如何在不运行Docker的情况下测试AWS Batch的boto3/moto Mock

Absolutely! You can skip Docker entirely when using Moto to mock AWS Batch's submit_job by creating a custom backend stub that overrides the default behavior. This eliminates the Docker dependency and drastically speeds up your tests—no more waiting 2 seconds per job submission.

Here's a step-by-step implementation:

1. Create a Custom Batch Backend #

First, we'll subclass Moto's existing BatchBackend and override the submit_job method to return your desired mock response instead of spinning up Docker containers.

from moto.batch.models import BatchBackend

class NoDockerBatchBackend(BatchBackend):
    def submit_job(self, job_name, job_queue, job_definition, **kwargs):
        # Return a simplified mock response without touching Docker
        return {
            "jobName": job_name,
            "jobId": f"mock-job-{hash(job_name)}",  # Generate a unique mock job ID
            "status": "SUBMITTED",
            "message": "This job is submitted"
        }

2. Create a Custom Mock Decorator/Context Manager #

Next, we'll create a custom version of Moto's mock_batch that replaces the default backend with our Docker-free one.

from moto.batch import mock_batch

def mock_batch_no_docker():
    # Start the standard Batch mock
    mock = mock_batch()
    mock.start()
    
    # Swap out the default backend with our custom one
    from moto.batch.models import batch_backends
    # Replace the backend for your target region (adjust if needed)
    batch_backends["us-east-1"] = NoDockerBatchBackend("us-east-1")
    
    return mock

3. Use the Custom Mock in Your Tests #

Now you can use this custom mock in your test cases just like you'd use the regular mock_batch. You'll still need to set up prerequisite resources (like job queues and definitions) since Moto expects those to exist, but job submission will be instant.

import boto3

def test_batch_submit_job_without_docker():
    with mock_batch_no_docker():
        batch_client = boto3.client("batch", region_name="us-east-1")
        
        # Set up required resources (Moto validates these exist)
        batch_client.create_job_queue(
            jobQueueName="test-queue",
            state="ENABLED",
            priority=1
        )
        batch_client.register_job_definition(
            jobDefinitionName="test-job-def",
            type="container",
            containerProperties={
                "image": "dummy-image",  # No need for a real Docker image
                "vcpus": 1,
                "memory": 256
            }
        )
        
        # Submit your job
        response = batch_client.submit_job(
            jobName="test-job",
            jobQueue="test-queue",
            jobDefinition="test-job-def"
        )
        
        # Verify the mock response
        assert response["jobName"] == "test-job"
        assert response["message"] == "This job is submitted"
        assert response["status"] == "SUBMITTED"

Key Notes #

• Region Flexibility: If you're testing with a different AWS region, update the region string in both the backend setup and your Boto3 client.
• Moto Version Compatibility: This approach works with recent Moto versions, but if you're using an older release, double-check the BatchBackend method signatures (they can change slightly between versions).
• Extensibility: You can expand the mock response to include any fields your application expects, like jobArn or createdAt, by adding them to the return dictionary.

This approach gives you full control over the mock behavior, removes the Docker dependency, and makes your tests run lightning-fast.

内容的提问来源于stack exchange，提问作者mrówa