aws-samples · ginjups · Dec 24, 2025
diff --git a/lambda-durable-esm-and-chaining/README.md b/lambda-durable-esm-and-chaining/README.md
@@ -0,0 +1,144 @@
+# Event-Driven Data Pipeline with Lambda Durable Functions
+
+This serverless pattern demonstrates how to build an event-driven data processing pipeline using AWS Lambda Durable Functions with **direct SQS Event Source Mapping** and Lambda invoke chaining.
+
+## How It Works
+
+This pattern demonstrates an event-driven data processing pipeline using AWS Lambda Durable Functions with direct SQS Event Source Mapping. When a message arrives in the SQS queue, it directly triggers the durable function (no intermediary Lambda needed). The durable function then orchestrates a series of specialized processing steps using Lambda invoke chaining - first validating the incoming data, then transforming it (converting data_source to uppercase), and finally storing the processed results in DynamoDB. Throughout this process, the durable function automatically creates checkpoints, enabling fault-tolerant execution that can recover from failures without losing progress. The entire pipeline operates within the 15-minute ESM execution limit, making it ideal for reliable batch processing workflows.
+
+## Architecture Overview
+
+The pattern showcases two key Durable Functions capabilities:
+1. **Direct Event Source Mapping**: SQS directly triggers the durable function (15-minute limit)
+2. **Lambda Invoke Chaining**: Orchestrates specialized processing functions
+
+![Architecture Diagram](architecture-diagram.png)
+
+## Key Features
+
+- **Direct ESM Integration**: No intermediary function needed
+- **15-minute execution constraint**: Demonstrates ESM time limits
+- **Fault-tolerant processing**: Automatic checkpointing and recovery
+- **Microservices coordination**: Chains specialized Lambda functions
+- **Batch processing**: Handles multiple SQS records per invocation
+- **Simple storage**: Uses DynamoDB for processed data
+
+## Important ESM Constraints
+
+⚠️ **15-Minute Execution Limit**: When using Event Source Mapping with Durable Functions, the total execution time cannot exceed 15 minutes. This includes:
+- All processing steps
+- Function invocations
+- No long wait operations
+
+## Use Cases
+
+- ETL pipelines with validation and transformation
+- Event-driven microservices orchestration
+- Batch processing with fault tolerance
+- Data processing workflows requiring checkpointing
+
+## Prerequisites
+
+- [AWS CLI](https://docs.aws.amazon.com/cli/latest/userguide/getting-started-install.html) configured with appropriate permissions
+- [AWS SAM CLI](https://docs.aws.amazon.com/serverless-application-model/latest/developerguide/install-sam-cli.html) latest version installed
+- [Python 3.14](https://www.python.org/downloads/release/python-3140/) runtime installed
+
+## Deployment
+
+1. **Build the application**:
+   ```bash
+   sam build
+   ```
+
+2. **Deploy to AWS**:
+   ```bash
+   sam deploy --guided
+   ```
+
+   Note the outputs after deployment:
+   - `DataProcessingQueueUrl`: Use this for `<QUEUE_URL>`
+   - `ProcessedDataTable`: Use this for `<PROCESSED_DATA_TABLE>`
+
+3. **Test the pipeline**:
+   ```bash
+   # Send a test message to SQS
+   aws sqs send-message \
+     --queue-url <QUEUE_URL> \
+     --message-body '{"data_source": "test.csv", "processing_type": "standard"}'
+   ```
+
+4. **Verify successful processing**:
+   ```bash
+   # Check if data was processed and stored in DynamoDB
+   aws dynamodb scan --table-name <PROCESSED_DATA_TABLE> --query 'Items[*]'
+   ```
+
+   **Success indicators:**
+   - You should see at least one item in the DynamoDB table
+   - Original input data: `"data_source": "test.csv"`
+   - Transformed data: `"data_source": "TEST.CSV"` (uppercase transformation applied)
+   - Execution tracking with unique `execution_id`
+   - Timestamps showing when data was processed and stored
+
