Learn cloud-service concepts, resource configuration and operational tradeoffs.
Before starting: Networking basics, HTTP, permissions, and cloud billing awareness.
Practice project: Plan and document a minimal cloud lab with permission checks and cleanup.
Cloud Foundations & Infrastructure
- What is Cloud Computing, Deployment Models & AWS Value Proposition
- AWS Global Infrastructure: Regions, Availability Zones (AZs), Edge Locations & Local Zones
Read the example and try it
A region contains several independent zones. List the zone IDs for one region and use those IDs to compare placement across accounts.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-availability-zones --region eu-west-1 --query "AvailabilityZones[].ZoneId"Compare the zone IDs for two subnets.
Two offline subnet records have AvailabilityZoneId values use1-az2 and use1-az4.
Expected result
Different zones. If the second ID were also use1-az2, they would identify the same zone even across accounts. - AWS Management Console, AWS CLI v2 Installation & AWS CloudShell
Read the example and try it
Check the identity that will make the request before inspecting a resource. The account and ARN tell you which credentials the CLI selected.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws sts get-caller-identity --query "{Account:Account,Caller:Arn}"Find the account ID and role ARN in a sample identity response.
Read this fictional identity: {"Account":"111122223333","Arn":"arn:aws:sts::111122223333:assumed-role/Lab/Session1"}.
Expected result
Account: 111122223333. Session ARN: arn:aws:sts::111122223333:assumed-role/Lab/Session1. The role name is Lab. - AWS Shared Responsibility Model & Well-Architected Framework (6 Pillars)
Read the example and try it
For an EC2 guest, the team running the application still manages its operating system. Inspect the image ID as a starting point for a patch inventory.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-instances --query "Reservations[].Instances[].{Instance:InstanceId,Image:ImageId}"Divide a server patch checklist into guest and host tasks.
An EC2 checklist lists installing a guest Linux security update and replacing defective host hardware.
Expected result
Customer: patch the guest OS. AWS: maintain the underlying host hardware. - AWS Pricing Fundamentals, Free Tier, Total Cost of Ownership (TCO) & Cost Explorer
Read the example and try it
Estimate a lab's compute charge with hours multiplied by a supplied hourly rate. These classroom rates omit storage, traffic, taxes and discounts.
Standalone Python 3 simulation. No AWS account or network call is needed.
hours = 12 example_hourly_rate = 0.25 print(hours * example_hourly_rate)Sketch separate steps for a spending alert and a response.
A $50 practice budget sends a notification at 80% of actual spend. No action is configured.
Expected result
Threshold: $40. The alert informs the owner. It does not shut down resources. A separate response could review cost drivers and stop a nonessential lab.
Identity, Access & Governance
- AWS Identity and Access Management (IAM): Users, User Groups & MFA
Read the example and try it
MFA adds a second check during sign-in. Inspect the devices registered to a named lab user before changing any credentials.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws iam list-mfa-devices --user-name lab-reader --query "MFADevices[].SerialNumber"Describe a sign-in using a password and a security key.
Sketch a login with a valid password and a registered hardware security key.
Expected result
Password accepted → second-factor challenge → user confirms with the registered key → session issued. A password alone is insufficient. - IAM Policies: Identity-Based, Resource-Based, JSON Policy Grammar & Conditions
Read the example and try it
An IAM statement names an action and the resource it applies to. This example grants bucket-level listing for just one example bucket.
IAM policy document for a fictitious S3 bucket. Inspect the policy before adapting it for a lab.
{"Version":"2012-10-17","Statement":[{"Effect":"Allow","Action":"s3:ListBucket","Resource":"arn:aws:s3:::example-study-bucket"}]}Trace an S3 request through an allow and a matching deny.
One policy allows GetObject on lab/* and another explicitly denies GetObject on lab/private/*. Test lab/private/notes.
Expected result
Denied. Both patterns match and the explicit deny wins. An object outside private needs its own applicable allow. - IAM Roles & Temporary Security Credentials (AWS STS, AssumeRole & Instance Profiles)
Read the example and try it
A role has a trust policy describing who can assume it. Read that part separately from the policies that grant access after assumption.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws iam get-role --role-name StudyReader --query "Role.AssumeRolePolicyDocument"Draw how an instance obtains and refreshes credentials.
An EC2 workload has an attached instance profile. Its first temporary credentials approach expiration.
