Example docker-compose.yml not working (port mapping + loopback)

[btxtiger]

As mentioned in some earlier issues (#611, #493, #480), the given docker-compose.yml on hub.docker.com with port mapping 8080:80 does not work, because the port mapping itself breaks the loopback, resulting in two critical errors:

So it should be either added a necessary step to fix this issue with port mapping (which I am currently not aware of), or it should be changed to port 80:80, which currently solves the issue, mentioning that other port mappings currently will not work.

[wglambert]

I'm not able to reproduce

Using the docker-compose.yml from https://github.com/docker-library/docs/tree/master/wordpress#-via-docker-stack-deploy-or-docker-compose

[btxtiger]

I'm using MacOS 12.4, already disabled the firewall, tried different browsers.
Below is my compose file, just using a different DB. Using Network or Docker Volumes makes no difference.

version: '3.1'

services:
  mariadb:
      container_name: APP_mariadb
      image: mariadb:10.5
      command:
         - '--default-authentication-plugin=mysql_native_password'
         - '--character-set-server=utf8mb4'
         - '--collation-server=utf8mb4_unicode_ci'
         - '--skip-name-resolve'
      volumes:
         - ./db_volume:/var/lib/mysql
      cap_add:
         - SYS_NICE # CAP_SYS_NICE
      environment:
        MYSQL_INITDB_SKIP_TZINFO: '1'
        MYSQL_ALLOW_EMPTY_PASSWORD: 'false'
        ### Database initialization config:
        MARIADB_ROOT_PASSWORD: 'root123'
        MARIADB_USER: 'app'
        MARIADB_PASSWORD: 'app123'
        MARIADB_DATABASE: 'applights'
      ports:
         - 3380:3306
      restart: always
      # networks:
      #    - app_group

  wordpress:
      container_name: APP_wp_applights
      image: wordpress
      restart: always
      ports:
         - 8080:80
      environment:
         WORDPRESS_DB_HOST: mariadb
         WORDPRESS_DB_USER: 'app'
         WORDPRESS_DB_PASSWORD: 'app123'
         WORDPRESS_DB_NAME: 'applights'
      volumes:
         - ./wp_applights_volume:/var/www/html
         # - wp_applights_volume:/var/www/html
         - ./docker/config/php-custom.ini:/usr/local/etc/php/conf.d/custom.ini:cached,ro
      # networks:
      #    - app_group


# networks:
#    app_group:
#       external: false

# volumes:
#     wp_applights_volume:
#         driver: local
#     db_volume:
#         driver: local

[PPAGabriel]

Nice repo! 😄

[btxtiger]

I just installed Docker on Windows 11, and tried the compose file above with a fresh installation. Getting the same result.

Then just copied the plain docker-compose from Docker Hub, and started it on Windows 11. Same result:

[tianon]

WordPress is very particular about the URL that you use to access it (it stores it into the database so that it can use it for things later, including doing redirects if the incoming requests don't match what it thinks is the "canonical URL"), so if there's a hostname besides localhost you're planning to use after your instance is set up, I'd suggest using that right from the start.

[btxtiger]

@tianon Thanks, I am aware of that. However I am just doing some localhost testing/development.
localhost:80 works, but could not get any other port working yet. That makes multi-project setups a bit hard.
Do you have any idea what logs might provide further debug info?

[tianon]

I'm not sure, but it's definitely coming from https://github.com/WordPress/WordPress/blob/1d2e969b502da4b6353a6fec7a3ade8c6390b7d6/wp-admin/includes/class-wp-site-health.php#L2002-L2106 -- I'm not sure if there's a way to tell WordPress a different way to access itself internally vs the way it gets accessed externally, but using your laptop/system's external IP is one solution (if it's stable). I imagine something like 192.168.1.128:8080 would route correctly through all the layers (out and back again) where localhost can't really do so.

