基于HTTP请求自动启停AWS EC2开发服务器方案咨询

Great question—this is such a common pain point with dev servers on EC2, and AWS has all the tools to build a clean, cost-effective solution. Let me walk you through a step-by-step setup that I’ve used for my own projects:

We’ll use API Gateway + Lambda to handle the HTTP trigger and kick off the EC2 instance. This is serverless, so you only pay for actual requests, no idle costs.

Setup Steps: #

• Create a Lambda Function:
  • Use Python (or your preferred runtime) with permissions to call ec2:StartInstances and ec2:DescribeInstances (to avoid trying to start an already running instance). Attach an IAM role with these permissions plus CloudWatch Logs access for debugging.
  • Here’s a sample function:

    import boto3
    from datetime import datetime
    
    ec2 = boto3.client('ec2')
    INSTANCE_ID = 'your-ec2-instance-id' # Replace with your actual instance ID
    
    def lambda_handler(event, context):
        # Check current instance state first
        response = ec2.describe_instances(InstanceIds=[INSTANCE_ID])
        state = response['Reservations'][0]['Instances'][0]['State']['Name']
    
        if state == 'stopped':
            ec2.start_instances(InstanceIds=[INSTANCE_ID])
            return {
                'statusCode': 200,
                'body': f"Starting instance {INSTANCE_ID}. It'll be ready in ~2-5 minutes."
            }
        elif state == 'running':
            return {
                'statusCode': 200,
                'body': f"Instance {INSTANCE_ID} is already up and running."
            }
        else:
            return {
                'statusCode': 400,
                'body': f"Instance {INSTANCE_ID} is in {state} state—can't start right now."
            }
    

• Configure API Gateway:
  • Create an HTTP API (cheaper and simpler than REST API for this use case) and add a route (e.g., /start-dev-server) that integrates with your Lambda function. Enable CORS if you plan to call this endpoint from a browser.
  • Test it out: Send a GET/POST request to your API endpoint, and you’ll get a clear response confirming the instance’s status.

To handle idle shutdown, we’ll track the last request time and use EventBridge + Lambda + DynamoDB to check for inactivity and stop the instance automatically.

Setup Steps: #

• Create a DynamoDB Table:
  • Make a simple table with InstanceID as the string primary key, and add an attribute LastRequestTimestamp (number, Unix timestamp). The free tier is more than enough for this single record.
• Update the Start Lambda:
  • Add code to update the DynamoDB table every time a request comes in, so we always have the latest activity time:

    # Add these imports/initializations at the top
    dynamodb = boto3.resource('dynamodb')
    table = dynamodb.Table('DevServerActivity') # Replace with your table name
    
    # ... after handling the instance start/running check ...
    table.put_item(
        Item={
            'InstanceID': INSTANCE_ID,
            'LastRequestTimestamp': int(datetime.now().timestamp())
        }
    )
    

• Create a Stop Check Lambda:
  • This function runs on a schedule, checks the last activity time, and stops the instance if it’s been idle for 1 hour:

    import boto3
    from datetime import datetime
    
    ec2 = boto3.client('ec2')
    dynamodb = boto3.resource('dynamodb')
    table = dynamodb.Table('DevServerActivity')
    INSTANCE_ID = 'your-ec2-instance-id'
    IDLE_THRESHOLD = 3600 # 1 hour in seconds
    
    def lambda_handler(event, context):
        # Fetch the last activity timestamp
        response = table.get_item(Key={'InstanceID': INSTANCE_ID})
        if 'Item' not in response:
            return {'statusCode': 200, 'body': "No activity recorded for this instance yet."}
        
        last_request = response['Item']['LastRequestTimestamp']
        current_time = int(datetime.now().timestamp())
        time_since_last = current_time - last_request
    
        # Check instance state and idle duration
        ec2_response = ec2.describe_instances(InstanceIds=[INSTANCE_ID])
        state = ec2_response['Reservations'][0]['Instances'][0]['State']['Name']
    
        if state == 'running' and time_since_last >= IDLE_THRESHOLD:
            ec2.stop_instances(InstanceIds=[INSTANCE_ID])
            return {
                'statusCode': 200,
                'body': f"Instance {INSTANCE_ID} stopped due to 1 hour of inactivity."
            }
        else:
            return {
                'statusCode': 200,
                'body': f"Instance is {state} | Last activity {time_since_last} seconds ago."
            }
    

• Set Up EventBridge Schedule:
  • Create a rule that runs the Stop Check Lambda every 5 minutes (or 10 minutes, depending on how quickly you want to shut down). This ensures we catch the 1-hour mark without unnecessary frequent runs.

If your instance takes a few minutes to boot and start serving requests, you can modify the Start Lambda to poll the instance’s status (and even your dev server’s HTTP endpoint) before returning a "ready" response. For more complex workflows, use Step Functions to orchestrate the start → wait for running → wait for service ready → notify user flow.

Make sure your Lambda roles have these permissions:

• For the Start Lambda: ec2:DescribeInstances, ec2:StartInstances, dynamodb:PutItem, logs:CreateLogGroup, logs:CreateLogStream, logs:PutLogEvents
• For the Stop Check Lambda: ec2:DescribeInstances, ec2:StopInstances, dynamodb:GetItem, logs:*

This setup is lightweight, cost-efficient, and requires almost no maintenance once configured. I’ve been using this for my own EC2 dev server for months, and it’s saved me a ton of unnecessary charges.

内容的提问来源于stack exchange，提问作者ozkolonur