Expected result
The SDK credential provider obtains role credentials through instance metadata, refreshes them before expiry, and signs later calls with the refreshed credentials. - AWS Organizations, Multi-Account Strategy & Service Control Policies (SCPs)
Read the example and try it
An organization can constrain accounts through service control policies. An SCP limits available permissions, while identities still need grants of their own.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws organizations list-policies --filter SERVICE_CONTROL_POLICY --query "Policies[].Name"Show why an SCP allowing S3 is insufficient by itself.
A member-account role has no S3 grants. An applicable SCP allows s3:GetObject; assume no other grants.
Expected result
GetObject is denied. The SCP permits a possible grant but supplies none. Add an applicable permission grant within the limits to permit access. - Governance & Compliance: AWS CloudTrail, AWS Config & AWS Audit Manager
Read the example and try it
To investigate who called an API, look for a management event in the relevant region. Configuration history answers a separate question about changes in resource state.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws cloudtrail lookup-events --max-results 5 --query "Events[].{Action:EventName,Who:Username}"Identify the actor, time and action in a sample audit event.
Read this sample event: {"userIdentity":{"arn":"arn:aws:iam::111122223333:user/ana"},"eventTime":"2025-01-02T10:00:00Z","eventName":"CreateBucket"}.
Expected result
Actor: user/ana in account 111122223333. UTC time: 10:00. Action: CreateBucket.
Virtual Networking & Content Delivery
- Amazon VPC Architecture: CIDR Blocks, Subnets (Public vs Private) & Route Tables
Read the example and try it
A subnet occupies part of a VPC address range. Compare each subnet prefix with its VPC ID before choosing an address for an instance.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-subnets --query "Subnets[].{Range:CidrBlock,Vpc:VpcId,Zone:AvailabilityZone}"Route 10.2.1.7 using a /16 route and a default route.
The routes are 10.2.0.0/16 → peer and 0.0.0.0/0 → internet gateway.
Expected result
10.2.1.7 follows the /16 route to the peer. An address such as 8.8.8.8 follows the default. - VPC Connectivity: Internet Gateways (IGW), NAT Gateways & Egress-Only Gateways
Read the example and try it
A route lists a destination and its next hop. Look for the default IPv4 route and identify whether it points to an internet gateway or a NAT gateway.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-route-tables --query "RouteTables[].Routes[?DestinationCidrBlock=='0.0.0.0/0']"Trace a private subnet request through NAT to the internet.
Draw one outbound HTTPS request from a private IPv4 instance through a public NAT gateway.
Expected result
Instance → private route table → NAT gateway in a public subnet → internet gateway → server. The reply follows the existing translation back to the client. - VPC Security: Security Groups (Stateful) vs. Network Access Control Lists (NACLs, Stateless)
Read the example and try it
Security groups follow connections. Network ACLs inspect packets crossing the subnet boundary in each direction, so return traffic needs a matching rule.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-network-acls --query "NetworkAcls[].Entries[].{Number:RuleNumber,Outbound:Egress,Action:RuleAction}"Mark both directions of a TCP exchange on a subnet diagram.
A client connects from port 51000 to server port 443. Mark the stateless subnet ACL on the server side.
Expected result
Inbound allows traffic to destination port 443. Outbound allows the reply to the client on destination port 51000 or its permitted ephemeral range. - VPC Peering, Transit Gateway & VPC Endpoints (Interface PrivateLink vs Gateway)
Read the example and try it
A VPC endpoint gives services a path that can avoid a public internet route. Check the endpoint's service and VPC before diagnosing DNS or policy problems.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-vpc-endpoints --query "VpcEndpoints[].{Service:ServiceName,Vpc:VpcId}"Draw the missing connection in a three-VPC design.
A peers with B and B peers with C. A must reach C directly using peering.
Expected result
Add an A–C peer with nonoverlapping CIDRs, routes on both sides and suitable security rules. A→B→C is not a transitive peering path. - Edge Networking: Amazon Route 53 DNS Routing Policies & Amazon CloudFront CDN
Read the example and try it
DNS sends a client toward an endpoint. A distribution can then serve a cached object near that client. Inspect the distribution's assigned domain.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws cloudfront list-distributions --query "DistributionList.Items[].DomainName"Trace a first request and a cache hit for the same image.
A cacheable image is requested twice at the same edge before its TTL expires; begin with an empty cache.