+   This confirms the entire pipeline worked: SQS → Durable Function → Validation → Transformation → Storage → DynamoDB
+
+## Components
+
+### 1. Durable Pipeline Function (`src/durable_pipeline/`)
+- **Direct SQS Event Source Mapping**: Receives SQS events directly
+- **15-minute execution limit**: Must complete all processing within ESM constraints
+- **Batch processing**: Handles multiple SQS records per invocation
+- **Lambda invoke chaining**: Orchestrates validation, transformation, and storage
+- **Automatic checkpointing**: Recovers from failures without losing progress
+
+### 2. Specialized Processing Functions
+- **Validation Function**: Simple data validation checks
+- **Transformation Function**: Basic data transformation
+- **Storage Function**: Persists processed data to DynamoDB
+
+## Monitoring
+
+- CloudWatch Logs for execution tracking
+- DynamoDB table for processed data
+- SQS DLQ for failed messages
+
+## Configuration
+
+Key environment variables:
+- `ENVIRONMENT`: Deployment environment (dev/prod)
+- `PROCESSED_DATA_TABLE`: DynamoDB table for processed data
+- `VALIDATION_FUNCTION_ARN`: ARN of validation function
+- `TRANSFORMATION_FUNCTION_ARN`: ARN of transformation function
+- `STORAGE_FUNCTION_ARN`: ARN of storage function
+
+## ESM-Specific Considerations
+
+- **Execution Timeout**: Set to 900 seconds (15 minutes) maximum
+- **Batch Size**: Configured for optimal processing (5 records)
+- **Error Handling**: Uses SQS DLQ for failed batches
+- **Efficient Processing**: Optimized for speed to stay within time limits
+
+## Error Handling
+
+- Automatic retries with exponential backoff
+- Dead Letter Queue for failed messages
+- Partial batch failure support
+- Checkpoint-based recovery
+
+## Cost Optimization
+
+- Pay only for active compute time
+- Efficient batch processing
+- Automatic scaling based on queue depth
+
+## Cleanup
+
+```bash
+sam delete
+```
+
+## Learn More
+
+- [AWS Lambda Durable Functions Documentation](https://docs.aws.amazon.com/lambda/latest/dg/durable-functions.html)
+- [Event Source Mappings with Durable Functions](https://docs.aws.amazon.com/lambda/latest/dg/durable-invoking-esm.html)
+- [Lambda Invoke Chaining](https://docs.aws.amazon.com/lambda/latest/dg/durable-examples.html#durable-examples-chained-invocations)
diff --git a/lambda-durable-esm-and-chaining/architecture.png b/lambda-durable-esm-and-chaining/architecture.png
diff --git a/lambda-durable-esm-and-chaining/example-pattern.json b/lambda-durable-esm-and-chaining/example-pattern.json
@@ -0,0 +1,67 @@
+{
+    "title": "Event-Driven Data Pipeline with Lambda Durable Functions",
+    "description": "This serverless pattern demonstrates how to build an event-driven data processing pipeline using AWS Lambda Durable Functions with direct SQS Event Source Mapping and Lambda invoke chaining.",
+    "language": "Python",
+    "level": "200",
+    "framework": "SAM",
+    "services": ["sqs","lambda", "dynamoDB"],
+    "introBox": {
+      "headline": "How it works",
+      "text": [
+        "This pattern demonstrates an event-driven data processing pipeline using AWS Lambda Durable Functions with direct SQS Event Source Mapping. When a message arrives in the SQS queue, it directly triggers the durable function (no intermediary Lambda needed). The durable function then orchestrates a series of specialized processing steps using Lambda invoke chaining - first validating the incoming data, then transforming it (converting data_source to uppercase), and finally storing the processed results in DynamoDB. Throughout this process, the durable function automatically creates checkpoints, enabling fault-tolerant execution that can recover from failures without losing progress. The entire pipeline operates within the 15-minute ESM execution limit, making it ideal for reliable batch processing workflows."
+      ]
+    },
+    "testing": {
+      "headline": "Testing",
+      "text": [
+        "See the GitHub repo for detailed testing instructions."
