Calling local-data-api from another container

[eddiepease]

Hi,

I have just spent over the day fixing an issue where the local-data-api is being called from another docker container rather than localhost (in this case, I was using SAM to invoke a local lambda container but the situation applies more generally). While this is no a bug with this repo, I think that the repo would benefit from a few small additions / extra clarification.

The Problem

After creating a user-defined docker network (docker network create dev-net), I did the following:

• I used the following postgres docker compose file below (exactly the same as in the README with the addition of a docker network)
• Set up a test table in the database
• Ensured that the lambda container was being launched in this same network

version: '3.1'

services:
  local-data-api:
    image: koxudaxi/local-data-api
    container_name: api
    networks:
      - dev-net
    restart: always
    environment:
      ENGINE: PostgreSQLJDBC
      POSTGRES_HOST: db
      POSTGRES_PORT: 5432
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: example
      RESOURCE_ARN: 'arn:aws:rds:us-east-1:123456789012:cluster:dummy'
      SECRET_ARN: 'arn:aws:secretsmanager:us-east-1:123456789012:secret:dummy'
    ports:
      - "8080:80"
  db:
    image: postgres:10.7-alpine
    container_name: db
    networks: 
      - dev-net
    restart: always
    environment:
      POSTGRES_PASSWORD: example
      POSTGRES_DB: test
    ports:
        - "5432:5432"
        
networks:
  dev-net:
    external: true

I tried to connect to the local-data-api with the following command, with a similar command to that in the README:

rds_client = boto3.client('rds-data', endpoint_url='http://local_local-data-api_1:8080', aws_access_key_id='aaa', aws_secret_access_key='bbb')
with local_local-data-api_1 being the automatically generated name for the api container in the docker compose.

The Solution

It turns out that there were two overlapping issues:

1. the URL above is invalid because it has an underscore
2. port 8080 sees to be already used so cannot be referred to in the lambda container

I solved this by explicitly naming the containers in the docker compose file as follows:

version: '3.1'

services:
  local-data-api:
    image: koxudaxi/local-data-api
    container_name: api
    networks:
      - dev-net
    restart: always
    environment:
      ENGINE: PostgreSQLJDBC
      POSTGRES_HOST: db
      POSTGRES_PORT: 5432
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: example
      RESOURCE_ARN: 'arn:aws:rds:us-east-1:123456789012:cluster:dummy'
      SECRET_ARN: 'arn:aws:secretsmanager:us-east-1:123456789012:secret:dummy'
    ports:
      - "8080:80"
  db:
    image: postgres:10.7-alpine
    container_name: db
    networks: 
      - dev-net
    restart: always
    environment:
      POSTGRES_PASSWORD: example
      POSTGRES_DB: test
    ports:
        - "5432:5432"
        
networks:
  dev-net:
    external: true

When calling the database from the lambda container, I also switched to port 80, which seemed to work, as follows:

rds_client = boto3.client('rds-data', endpoint_url='http://api:80', aws_access_key_id='aaa', aws_secret_access_key='bbb')

Suggestion

Given the above, I suggest that the following changes should be made to the repo:

1. The docker compose files in the repo should explicitly name the containers such that no underscores are generated when they are created
2. An extra section should be added to the README to discuss this issue, in particular the required switch to port 80 (unless a fix can be found to switch to 8080)