Expected result
First request: edge miss → origin → edge → visitor. Second: edge hit → visitor. The second response needs no origin download.
Elastic Compute Services
- Amazon EC2 Fundamentals: Instance Types, AMIs, Key Pairs & User Data Bootstrapping
Read the example and try it
An AMI supplies the initial disk image. Two instances launched from one image can subsequently have different files and patch levels.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-images --owners self --query "Images[].{Name:Name,Image:ImageId}"Separate boot image settings from first-boot commands.
One setting chooses an OS disk, another installs a package on first boot.
Expected result
The AMI supplies the initial disk. User data can run the package installation at startup. Check boot logs to confirm the install succeeded. - EC2 Purchasing Options: On-Demand, Reserved (RI), Savings Plans & Spot Instances
Read the example and try it
A flexible batch job can retry interrupted work. Check the instance lifecycle field when investigating whether a worker was launched as Spot.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-instances --query "Reservations[].Instances[].{ID:InstanceId,Lifecycle:InstanceLifecycle}"Describe how a batch job could resume after interruption.
A batch has ten chunks and has committed results for chunks 1–6 when a Spot worker stops.
Expected result
A replacement reads the checkpoint and resumes with chunk 7. Retrying a completed chunk must leave a single committed result for it. - Elastic Load Balancing (ELB): ALB, NLB & Gateway Load Balancer (GWLB)
Read the example and try it
A load balancer receives traffic before its targets. Inspect its type and scheme to see which kind of entry point your application exposes.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws elbv2 describe-load-balancers --query "LoadBalancers[].{Name:LoadBalancerName,Type:Type,Scheme:Scheme}"Route /images and /api to two target groups on paper.
Set path rules for /images/* and /api/* on an HTTP Application Load Balancer.
Expected result
GET /images/logo.png → image targets. GET /api/rooms → API targets. Specify a default action for paths matching neither rule. - Amazon EC2 Auto Scaling: Launch Templates, Scaling Policies & Lifecycle Hooks
Read the example and try it
Desired capacity is the group's target. Compare it with the minimum and maximum to understand the range available to a scaling policy.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws autoscaling describe-auto-scaling-groups --query "AutoScalingGroups[].{Goal:DesiredCapacity,Low:MinSize,High:MaxSize}"Separate the settings for what to launch and when to launch it.
A launch template points to the AMI ID for revision lab-v2. The group has one instance and a maximum of three. A CloudWatch alarm invokes a policy that adds one instance.
Expected result
Desired capacity rises from 1 to 2. The new instance uses the AMI ID from the group's selected launch template version. The alarm and policy determine when capacity changes. Check the instance health separately. - Amazon Lightsail & AWS Elastic Beanstalk: PaaS Deployments & Rapid Prototyping
Read the example and try it
A managed environment bundles deployment and infrastructure tasks. List its health and status before looking inside an individual instance.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws elasticbeanstalk describe-environments --query "Environments[].{App:ApplicationName,Health:Health,State:Status}"List what you would check after deploying a new app version.
Release revision B to a previously healthy environment running A.
Expected result
Verify the deployed revision is B, target health is good, the public endpoint passes a functional check, and logs show no startup failure.
Scalable Storage Solutions
- Amazon S3 Architecture: Buckets, Objects, Versioning & Presigned URLs
Read the example and try it
An object key identifies a file-like item. With versioning enabled, saving another version can keep earlier contents recoverable.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws s3api get-bucket-versioning --bucket example-study-bucketSketch one key after two uploads with versioning enabled.
Enable versioning before writing A then B to the same key notes.txt.
Expected result
Two versions of notes.txt remain. An ordinary GET returns B. A GET with the first version ID returns A. Version IDs vary. - Amazon S3 Storage Classes: Standard, Intelligent-Tiering, Glacier & Lifecycle Rules
Read the example and try it
The cost of storing a file depends partly on when you need it back. Read an object's storage class before planning retrieval.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws s3api head-object --bucket example-study-bucket --key notes/week1.txt --query StorageClassCompare storage choices for tomorrow's restore and a yearly archive.
One file must be readable immediately tomorrow; another can take hours to restore once a year.
