10 expert AWS Lambda Layers answers researched from official documentation. Every answer cites authoritative sources you can verify.
Distribution mechanism for libraries, custom runtimes, and dependencies shared across Lambda functions. Layer = ZIP archive extracted to /opt in function execution environment. Benefits: reduce deployment package size, share code across functions, separate dependencies from business logic. Limit: 5 layers per function, 250MB unzipped total.
Package dependencies in correct directory structure: python/lib/python3.x/site-packages/ (Python), nodejs/node_modules/ (Node.js), ruby/gems/2.7.0/ (Ruby). ZIP structure, upload via: aws lambda publish-layer-version --layer-name my-layer --zip-file fileb://layer.zip --compatible-runtimes python3.12. Attach to function: aws lambda update-function-configuration --function-name my-func --layers arn:aws:lambda:region:account:layer:my-layer:1.
Python libraries: python/lib/python3.x/site-packages/. Binaries: python/bin/. Example: python/lib/python3.12/site-packages/requests/. Lambda adds /opt/python/lib/python3.12/site-packages to sys.path. Alternative structure: python/ (directly, for shared modules). ZIP from layer directory: zip -r layer.zip python/.
Node.js modules: nodejs/node_modules/ or nodejs/node14/node_modules/ (runtime-specific). Example: nodejs/node_modules/axios/. Lambda adds /opt/nodejs/node_modules to NODE_PATH. Binary dependencies: include platform-specific builds (compile on Amazon Linux 2). Package: zip -r layer.zip nodejs/.
Best practices: (1) Immutable versions (each update creates new version), (2) Semantic versioning in layer name (my-layer-v1-2-0), (3) ARN with version in function config (arn:...:layer:my-layer:3, not latest), (4) Test new versions before updating production functions, (5) Deprecate old versions after migration (DeleteLayerVersion). Versions retained indefinitely unless deleted.
Grant permission: aws lambda add-layer-version-permission --layer-name my-layer --version-number 1 --principal 123456789012 --action lambda:GetLayerVersion. For public: --principal '*' (use with caution). Other account references layer ARN from original account. Common use case: shared libraries across organization accounts.
AWS provides layers for: (1) AWS SDK (updated regularly, avoid bundling in function), (2) Lambda Insights (CloudWatch monitoring), (3) AWS Parameters and Secrets extension, (4) X-Ray SDK. Regional ARNs published in docs. Example: AWS SDK for Python (boto3) layer ARN per region. Use to reduce deployment size.
Size limits: 50MB zipped per layer, 250MB total unzipped (all layers + function). Cold start: layers add ~50-200ms extraction time (cached between invocations). Optimization: minimize layer count (combine related dependencies), use sparse layers (only add to functions that need), compress (zstd for faster extraction). Monitor cold start with X-Ray.
Debug techniques: (1) Print sys.path (Python) or process.env.NODE_PATH (Node.js) in handler, (2) Verify ZIP structure (unzip layer.zip locally, check paths), (3) Test locally with SAM CLI (sam local invoke with --layers), (4) CloudWatch Logs: 'ModuleNotFoundError' indicates wrong path, (5) List /opt contents in handler (os.listdir('/opt')).
Automation: (1) SAM template: AWS::Serverless::LayerVersion resource with ContentUri, (2) Terraform: aws_lambda_layer_version resource, (3) CDK: new LayerVersion(this, 'Layer', {code: Code.fromAsset('layer/')}), (4) CI/CD: build layer in pipeline, publish new version, update function configs. Version in parameter store for reference.