Trouble running startup scripts for containers - docker

Question

I am trying to start three docker containers using docker-compose; a postgresql database, a rest api and a frontend web app.
Previously I had the rest api and the database working perfectly (ran migrations and started the server) but when I wanted to send requests to it from my web app I had trouble connecting to the docker network. All the discussion on the internet was centered around connecting a web app in a container to an api in another container and I didn't find any promising method to connect to it (besides enabling port forwarding in the kernel and exposing myself to the network?) so I just decided to package the web app in a container as well.
My directory structure:
ProjectName
|-> projectapi
|-> |-> api.docker
|-> |-> api_start.sh
|-> projectapp
|-> |-> front.docker
|-> |-> front_start.sh
|-> docker-compose.yml
The problem is running these startup scripts in the right context (tbh I only need to run a npm start and a python manage.py runserver).
I can't think of anything I haven't tried, but most of my efforts have been centered around mucking around with paths because for the longest time the issue was not finding the file which I think I have fixed.
This is the docker-compose file, I have tried having all sorts of command entries in here to run the desired startup script. I have also used entrypoint.
version: '3.7'
services:
db:
container_name: projectdb
image: postgres:9.6-alpine
restart: always
volumes:
- projectdb:/var/lib/postgresql/data/
environment:
POSTGRES_DB: projectdb
POSTGRES_PASSWORD: root
ports:
- "8001:5432"
api:
container_name: projectapi
build:
context: projectapi/
dockerfile: api.docker
ports:
- "8000:8000"
expose:
- "8000"
depends_on:
- db
restart: always
environment:
POSTGRES_DB: 'projectdb'
POSTGRES_USER: 'postgres'
POSTGRES_PASSWORD: 'root'
POSTGRES_HOST: 'db'
front:
container_name: projectapp
build:
context: projectapp/
dockerfile: front.docker
ports:
- "3000:3000"
restart:
always
volumes:
projectdb:
Then there are the Dockerfiles:
front
FROM node:8
RUN mkdir /projectapp
COPY $HOSTDIR/package*.json /projectapp/
RUN npm install /projectapp
COPY $HOSTDIR/* /projectapp/
ENTRYPOINT ["npm", "start", "/projectapp"]
back
FROM python:3.6-slim
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONBUFFERED 1
RUN mkdir /projectapi
COPY $HOSTDIR/requirements.txt /projectapi/
RUN pip install -r /projectapi/requirements.txt
COPY $HOSTDIR/* /projectapi/
CMD ["python", "manage.py", "migrate"]
That last line could be runserver as well. These are just some examples of the permutations I've gone through but at this point I feel the problem is some conceptual misunderstanding, I've read the docs.
The error messages are all different permutations of not finding the startup script. I think there was a point where I managed to run a startup script and the error became it couldn't find manage.py, at which point I started to look into how to write the script better than just: python manage.py runserver but didn't get very far.

Answer 1

Try using a WORKDIR. Your current run path doesn't have manage.py -
I tried changing it momentarily -
WORKDIR /projectapi
COPY requirements.txt ./
RUN pip install -r requirements.txt
COPY . .
CMD ["python", "manage.py", "runserver"]

Answer 2

I have solved this issue but still haven't gotten the network to work, however that's another question. While I solved the problem with the startup scripts it turned out not to be needed, here's the current state of things:
Directory structure is the same as before.
docker-compose.yml
version: '3.7'
services:
db:
container_name: compdb
image: postgres:9.6-alpine
restart: always
volumes:
- compdb:/var/lib/postgresql/data/
environment:
POSTGRES_DB: compdb
POSTGRES_PASSWORD: root
networks:
- internal
ports:
- "8001:5432"
api:
container_name: back
build:
context: back/
dockerfile: api.docker
entrypoint: ["python", "/back/manage.py", "runserver", "0.0.0.0:8000"]
networks:
- internal
ports:
- "8000:8000"
expose:
- "8000"
depends_on:
- db
restart: always
environment:
POSTGRES_DB: 'compdb'
POSTGRES_USER: 'postgres'
POSTGRES_PASSWORD: 'root'
POSTGRES_HOST: 'db'
front:
container_name: front
build:
context: front/
dockerfile: front.docker
entrypoint: ["npm","start", "--prefix", "/front/"]
networks:
- internal
ports:
- "3000:3000"
expose:
- "3000"
depends_on:
- api
restart:
always
staff:
container_name: staff
build:
context: staff/
dockerfile: staff.docker
entrypoint: ["npm","start","--prefix","/staff/"]
networks:
- internal
ports:
- "3006:3006"
expose:
- "3006"
depends_on:
- api
restart:
always
volumes:
compdb:
networks:
internal:
front
FROM node:8
RUN mkdir /front
COPY package*.json /front/
RUN npm install /front
COPY . /front/
back
FROM python:3.6-slim
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONBUFFERED 1
RUN mkdir /back
COPY requirements.txt /back/
RUN pip install -r /back/requirements.txt
COPY . /back/
staff is similar to front.
The problem was solved by moving the build context into each directory with docker compose. Running startup scripts can be done by changing the entrypoint, however for local development attaching to the container to run migrations or similar is more convenient.