Expected result
One workable choice is S3 Standard for the first file and S3 Glacier Flexible Retrieval for the archive. For the archive, request a temporary restored copy and wait until it is ready before reading it. Check retrieval time, charges and minimum storage duration against the requirement. - S3 Security & Compliance: Bucket Policies, S3 Block Public Access & KMS Encryption
Read the example and try it
Block Public Access can prevent public sharing even if someone proposes a permissive bucket policy. Check the bucket's four settings together.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws s3api get-public-access-block --bucket example-study-bucketReview a hypothetical public grant alongside the block settings.
A proposed bucket policy grants public read while BlockPublicPolicy is active for that bucket.
Expected result
The attempt to set that public policy is rejected. Turning this block off would allow evaluation of the policy change; it would not itself grant public access. - Amazon Elastic Block Store (EBS): Volume Types (gp3, io2, st1), Snapshots & DLM
Read the example and try it
A snapshot records a volume's state for recovery. Inspect the source volume and completion state before using a snapshot in a recovery plan.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ec2 describe-snapshots --owner-ids self --query "Snapshots[].{Source:VolumeId,State:State}"Plan a restore to a separate volume and describe a verification step.
An EBS snapshot contains report.txt with known contents “April totals”. Restore to a separate test instance.
Expected result
Create a volume from the snapshot in the test instance's zone, attach and mount it, then read and compare report.txt. The source volume is untouched. - Amazon Elastic File System (EFS) & AWS Storage Gateway
Read the example and try it
A shared file system can serve multiple clients. Mount targets provide network entry points, so check their subnets when a client cannot connect.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws efs describe-mount-targets --file-system-id fs-EXAMPLEShow where both clients mount the same file system.
Two Linux clients mount the same EFS file system under /mnt/shared with access to its mount target.
Expected result
Client A writes hello.txt and closes it. Client B opens /mnt/shared/hello.txt and reads the same contents, given suitable permissions.
Cloud Databases & Caching
- Amazon Relational Database Service (RDS): Multi-AZ & Read Replicas
Read the example and try it
A standby helps recovery. A read replica can handle eligible read traffic. Inspect the instance configuration before pointing reports at a standby address.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws rds describe-db-instances --query "DBInstances[].{Name:DBInstanceIdentifier,MultiAZ:MultiAZ,Readers:ReadReplicaDBInstanceIdentifiers}"Separate a read-scaling requirement from a failover requirement.
An RDS application needs more read capacity and also automatic recovery from writer failure.
Expected result
For a Multi-AZ DB instance, the standby is for failover. Send reports to a separate read replica. A supported Multi-AZ DB cluster can instead provide readable instances and automatic failover. Check read capacity and writer recovery separately. - Amazon Aurora: Architecture, Global Databases & Serverless v2
Read the example and try it
An Aurora cluster exposes a writer address and a reader address. Choose an endpoint by what the application needs to do.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws rds describe-db-clusters --query "DBClusters[].{Write:Endpoint,Read:ReaderEndpoint}"Label the writer and reader connections in a cluster diagram.
An Aurora cluster has a writer and two available read replicas. Route an order update and a read-only report. What changes if you remove both replicas?
Expected result
Send the update to the cluster (writer) endpoint. Use the reader endpoint for the report. It balances new connections across available replicas, which can lag behind the writer. With no replicas the reader endpoint connects to the writer. Use a database role with read-only privileges for the report. - Amazon DynamoDB: NoSQL Foundations, Partition/Sort Keys, RCU/WCU & DAX
Read the example and try it
A partition key groups items for lookup. Inspect the key schema before designing a query for a pupil's attempts.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws dynamodb describe-table --table-name StudyAttempts --query Table.KeySchemaCompare a constant key with a per-customer key for a busy service.
Compare keys "all-customers" and customerId on 10,000 writes spread across 1,000 active customers.
Expected result
The constant key concentrates all traffic. Per-customer keys spread that workload, although one exceptionally busy customer can still create a hot key. - DynamoDB Advanced: Secondary Indexes (LSI vs GSI), Streams, TTL & Global Tables
Read the example and try it
An index gives a table another lookup path. Check its key schema and status before using it in a new query.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws dynamodb describe-table --table-name StudyAttempts --query "Table.GlobalSecondaryIndexes[].{Name:IndexName,Keys:KeySchema,State:IndexStatus}"Trace an update from a table item to a change consumer.
Enable a DynamoDB stream with new images and update item A from 4 seats to 5.
Expected result
A MODIFY record identifies A and contains the new value 5. A consumer can process the record; deduplicate retries before external side effects. - Amazon ElastiCache: In-Memory Caching with Redis & Memcached
Read the example and try it
A cache holds frequently read values. Decide what the application should do when a cached timetable expires.
