Lambda Flashcards

Question

What are various storage systems available for Lambda?

Answer 1

1. /tmp Directory (Ephemeral Storage) Size Limit Default: 512 MB (configurable up to 10 GB) Persistence Only during the lifetime of the execution environment Use Cases Temporary file storage, intermediate processing, caching 2. Amazon S3 (Object Storage) Size Limit Virtually unlimited Persistence Long-term, highly durable (11 9s durability) Use Cases File storage, logs, backup, static assets 3. Amazon EFS (Elastic File System) Size Limit Virtually unlimited Persistence Persistent and shared across Lambda and EC2 Use Cases Large datasets, ML models, shared files 4. Amazon DynamoDB Size Limit Per-item limit: 400 KB Persistence Highly available and persistent Use Cases Key-value or document store, state storage 5. Amazon RDS / Aurora Size Limit Depends on DB engine Persistence Persistent and durable Use Cases Complex queries, transactional data

Answer 2

1. We can have upto 1000 concurrent execution of Lambda. If we need more we can raise a support ticket 2. We can set a reserved concurrency at function level ie set a limit for the number of concurrent executions 3. If the limit exceeds, it will throw 429 status code for synchronous invocations. For asynchronous invocations it will retry automatically and then go to DLQ. The retry will happen as exponential backoff starting from 1 second to 5 minutes 4. Concurrency limit applies to all the Lambda functions in our account. So if there is too much load on our application and suppose one Lambda function is behind an Application Load Balancer. Now if the function being invoked by ALB automatically scales to 1000 concurrent executions, then every other Lambda function in our application will start to throttle. Hence we have to carefully set this reserved concurrency limit for our Lambda functions

Answer 3

Cold start refers to the initial delay that occurs when AWS Lambda has to: • Set up a new execution environment, • Initialize your function code (including any dependencies), • Then run your function handler. This typically happens when: • Your function is invoked after a period of inactivity, • AWS scales out by creating new instances to handle increased traffic. Provisioned concurrency is an AWS feature that pre-warms Lambda instances, keeping them initialized and ready to respond immediately—eliminating cold starts. Key aspects: • You specify the number of pre-initialized instances. • AWS keeps those instances always ready, even before invocation. • Helps achieve consistent low-latency performance, ideal for latency-sensitive applications. How it’s billed: • You pay for the amount of provisioned concurrency (per GB-second), • Plus the usual Lambda invocation charges.

Answer 4

1. Include Dependencies in the Deployment Package • Best for small functions or when using runtimes like Node.js, Python, or Java. • Steps: • Create a project directory. • Install dependencies locally (npm install, pip install -t, etc.). • Zip the entire directory, including the dependencies and your function code. • Upload the zip to Lambda (via console, CLI, or CI/CD). 2. Use Lambda Layers • Best for sharing dependencies across multiple functions or separating dependencies from business logic. • Layers are zip archives containing libraries, custom runtimes, or configuration files. • A Lambda function can use up to 5 layers.

Answer 5

1. Basic CloudFormation Template for Lambda or Inline code for very basic functions 2. Uploading Lambda Code to S3 Zip your code, upload it to S3 and then deploy your Cloudformation stack

Answer 6

Deploy Lambda function as a container image for upto 10GB from ECR. Pack large and complex dependencies in the container. Base images are available for Python, Node.js, Java, .Net, Go, Ruby. We can create our own Lambda image as long as it implements Lambda Runtime API. 1. Build a Dockerfile for Lambda Each Lambda container image must: • Be based on an AWS Lambda base image • Contain your application code • Define the CMD as the Lambda handler 2. Build the Docker Image 3. Create ECR Repository 4. Tag and Push the Image to ECR

Answer 7

1. When we work on a Lambda function the version is $Latest 2. When we publish the version it becomes V1 3. Versions are immutable ie we can’t change the code once we publish 4. Subsequent published codes will increment the versions 5. A version of Lambda contains Code + Configuration 6. Each version of Lambda function can be accessed Aliases are pointers to the Lambda versions for example dev, test, prod Aliases allow Canary deployment

Answer 8

1. All-at-once Deployment • What it does: Instantly shifts 100% of the traffic from the old version to the new version. • Use case: Low-risk deployments, testing environments, or small functions where downtime is acceptable. • Rollback behavior: Entire traffic switches back if deployment fails. 2. Canary Deployment • What it does: Shifts a small portion of traffic to the new version initially, waits for a specified time, then shifts the rest. • Example: Canary(10%, 5 minutes): • 10% of traffic goes to the new version. • Waits 5 minutes. • If no errors, shifts remaining 90%. 3. Linear Deployment • What it does: Gradually shifts traffic to the new version in equal increments over time. • Example: Linear(10%, every 2 minutes): • Moves 10% of traffic every 2 minutes until it reaches 100%. We can create pre and post deployment hooks to monitor health of the Lambda function and rollback if there are errors

Answer 9

What is appspec.yml? It’s a YAML file that defines: • The Lambda function name • The alias used for traffic shifting • The target version (usually the version you are publishing and deploying) • Optional hooks (pre/post traffic shifting logic using other Lambda functions) Basic Structure of appspec.yml Explanation: • Name: The Lambda function name. • Alias: The Lambda alias to shift traffic (CodeDeploy only works with aliases). • CurrentVersion: The version currently associated with the alias (optional). • TargetVersion: The new published version to be deployed.

Answer 10

Lambda Function URL (introduced by AWS in 2022) is a built-in HTTPS endpoint that you can enable for your Lambda function without needing to configure API Gateway or any other proxy. When to Use Lambda Function URL • For simple, public-facing microservices • Internal tools and automation scripts • Webhooks or callbacks • Prototyping without API Gateway overhead Invoke example curl https://.lambda-url..on.aws/

Answer 11

To integrate AWS Lambda with Amazon CodeGuru Profiler, you can use it to analyze performance bottlenecks and optimize Lambda functions written in Java or Python. Step 1: Ensure Lambda uses a Supported Runtime • Java 8/11 (Corretto) • Python 3.8 or later Step 2: Add CodeGuru Profiler SDK to Your Code Step 3: Create a Profiling Group in CodeGuru Step 4: Grant IAM Permissions to Lambda

Answer 12

These limits are PER REGION Lambda Execution Limits • Memory allocation ranges from 128 MB to 10,240 MB (10 GB), adjustable in 1 MB increments. • Maximum execution timeout is 15 minutes (900 seconds). • Ephemeral /tmp disk storage is 512 MB by default, can be increased up to 10 GB. • Environment variables size is limited to 4 KB total (key + value). • Maximum response/request payload size is: • 6 MB for synchronous invocation • 256 KB for asynchronous event sources (e.g., SQS, SNS) • File descriptors limit is 1,024 per function instance. • Maximum execution threads per function instance is 1,024. • Lambda execution role (IAM) is limited to one role per function. ⸻ Lambda Deployment Limits • Deployment package size (direct upload via console) must be ≤ 50 MB (zipped). • Deployment package size (via S3) can be up to 250 MB (zipped). • Unzipped code + dependencies size must be ≤ 250 MB. • Maximum number of Lambda layers per function is 5. • Maximum size per Lambda layer is 250 MB (unzipped). • Maximum number of versions per function is 1,000. • Lambda function name length is limited to 64 characters.

Lambda Flashcards

(36 cards)