Running Caddy with Custom CMD in Dockerfile

Help
1

I am packaging my caddy with frontend which is a React frontend produced by vite and as far as I can see I am doing everything correct, but when I access my domain it says site can’t be reached. I tried to simplify this even further and use a static index.html.

Here’s my Dockerfile for the frontend:

FROM node:14-alpine as build

WORKDIR /frontend

COPY package*.json .

COPY Caddyfile .

RUN npm install

COPY . .

RUN npm run build

FROM caddy:2.7.6-alpine

EXPOSE 80

EXPOSE 443

COPY --from=build /frontend/Caddyfile /etc/caddy/Caddyfile
COPY --from=build /frontend/testDir /frontend/testDir

CMD ["caddy", "run"]

I double checked here and testDir is there in the container as well as the Caddyfile is there as well.

Here’s my Caddyfile, domain name is redacted:

domain.pro, www.domain.pro {
  root * frontend/testDir
  file_server
  reverse_proxy /api/* localhost:3000
}

I would like to relay requests to my other container called backend.
Finally here’s my compose.yml file:

services:
  backend:
    container_name: domain-backend
    build: ./backend
    ports:
      - "3000:3000"
  frontend:
    container_name: domain-frontend
    build: ./frontend
    ports:
      - "80:80"
      - "443:443"

I am setting this up both as production and dev environment so I’d be happy if I could locally have https://localhost working as well on the actual domain name. Any guidance is much appreciated. I simplified it as much as I can.

2

Are you missing a leading / here?

4

I tried that it still doesn’t work. The docs imho should explain things a little bit better.

I tried a similar approach outside a docker file and after root * it expects an absolute path.

5

Enable debug in Caddyfile and share the log output.

6

I simplified this even further:

I enabled debug mode for the simplified issue and here’s what the logs say:

2025-03-25 19:14:19 {"level":"info","ts":1742926459.2633002,"logger":"admin","msg":"admin endpoint started","address":"localhost:2019","enforce_origin":false,"origins":["//127.0.0.1:2019","//localhost:2019","//[::1]:2019"]}
2025-03-25 19:14:19 {"level":"info","ts":1742926459.2634373,"msg":"serving initial configuration"}```
7

What does it mean “nothing happens”?

What do you get from curl -v?

4 Likes
8

Looks like now I am running into certificate issues, curl -k https://domain.pro works and shows the page, however, I still can’t open it through the browser.

Wasn’t this done automatically? Does running it in a docker container change something related to SSL certificates?

(for those wondering,

CMD ["caddy", "run", "--config", "/etc/caddy/Caddyfile"]
needed to be explicit with the Caddyfile)

9

The logs will tell you what is happening with issuing certificates. By default you need http (port 80) exposed to the Internet and a resolvable host name for this to work.

3 Likes
10

I think you mean that needs to be a line specified within the Dockerfile, but yes, that is correct.

What do you mean it shows the page? It’s showing the index.html file within /frontend/testDir/api? I’m curious if your Caddyfile configuration is incorrect and requests are not landing where they should. debug would show the HTTP request Caddy is receiving.

When specifying the root folder, it needs a full path. So when @timelordx was asking if you’re forgetting a leading slash, that is why he asked. I’m not sure if Caddy is configured to search from the directory it is located in if there is no leading slash.

1 Like