Standalone Python 3 simulation. No AWS account or network call is needed.
cache = {"route-A": "09:15"} print(cache.get("route-B", "Look up the timetable"))Draw the hit and miss paths for a product lookup.
Item 7 exists in the database at price 12. Begin with an empty cache.
Expected result
First lookup: miss → database returns 12 → cache stores 12. Second lookup before expiry: hit → 12, without another database lookup.
Serverless Compute & Event-Driven Architectures
- AWS Lambda: Serverless Compute Foundations & Execution Model
Read the example and try it
A Lambda handler receives an event for one invocation. Test its calculation locally before connecting an event source.
Standalone Python 3 simulation. No AWS account or network call is needed.
def handler(event, context): return {"remaining": event["capacity"] - event["booked"]} print(handler({"capacity": 12, "booked": 7}, None))Choose durable storage for a counter that survives fresh invocations.
Two fresh Lambda environments each receive one increment for the same counter initially at zero.
Expected result
Use a durable shared counter with atomic increments, for example DynamoDB UpdateItem. Two distinct successful increments produce 2. Handle duplicate events separately. - AWS Lambda: Configuration, Memory, Ephemeral Storage & Lambda Layers
Read the example and try it
Memory and timeout are properties of a function configuration. Check the current values before comparing two performance measurements.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws lambda get-function-configuration --function-name study-summary --query "{MemoryMB:MemorySize,Seconds:Timeout}"Separate a library packaging problem from a timeout problem.
A function fails to import a library; a second function imports it but runs out of time.
Expected result
For the import error, deploy the dependency in the package or a compatible layer and test the import again. For the timeout, measure the slow operation and check both the function timeout and downstream deadlines. Fix the bottleneck or adjust the deadline if appropriate. - Amazon API Gateway: REST APIs, HTTP APIs & WebSocket Endpoints
Read the example and try it
An API route maps an HTTP method and path to an integration. List route keys to distinguish GET from POST on the same path.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws apigatewayv2 get-routes --api-id EXAMPLE --query "Items[].RouteKey"Sketch how the server could push a new chat message.
Two signed-in WebSocket clients join room-7 and one sends “hello”.
Expected result
Authenticate and authorize membership, look up active connection IDs, and send the message using the management API. Remove IDs that are no longer connected. - Message Decoupling: Amazon SQS (Standard vs FIFO) & Amazon SNS
Read the example and try it
A queue holds work until a consumer processes it. The count returned by SQS is approximate, so use several observations when diagnosing a backlog.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws sqs get-queue-attributes --queue-url https://sqs.eu-west-1.amazonaws.com/111122223333/study --attribute-names ApproximateNumberOfMessagesDraw two queues behind one topic.
Billing and email each require a copy of order 17.
Expected result
Topic → billing queue and email queue via separate subscriptions. Each queue receives its own message and is acknowledged independently. - Event-Driven Orchestration: Amazon EventBridge & AWS Step Functions
Read the example and try it
An event rule selects messages by their fields. This example pattern matches only quiz submissions from the classroom application.
EventBridge event-pattern JSON. Use a matching sample event to test the pattern.
{"source":["study.classroom"],"detail-type":["QuizSubmitted"]}Model retry and failure branches for a two-step order.
Reserve stock then charge. The first charge times out.
Expected result
Retry the charge within a bounded policy using the same idempotency key. On final failure enter a failure or compensation step, for example release the reservation after checking payment state.
Containers & Modern Application Delivery
- Amazon Elastic Container Registry (ECR): Repositories, Image Scanning & Lifecycle Policies
Read the example and try it
A repository can contain many image versions. A digest identifies the content even when a tag is moved to a newer build.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ecr describe-images --repository-name study-web --query "imageDetails[].{Tags:imageTags,Digest:imageDigest}"Distinguish a repository upload receipt from a scan result.
An image push returns digest sha256:… and its vulnerability scan is still pending.
Expected result
The digest identifies the pushed image. Security status is pending. Examine the completed findings for that digest before recording a scan result. - Amazon Elastic Container Service (ECS): Task Definitions, Services & Fargate vs EC2
Read the example and try it
An ECS service tries to maintain the desired number of tasks. A gap between desired and running tasks is a reason to inspect service events.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws ecs describe-services --cluster study --services web --query "services[].{Goal:desiredCount,Running:runningCount}"Draw the relationship between a task definition and two running tasks.
