image pull, but not for containers built from a local dockerfile.

[shin-]

Are you using the --service-ports flag?

[akadoya]

@shin-
I'm sorry I meant docker-compose up -d

[akadoya]

this probably is a better sample. using the same image and doing the same thing but ports setting under build for port 9999 is not published.

version: '3.1'

services:
  nginx_build:
    build:
      context: "."
    ports:
      - "9999:80"
    network_mode: 'host'
  nginx_image:
    image: 'nginx'
    volumes:
      - .nginx.conf:/etc/nginx.conf
    ports:
      - "19999:80"
===
$ docker-compose ps
      Name                Command          State           Ports
------------------------------------------------------------------------
tmp_nginx_build_1   nginx -g daemon off;   Up
tmp_nginx_image_1   nginx -g daemon off;   Up      0.0.0.0:19999->80/tcp
$ netstat -lna |grep 9999
tcp6       0      0  ::1.19999                                     *.*                                           LISTEN
tcp4       0      0  *.19999                *.*                    LISTEN

[akadoya]

tried with run --service-ports but no luck.

$ docker-compose run -d --service-ports nginx_build
tmp_nginx_build_run_1
$ docker-compose ps
        Name                  Command          State   Ports
------------------------------------------------------------
tmp_nginx_build_run_1   nginx -g daemon off;   Up

[shin-]

Port mapping is incompatible with network_mode: host:

The host network adds a container on the host’s network stack. As far as the network is concerned, there is no isolation between the host machine and the container. For instance, if you run a container that runs a web server on port 80 using host networking, the web server is available on port 80 of the host machine.

https://docs.docker.com/engine/userguide/networking/#default-networks

[akadoya]

AH!! great catch. totally missed it even when I created this test files.
Thank you so much!

[b4dnewz]

so what was the solution exactly?
I'm having the same problem and can't manage to make it working yet

[gaillota]

I'm having the same issue, and I'm not using the network_mode: host
```
// Dockerfile
EXPOSE 3000

// docker-compose.yml
ports:

• "80:3000"

$ docker-compose run --rm web yarn run dev
$ docker-compose ps

Name Command State Ports

ddjtb_web_run_1 yarn run dev Up 3000/tcp
````

[akadoya]

@b4dnewz you cannot use port-mapping when using host network_mode because it's designed to expose all ports as written in Dockerfile. (see details in the link @shin- mentioned above) = use other network mode.

@gaillota not sure about your issue but I suspect that it was run by a regular user? (port 80 needs system admins' privilege to open)

[gaillota]

@akadoya Thank you for you quick answer and sorry if my issue was not clear enough.
My problem is pretty simple, I have the configuration above in my Dockerfile and docker-compose.yml which basically map the port 80 of my host machine to the port 3000 of my container.
But when I run docker-compose run ..., the mapping is not set between the two ports, and I cannot access my app through port 80 ?

[akadoya]

@gaillota oh I see, port should be mapped as HOST:CONTAINER so you need to swap position of 3000/80 port numbers in your compose file.

[akadoya]

never mind... your file seems right..

[albertorm95]

I'm having a problem with the ports on the .yml file it seems to be not working, or maybe I'm doing something wrong.

Dockerfile

FROM node:8.9.1-alpine

USER node
RUN mkdir /home/node/.npm-global
ENV PATH=/home/node/.npm-global/bin:$PATH
ENV NPM_CONFIG_PREFIX=/home/node/.npm-global
RUN npm install -g @angular/cli

USER root
RUN npm install -g nodemon

docker-compose.yml

version: "3.2"

services:

  backend:
    build:
      context: .
      dockerfile: dev.Dockerfile
    volumes:
      - ".:/app"
    working_dir: "/app"
    ports:
      - "1234:443"
    command: "nodemon app.docker.js"

And when I inspect the container the Network output is the following:

"NetworkSettings": {
            "Bridge": "",
            "SandboxID": "68d492e7bf09dc7f8a625162f62f812cf56950391d2f7455359ee472257e4f94",
            "HairpinMode": false,
            "LinkLocalIPv6Address": "",
            "LinkLocalIPv6PrefixLen": 0,
            "Ports": {
                "443/tcp": null
            },
            "SandboxKey": "/var/run/docker/netns/68d492e7bf09",
            "SecondaryIPAddresses": null,
            "SecondaryIPv6Addresses": null,
            "EndpointID": "79456da25fb3baecc07105c704f15dd0122902107636dbd41fdefa8721f61e0f",
            "Gateway": "172.17.0.1",
            "GlobalIPv6Address": "",
            "GlobalIPv6PrefixLen": 0,
            "IPAddress": "172.17.0.2",
            "IPPrefixLen": 16,
            "IPv6Gateway": "",
            "MacAddress": "02:42:ac:11:00:02",
            "Networks": {
                "bridge": {
                    "IPAMConfig": null,
                    "Links": null,
                    "Aliases": null,
                    "NetworkID": "8a0fb429170e347ffbc26afe7ac48d1c45e6eb77ddc02e4f595b0fc72855b2e4",
                    "EndpointID": "79456da25fb3baecc07105c704f15dd0122902107636dbd41fdefa8721f61e0f",
                    "Gateway": "172.17.0.1",
                    "IPAddress": "172.17.0.2",
                    "IPPrefixLen": 16,
                    "IPv6Gateway": "",
                    "GlobalIPv6Address": "",
                    "GlobalIPv6PrefixLen": 0,
                    "MacAddress": "02:42:ac:11:00:02",
                    "DriverOpts": null
                }
            }
        }

The only work around for this is running: docker-compose run -p 1234:443 backend

With that I got:

"NetworkSettings": {
            "Bridge": "",
            "SandboxID": "5a877c24a3f3487af570b09ca352804fccf459a5b677c62f573ba470944b60ea",
            "HairpinMode": false,
            "LinkLocalIPv6Address": "",
            "LinkLocalIPv6PrefixLen": 0,
            "Ports": {
                "443/tcp": [
                    {
                        "HostIp": "0.0.0.0",
                        "HostPort": "1234"
                    }
                ]
            },
            "SandboxKey": "/var/run/docker/netns/5a877c24a3f3",
            "SecondaryIPAddresses": null,
            "SecondaryIPv6Addresses": null,
            "EndpointID": "",
            "Gateway": "",
            "GlobalIPv6Address": "",
            "GlobalIPv6PrefixLen": 0,
            "IPAddress": "",
            "IPPrefixLen": 0,
            "IPv6Gateway": "",
            "MacAddress": "",
            "Networks": {
                "ecfgcloudapp_default": {
                    "IPAMConfig": null,
                    "Links": null,
                    "Aliases": [
                        "e919a70179d7"
                    ],
                    "NetworkID": "93325bd6ac1e86c9ed61c3f03eb6ccc6a13927322ff6c8337de96395c6bec941",
                    "EndpointID": "a3b8ad12664ded6e8e0c7b1f725b045b63c18c4c9544fba48c6c4d460a51cfb0",
                    "Gateway": "172.19.0.1",
                    "IPAddress": "172.19.0.2",
                    "IPPrefixLen": 16,
                    "IPv6Gateway": "",
                    "GlobalIPv6Address": "",
                    "GlobalIPv6PrefixLen": 0,
                    "MacAddress": "02:42:ac:13:00:02",
                    "DriverOpts": null
                }
            }
        }

Any idea?

[shin-]

docker-compose version?

[albertorm95]

docker version

Client:
 Version:      17.09.0-ce
 API version:  1.32
 Go version:   go1.8.3
 Git commit:   afdb6d4
 Built:        Tue Sep 26 22:40:09 2017
 OS/Arch:      darwin/amd64

Server:
 Version:      17.09.0-ce
 API version:  1.32 (minimum version 1.12)
 Go version:   go1.8.3
 Git commit:   afdb6d4
 Built:        Tue Sep 26 22:45:38 2017
 OS/Arch:      linux/amd64
 Experimental: true

docker-compose version

docker-compose version 1.16.1, build 6d1ac21
docker-py version: 2.5.1
CPython version: 2.7.12
OpenSSL version: OpenSSL 1.0.2j  26 Sep 2016

[shin-]

oh, you didn't specify, but I'm assuming you're using run instead of up? If so, you need --service-ports to publish the ports from the service definition. See our docs

[albertorm95]

What's the error?

2018-02-27 6:36 GMT-06:00 Jáchym Toušek [e-mail]:

I'm experiencing the same problem even when using docker-compose up. Any
ideas how to solve it?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/docker/compose/issues/4799#issuecomment-368859793,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AIaBKnXIPaz39XZS3ba5YZ8QRn9LAm7Zks5tY_bigaJpZM4NSip0
.

