Boto3 在 docker 中需要 AWS 区域，但在本地环境中不需要

Question

I have a script setup for a lambda that uses Python and will eventually connect to a DynamoDB table. I setup everything locally (a virtual environment using pipenv) using the docker image AWS provides for DynamoDB and it all worked without a hitch. Then I tried to dockerize the Python. When I run my table creation script in my local virtual environment, it runs without a problem. When I run the same script from within my docker container, I get an error. I'm not sure what the difference is. Right now, I use this line to connect:

dynamodb = boto3.resource('dynamodb', endpoint_url="http://dynamodb:8000")

dynamodb is the name of the service in docker, so this should work. When I replace dynamodb:8000 with localhost:8000 , it works fine in my local venv. When I run it in docker, I get

botocore.exceptions.NoRegionError: You must specify a region.

The big question is why it's looking for a region in docker, but not locally. Here's my docker-compose for good measure:

version: '3'
services:
    dynamodb:
        command: "-jar DynamoDBLocal.jar -sharedDb -optimizeDbBeforeStartup -dbPath ./data"
        image: "amazon/dynamodb-local:latest"
        container_name: dynamodb
        ports:
            - "8000:8000"
        volumes:
            - "./database_data:/home/dynamodblocal/data"
        working_dir: /home/dynamodblocal
    lambda:
        build: .
        container_name: user-rekognition-lambda
        volumes:
            - ./:/usr/src/app

Answer 1

In one of the AWS blogs, local AWS Glue , they share the ~/.aws/ in read-only mode with the docker container using volume option:

-v ~/.aws:/root/.aws:ro

This would be the easiest way for you to reuse your credentials from host workstation inside the docker.

Answer 2

boto3 has a bug that even if you explicitly give it an endpoint_url , it still wants to know a region_name . Why? I don't know, and it doesn't seem to use this name for anything as far as I can tell.

In most people's setups, $HOME/.aws/config contains some default choice of region, so boto3 picks up the region_name from that configuration, and then just ignores it...

Since probably your docker image doesn't have that file, the trivial solution is just to add region_name='us-east-1' (for example) explicitly to your boto3.resource() call. Again, the specific region name you choose won't matter - boto3 will connect to the URL you give it, not to that region.

So the full command becomes:

dynamodb = boto3.resource('dynamodb',
    endpoint_url="http://dynamodb:8000",
    region_name="us-east-1")