A service runs task definition lab:4 with desired count 2.
Expected result
Two running tasks can use lab:4. They have separate task identities. Desired count 2 describes what the scheduler should maintain. - Amazon Elastic Kubernetes Service (EKS): Managed Control Plane & Worker Nodes
Read the example and try it
An EKS cluster has a managed control plane. Worker health and workload readiness need their own checks after the control plane becomes active.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws eks describe-cluster --name study --query "cluster.{State:status,Version:version}"Separate control-plane work from workload configuration.
A Deployment asks for two pods of image lab:v2.
Expected result
AWS maintains the EKS control plane. The workload owner supplies the image, manifests, permissions and capacity configuration. Healthy result: two ready replicas. - AWS App Runner & AWS Elastic Beanstalk for Containerized Microservices
Read the example and try it
App Runner exposes a service URL for a deployed application. Check the service status before debugging a request to that URL.
AWS CLI v2 in an existing App Runner customer account. New customers can read the example without creating an App Runner service.
aws apprunner list-services --query "ServiceSummaryList[].{Name:ServiceName,URL:ServiceUrl,State:Status}"Trace a source revision through build and health checks.
For an eligible existing App Runner service, assume automatic deployments are enabled. Revision B fails its health check.
Expected result
Trace source B → build → deployment → health checks. Build success alone does not establish that B is serving traffic. Confirm the deployment result and active healthy revision. - Service Discovery & Mesh: AWS Cloud Map & AWS App Mesh
Read the example and try it
Service discovery maps a stable service name to registered instances. Look at the registry before assuming that a container still has yesterday's address.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws servicediscovery list-services --query "Services[].{Name:Name,Id:Id}"Show how a client resolves a service name after an endpoint changes.
The registry changes from endpoint 10.0.1.4 to 10.0.1.9. Assume the old endpoint is deregistered and DNS TTL has expired.
Expected result
A fresh discovery lookup returns the new endpoint. Connections or cached answers held before expiry may still need renewal.
Observability, Security & Secrets Management
- Amazon CloudWatch: Metrics, Alarms, Logs & Metric Math
Read the example and try it
An alarm evaluates a metric over time. Read its current state together with its comparison rule before interpreting an alert.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws cloudwatch describe-alarms --query "MetricAlarms[].{Name:AlarmName,State:StateValue,Rule:ComparisonOperator}"Choose a period, threshold and missing-data treatment for an alarm.
A metric reports one CPU sample per minute. Alert on three consecutive one-minute averages above 80%; treat missing as missing.
Expected result
Samples 82, 85, 83 satisfy the alarm. A missing sample is not a measured zero; the eventual state depends on available datapoints and alarm evaluation. - AWS X-Ray: Distributed Tracing, Service Maps & Latency Profiling
Read the example and try it
A trace connects spans from one request. A long child span can show where a student waited even if the whole request succeeded.
Standalone Python 3 simulation. No AWS account or network call is needed.
spans_ms = {"request": 210, "database": 160, "format": 12} print(round(spans_ms["database"] / spans_ms["request"] * 100))Follow a trace ID through two services and mark a sampling decision.
Use trace T7 for a sampled request crossing API → worker; a second request is not sampled.
Expected result
The instrumented API and worker segments share T7. The second request may have no recorded trace even when successful. Do not count traces as total requests. - AWS Key Management Service (KMS): Customer Managed Keys & Envelope Encryption
Read the example and try it
Envelope encryption uses a data key for the data and a wrapping key to protect the data key. Inspect key metadata without requesting secret material.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws kms describe-key --key-id alias/study-data --query "KeyMetadata.{Use:KeyUsage,State:KeyState}"Draw where the encrypted data key is stored alongside the payload.
Encrypt one report with a generated data key.
Expected result
Store the encrypted report and encrypted data key together. Keep the wrapping KMS key protected. Discard the plaintext data key after use. - AWS Secrets Manager vs AWS Systems Manager Parameter Store
Read the example and try it
A service can inspect rotation settings without putting a secret into its logs. This command requests metadata about the named secret.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws secretsmanager describe-secret --secret-id study/db --query "{Name:Name,Rotation:RotationEnabled}"Outline how an application picks up a rotated credential.
A database accepts a rotated credential after the rotation workflow succeeds.