+      ]
+    },
+    "cleanup": {
+      "headline": "Cleanup",
+      "text": [
+        "Delete the stack: <code>sam delete</code>."
+      ]
+    },
+    "deploy": {
+      "text": [
+        "sam build",
+        "sam deploy --guided"
+      ]
+    },
+    "gitHub": {
+      "template": {
+        "repoURL": "https://github.com/aws-samples/serverless-patterns/tree/main/lambda-durable-esm-and-chaining",
+        "templateURL":"serverles-patterns/lambda-durable-esm-and-chaining",
+        "templateFile": "template.yaml",
+        "projectFolder": "lambda-durable-esm-and-chaining"
+      }
+    },
+    "resources": {
+      "headline": "Additional resources",
+      "bullets": [
+        {
+          "text": "AWS Lambda Durable Functions Documentation",
+          "link": "https://docs.aws.amazon.com/lambda/latest/dg/durable-functions.html"
+        },
+        {
+          "text": "Event Source Mappings with Durable Functions",
+          "link": "https://docs.aws.amazon.com/lambda/latest/dg/durable-invoking-esm.html"
+        },
+        {
+          "text": "Durbale Function Lambda Invoke Chaining",
+          "link": "https://docs.aws.amazon.com/lambda/latest/dg/durable-examples.html#durable-examples-chained-invocations"
+        }
+      ]
+    },
+     "authors": [
+        {
+          "name": "Sahithi Ginjupalli",
+          "image": "https://drive.google.com/file/d/1YcKYuGz3LfzSxiwb2lWJfpyi49SbvOSr/view?usp=sharing",
+          "bio": "Cloud Engineer at AWS with a passion for diving deep into cloud and AI services to build innovative serverless applications.",
+          "linkedin": "ginjupalli-sahithi-37460a18b",
+          "twitter": ""
+        }
+      ]
+  }
+
diff --git a/lambda-durable-esm-and-chaining/src/durable_pipeline/handler.py b/lambda-durable-esm-and-chaining/src/durable_pipeline/handler.py
@@ -0,0 +1,100 @@
+import json
+import os
+import boto3
+from datetime import datetime
+from typing import Dict, Any, List
+from aws_durable_execution_sdk_python import DurableContext, durable_execution
+
+# Initialize AWS clients
+dynamodb = boto3.resource('dynamodb')
+lambda_client = boto3.client('lambda')
+
+@durable_execution
+def lambda_handler(event: Dict[str, Any], context: DurableContext) -> Dict[str, Any]:
+    """
+    Main durable pipeline function that processes SQS events directly via ESM.
+    Demonstrates lambda invoke chaining with checkpointing and recovery.
+    Limited to 15 minutes total execution time due to ESM constraints.
+    """
+
+    # Extract configuration from environment
+    validation_function_arn = os.environ['VALIDATION_FUNCTION_ARN']
+    transformation_function_arn = os.environ['TRANSFORMATION_FUNCTION_ARN']
+    storage_function_arn = os.environ['STORAGE_FUNCTION_ARN']
+    processed_data_table = os.environ['PROCESSED_DATA_TABLE']
+    environment = os.environ.get('ENVIRONMENT', 'dev')
+
+    print(f"Processing SQS batch with {len(event.get('Records', []))} records")
+
+    # Process each SQS record in the batch
+    batch_results = []
+
+    for record in event.get('Records', []):
+        try:
+            # Extract data from SQS record
+            message_id = record['messageId']
+            data = json.loads(record['body'])
+            execution_name = f"{environment}-esm-{message_id}"
+
+            print(f"Processing record: {message_id}")
+
+            # Step 1: Validate data by invoking validation function
+            validation_result = context.invoke(
+                validation_function_arn,
+                {'data': data, 'execution_id': execution_name},
+                name=f'validate-data-{message_id}'
+            )
+
+            if not validation_result.get('is_valid', False):
+                batch_results.append({
+                    'message_id': message_id,
+                    'status': 'failed',
+                    'reason': 'validation_failed'
+                })
+                continue
+
+            # Step 2: Transform data by invoking transformation function
+            transformation_result = context.invoke(
+                transformation_function_arn,
+                {'data': data, 'execution_id': execution_name},
+                name=f'transform-data-{message_id}'