--
Alberto Manuel Rojas Mendez

[sureshamk]

NOTE : The host networking driver only works on Linux hosts, and is not supported on Docker for Mac, Docker for Windows, or Docker EE for Windows Server.

[LotfiKobrosly]

Hi! I am having same issues: I want to send RPC commands using curl from the docker command line to the containers but it returns a timeout error.
Note: I created a test network for nano (a.k.a raiblocks) using docker-compose, I followed these instructions for Ubuntu 16.04 (I created a container from ubuntu image), changed the cmake environment variables following these instructions for the CMake variables and for testing RaiBlocks (see bottom of the page) then committed those changes to a new image "ubuntu" with tag "nano_test_2".

Here is my docker-compose.yml:

version: "2"
services:

  peer1:
    image: ubuntu:nano_test_2
    volumes:
      - /root:/root/nano_test
    ports:
      - "1001:7075/udp"
      - "1002:7076"
      - "54000"
    networks:
      vpcbr:
        ipv4_address: 172.20.0.1
    command: bash -c  "cd /root/rai_build && ./rai_node --daemon"

  peer2:
    image: ubuntu:nano_test_2
    volumes:
      - /root:/root/nano_test
    ports:
      - "1003:7075/udp"
      - "1004:7076"
      - "54000"
    networks:
      vpcbr:
        ipv4_address: 172.20.0.2
    command: bash -c  "cd /root/rai_build && ./rai_node --daemon"

  peer3:
    image: ubuntu:nano_test_2
    volumes:
      - /root:/root/nano_test
    ports:
      - "1005:7075/udp"
      - "1006:7076"
      - "54000"
    networks:
      vpcbr:
        ipv4_address: 172.20.0.3
    command: bash -c  "cd /root/rai_build && ./rai_node --daemon"

networks:
  vpcbr:
    driver: bridge
    ipam:
      config:
        - subnet: 172.20.0.0/16
          gateway: 172.20.0.21

docker-compose version:

docker-compose 
version 1.20.1, build 5d8c71b
docker-py version: 3.1.4
CPython version: 3.6.4
OpenSSL version: OpenSSL 1.0.1t  3 May 2016

I execute docker-compose up and have the nodes running (although sometimes for some reason, one or two of them (never all) stop working, I get them to start running again docker-compose start peer2 and I don't lose anything in the db).
Can anyone help me please?

[robert197]

try running docker-machine inspect default | grep IPAddress and then use this address instead of localhost.

[Kelbie]

@robert197 Thanks for that! I was really confused why my docker container wasn't working. Do you know how to get docker to use localhost instead of defaulting to address when running docker-machine inspect default | grep IPAddress?

[mattnedrich]

@KevinKelbie I think you can achieve that by using Docker Desktop on Mac instead of Docker Machine: https://docs.docker.com/docker-for-mac/docker-toolbox/

[Nonary]

I'm going to post this here in case anyone else ran into the same issue I did.

I discovered that if you use a dockerfile that functions like a builder script (containing the AS after from) like this example:

FROM node as frontend
WORKDIR /src
COPY ["package.json", "package.json"]
RUN  yarn install
COPY [".", "."]
CMD [ "cd src && yarn serve" ]

Docker Compose will not expose the ports properly, or name the container after the service either.
I wasted a lot of time trying to figure out what was wrong, as soon as I removed the "as frontend" everything worked perfectly.

I ran into this issue because I just started learning docker compose and I copied some code from another builder script I made... not catching the fact I forgot to remove the stage.

[chinmayj817]