[wglambert]

Yeah I'm not sure how it's even succeeding on the Linux host since it can't curl localhost:8080 or 127.0.0.1:8080 and it's the same behavior on Docker Desktop. And on both of them localhost:80 works as expected

Because the errors are Failed to connect to localhost port 8080 which is only erroring on Docker Desktop

The REST API request failed due to an error.
Error: cURL error 7: Failed to connect to localhost port 8080: Connection refused (http_request_failed)

The loopback request to your site failed, this means features relying on them are not currently working as expected.
Error: cURL error 7: Failed to connect to localhost port 8080: Connection refused (http_request_failed)

[btxtiger]

Unfortunately I currently don't have a real Linux env to test this.
I did some tests using $base = 'host.docker.internal:host-gateway/index.php?'; in functions.php:1187, but that didn't help so far. But I guess the error is just a test related thing. The API should still work, just the test does not work properly on hosts with different ports. (A reverse proxy with a port other than 80 should also not work).

The test itself has some character of e2e tests, but being executed not at the real end.

So I come to these conclusions:

• A possible solution would be to map 80 to 8080 also inside the container (could not find a possibility so far)
• Or the test must be a http request from the frontend to the backend. In that case localhost:8080 would be the correct host.
• I was thinking about how this is handled in Laravel, there is the possibility to run internal requests behind the web server, so it just tests the route internally.

[MadmanXML]

I had the same issue.
I tested port 6666/7777 and 80; just port 80 work.

I used image is wordpress:latest and wordpress:php8.1-fpm-alpine.(Neither will work on ports 6666/7777)
My host os is centos 9 steam latest and docker version is:

[root@testsrv ~]# docker version
Client: Docker Engine - Community
 Version:           20.10.18
 API version:       1.41
 Go version:        go1.18.6
 Git commit:        b40c2f6
 Built:             Thu Sep  8 23:12:02 2022
 OS/Arch:           linux/amd64
 Context:           default
 Experimental:      true

Server: Docker Engine - Community
 Engine:
  Version:          20.10.18
  API version:      1.41 (minimum version 1.12)
  Go version:       go1.18.6
  Git commit:       e42327a
  Built:            Thu Sep  8 23:09:37 2022
  OS/Arch:          linux/amd64
  Experimental:     false
 containerd:
  Version:          1.6.8
  GitCommit:        9cd3357b7fd7218e4aec3eae239db1f68a5a6ec6
 runc:
  Version:          1.1.4
  GitCommit:        v1.1.4-0-g5fd4c4d
 docker-init:
  Version:          0.19.0
  GitCommit:        de40ad0
[root@testsrv ~]#

[visionm45]

I have had the same issue and I was not able to continue until I found a solution. From what I can tell this is only an issue if you are mapping the port to something other than 80. What I did was edit the apache server config files to match the port that was set for the container.

In /etc/apache2/sites-enabled/000-default.conf
So in my case I added *:8080 to the first line
<VirtualHost *:80 *:8080>

then in /etc/apache2/ports.conf
just below
Listen 80

I added the line
Listen 8080

then after running
apachectl restart

everything was working as expected

[patwork]

ChatGPT suggested running the socat command in the container:

socat TCP-LISTEN:8080,fork TCP:localhost:80

Unfortunately, this requires socat to be installed after the container is created, in addition socat must be running in the background because otherwise it blocks the console

docker exec -it your_container_name_or_id apt-get update
docker exec -it your_container_name_or_id apt-get install -y socat

@visionm45 Is there any way to automate such configuration change in apache's files from docker-compose.yml?

[visionm45]

Im sure there would be a way to automate it. But i am not the one to ask about this as i am quite new to docker.

[visionm45]

It looks to me like the apply-templates.sh is the script which is doing all of the configuration and which produces the dockerfiles. Maybe they could be eddited or maybe a docker-compose would be easier. Im not sure.

[tianon]

If you need additional packages installed, you should write a short Dockerfile something like:

FROM wordpress
RUN apt-get update && apt-get install -y socat

And build/deploy that instead of using the wordpress image as-is (if you're using docker-compose, you can have compose build the image for you as part of docker-compose up)

[LaurentGoderre]

To make this work, you can't use the localhost host because it stays within the container which only has the port 80 mapped.

To achieve this:

1. Add the following to your docker-compose.yml

extra_hosts:
  - "my_site:host-gateway"

2. Add the following line to your host file in your host (vim /etc/hosts/)

127.0.0.1       my_site

3. Recreate your site (docker-compose down -v && docker-compose up)
4. Access you site via my_site:8080

[LaurentGoderre]