+            )
+
+            # Step 3: Store processed data by invoking storage function
+            storage_result = context.invoke(
+                storage_function_arn,
+                {
+                    'transformed_data': transformation_result,
+                    'execution_id': execution_name,
+                    'original_data': data
+                },
+                name=f'store-data-{message_id}'
+            )
+
+            batch_results.append({
+                'message_id': message_id,
+                'status': 'completed',
+                'execution_id': execution_name
+            })
+
+        except Exception as e:
+            print(f"Error processing record {record.get('messageId', 'unknown')}: {str(e)}")
+            batch_results.append({
+                'message_id': record.get('messageId', 'unknown'),
+                'status': 'error',
+                'error': str(e)
+            })
+
+    # Return batch processing summary
+    successful_records = len([r for r in batch_results if r['status'] == 'completed'])
+    failed_records = len([r for r in batch_results if r['status'] in ['failed', 'error']])
+
+    return {
+        'batch_summary': {
+            'total_records': len(batch_results),
+            'successful_records': successful_records,
+            'failed_records': failed_records
+        },
+        'record_results': batch_results,
+        'processed_at': datetime.utcnow().isoformat()
+    }
diff --git a/lambda-durable-esm-and-chaining/src/durable_pipeline/requirements.txt b/lambda-durable-esm-and-chaining/src/durable_pipeline/requirements.txt
@@ -0,0 +1 @@
+aws-durable-execution-sdk-python
diff --git a/lambda-durable-esm-and-chaining/src/storage/handler.py b/lambda-durable-esm-and-chaining/src/storage/handler.py
@@ -0,0 +1,47 @@
+import boto3
+import os
+from typing import Dict, Any
+from datetime import datetime
+
+dynamodb = boto3.resource('dynamodb')
+
+def lambda_handler(event: Dict[str, Any], context) -> Dict[str, Any]:
+    """Simple data storage function that saves to DynamoDB"""
+
+    transformed_data = event['transformed_data']
+    execution_id = event['execution_id']
+    original_data = event['original_data']
+
+    table_name = os.environ['PROCESSED_DATA_TABLE']
+
+    print(f"Storing processed data for execution: {execution_id}")
+
+    try:
+        table = dynamodb.Table(table_name)
+
+        # Store processed data in DynamoDB
+        item = {
+            'execution_id': execution_id,
+            'original_data': original_data,
+            'transformed_data': transformed_data,
+            'stored_at': datetime.utcnow().isoformat(),
+            'data_source': original_data.get('data_source', 'unknown'),
+            'processing_type': original_data.get('processing_type', 'standard')
+        }
+
+        table.put_item(Item=item)
+
+        return {
+            'success': True,
+            'execution_id': execution_id,
+            'table_name': table_name,
+            'stored_at': datetime.utcnow().isoformat()
+        }
+
+    except Exception as e:
+        print(f"Error storing data: {str(e)}")
+        return {
+            'success': False,
+            'execution_id': execution_id,
+            'error': str(e)
+        }
diff --git a/lambda-durable-esm-and-chaining/src/transformation/handler.py b/lambda-durable-esm-and-chaining/src/transformation/handler.py
@@ -0,0 +1,22 @@
+from typing import Dict, Any
+from datetime import datetime
+
+def lambda_handler(event: Dict[str, Any], context) -> Dict[str, Any]:
+    """Simple data transformation function"""
+
+    data = event['data']
+    execution_id = event['execution_id']
+
+    print(f"Transforming data for execution: {execution_id}")
+
+    # Simple transformation: add processing metadata and uppercase data_source
+    transformed_data = {
+        'original_data': data,
+        'data_source': data.get('data_source', '').upper(),
+        'processing_type': data.get('processing_type', 'standard'),
+        'processed_at': datetime.utcnow().isoformat(),
+        'execution_id': execution_id,
+        'transformation_applied': 'uppercase_data_source'
+    }
+
+    return transformed_data