I am facing the same problem my docker-compose is not exposing ports.
```version: "3.7"

services:
web:
build: ./src
command: uvicorn app.main:app --reload --workers 1 --host 0.0.0.0 --port 8000
volumes:
- ./src/:/usr/src/app/
ports:
- 8002:8000
`` The command I am using isdocker-compose up -d --build`
Can any one help me?
Thank You

[kmursk]

I've found that docker-compose doesn't do port mapping if custom nework only is used. For example,

ports:
  - "8080:8080"
networks:
 some_my_network_name:
  internal: true
  driver: nat
services:
 some_my_service_name
  networks: 
   some_my_network_name:
    aliases:
     - some_my_alias

This service is unavailable from the host machine. But add network used by all containers, started with docker run, and port 8080 will be mapped. I use docker-compose version 1.25.2 for windows and this network is named _default_, though there's no network with such name in docker network ls. I've found it with inspect to typical (i mean started by docker run) container. So make such change

services:
 some_my_service_name
  networks: 
   some_my_network_name:
    aliases:
     - some_my_alias
   default:

Then docker ps looks better
STATUS PORTS NAMES
Up 14 seconds 0.0.0.0:8080->8080/tcp some_my_service_name

[khskarl]

@kmursk That was it for me, thanks :tada:

Shouldn't docker at least print a WARNING: whenever it ignores a field? Implicit behaviors are rather time consuming to hunt down.

[sepiropht]

version: "3"
services:
  node_backend:
    container_name: "node_backend"
    build: ./node_backend/
    ports:
      - "3001:3001"

  pg:
    image: "postgres:12"
    container_name: "postgres"
    ports:
      - "5432:5432"
    volumes:
      - ./pg/db_data:/var/lib/postgresql/data

the port 3001 is not exposed, why ?
And the command docker-compose up only start the node container not the postgresl container

[kmursk]

https://github.com/docker/compose/issues/4799#issuecomment-608142032
Checked your case with docker desktop for Win and docker for linux - it works with some changes

version: "3"
services:
  node_backend:
    container_name: "node_backend"
#    build: ./node_backend/  _Here's some script, which i can't explore, therefore i have used ready webapp image._
    image: tomee:11-jre-8.0.1-plus
    ports:
      - "3001:8080"

  pg:
    environment:
     **POSTGRES_PASSWORD: password**
    image: "postgres:12"
    container_name: "postgres"
    ports:
      - "5432:5432"
    volumes:
      - ./pg/db_data:/var/lib/postgresql/data

Start and look at container's status. It should be something like that:

> docker ps -a        
IMAGE                     STATUS              PORTS                    NAMES
tomee:11-jre-8.0.1-plus   Up 3 seconds        0.0.0.0:3001->8080/tcp   node_backend
postgres:12               Up 3 seconds        0.0.0.0:5432->5432/tcp   postgres

So i assume there can be 2 kind of troubles:

1. webapp has not been started or does not listen port 3001.
2. you had not supplied postgres with password. It's a mandatory variable for this image. Also verify that folder _./pg/db_data_ exists where docker-compose is launching.

[sepiropht]

Thanks for the quick reply @kmursk, i have an express-app that listen on port 3001, and the console show "running on port 3001" in the container
dockerfile of the node container
```FROM node:12
WORKDIR /usr/src/app/
COPY package*.json ./
COPY tslint.json ./
COPY tsconfig.json ./
RUN npm install
ADD src/ ./src
RUN npm start

EXPOSE 3001
CMD [ "node" , "dist/index.js" ]

And even with  that changes postgres container doesn't run. But if i run it alone it works. This stuff are very confusing !
```version: "3"
services:
  node_backend:
    container_name: "node_backend"
    build: ./node_backend/
    ports:
      - "3001:3001"

  pg:
    image: "postgres:12"
    container_name: "postgres"
    environment:
      - POSTGRES_USER=postgres
      - POSTGRES_PASSWORD=postgres
    ports:
      - "5432:5432"
    volumes:
      - ./pg/db_data:/var/lib/postgresql/data

the command that i use docker-compose up

[gooney47]

For me the problem was that my service, that I was running in the docker container, was hosted on 127.0.0.1 instead of 0.0.0.0 which makes the service inaccessible from the host system even if the port is published.

[bluebrown]

Are you using the --service-ports flag?

This flag is invalid on compose up. You can only use it on compose run.

[rakesh4real]

@gooney47 was right

use

docker-compose run -p 8002:8002  web python -m uvicorn main:app --reload --host "0.0.0.0" --port 8002

[harleylang]

@gooney47 if I could give you more emojis I would, thank you!!!!

[savemetenminutes]

The problem in my case was that I had specified a port on the host which was already in use by another process. I had the following port configuration:

      - "20001:9001"
      - "20022:22"
      - "20080:80"
      - "20800:8000"

There was a process listening on 20080 (PhpStorm's zend debugger broadcast settings listener) on the host. As far as I remember docker-compose used to complain when starting services having published ports which are already in use. Output of docker-compose --version:

docker-compose version 1.26.2, build eefe0d31

What's even weirder is that port 20022 does get published but the other free ports 20001 and 20800 do not. If I close PhpStorm or comment out the 20080 port all ports are published successfully.