Expected result
The application refreshes its cached secret, opens a new connection and authenticates with the current version. Never print the credential as verification. - Cloud Perimeter Defense: AWS WAF, AWS Shield & Amazon GuardDuty
Read the example and try it
A web ACL filters requests using rules. Inspect regional ACLs in the region serving the application before investigating blocked requests.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws wafv2 list-web-acls --scope REGIONAL --region eu-west-1 --query "WebACLs[].Name"Separate blocking a request from investigating a threat finding.
A web request matches a blocking WAF rule, and GuardDuty separately raises a finding.
Expected result
WAF blocks the matching request. GuardDuty supplies a finding for investigation; blocking based on that finding requires a response mechanism.
Infrastructure as Code, CI/CD & Migration
- AWS CloudFormation: Stacks, Templates, Parameters & Change Sets
Read the example and try it
A change set previews how a stack update would affect resources. Review replacement decisions before applying a change.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws cloudformation describe-change-set --stack-name study --change-set-name preview --query "Changes[].ResourceChange.{Action:Action,Replacement:Replacement}"Identify a replacement in a sample change preview before execution.
Inspect this change-set excerpt: {"ResourceChange":{"Action":"Modify","LogicalResourceId":"Store","Replacement":"True"}}.
Expected result
Store is proposed for replacement even though Action says Modify. Review data retention and recovery before executing. A preview alone changes nothing. - AWS Cloud Development Kit (CDK): Construct Trees & Multi-Language Synthesis
Read the example and try it
CDK synthesis turns application constructs into a CloudFormation template. In a configured CDK project, inspect the template before deploying it.
Shell command inside an existing AWS CDK project.
cdk synth StudyStackInspect a synthesized template and locate a resource declaration.
Read this offline synthesis excerpt: {"Resources":{"Uploads":{"Type":"AWS::S3::Bucket"}}}.
Expected result
Logical ID Uploads declares an S3 bucket. Synthesis creates a template; it has not created a bucket in an AWS account. - AWS CodePipeline, CodeBuild & CodeDeploy: Cloud-Native CI/CD
Read the example and try it
A pipeline can halt between testing and release. Check the latest action status to find the stage that needs attention.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws codepipeline get-pipeline-state --name study-release --query "stageStates[].{Stage:stageName,Actions:actionStates[].latestExecution.status}"Put source, test, build and release steps in order.
Place the following stages: release, retrieve source, compile/package, run tests.
Expected result
One valid sequence: retrieve source → run tests → package a passing revision → release that artifact. For compiled tests, compile before running those tests. - Cloud Migration Framework: The 7 Rs & AWS Database Migration Service (DMS)
Read the example and try it
A database migration has its own progress and failure state. Compare task status with an application-level row-count check before switching readers.
AWS CLI v2. Use your own lab resource names and a profile with read permission in the intended account and region.
aws dms describe-replication-tasks --query "ReplicationTasks[].{Task:ReplicationTaskIdentifier,State:Status}"Compare a lift-and-shift move with replacing the application.
Compare moving an unchanged service to a VM with replacing it by a hosted equivalent.
Expected result
VM move with minimal changes: rehost. Buying the hosted replacement: repurchase. Both need data and cutover planning. - Disaster Recovery Strategies in the Cloud: Backup & Restore, Pilot Light, Warm Standby & Multi-Region Active-Active
Read the example and try it
A recovery plan needs a measurable loss limit. With backups every fifteen minutes, the last completed backup may omit nearly fifteen minutes of new work.
Standalone Python 3 simulation. No AWS account or network call is needed.
backup_minutes = [0, 15, 30] failure_minute = 43 print(failure_minute - max(backup_minutes))Give separate targets for restoring service and recovering recent writes.
Set a recovery objective of two hours and allow at most fifteen minutes of recent data loss.
Expected result
RTO = 2 hours. RPO = 15 minutes. In a drill, measure both elapsed restoration time and the timestamp of the latest restored write.
Plan a safe, measurable cloud practice lab
A cloud lab needs a learning goal and a cleanup plan. Price and free-tier eligibility depend on the service, region, account, and date, so verify them in the current provider documentation. A budget notification is not a universal hard spending cap. Avoid production accounts and never publish access keys in code or screenshots.
Planning exercise; verify current AWS pricing before creating resources
Goal: explain one storage permission
Resources: one practice bucket and synthetic text file
Checks: permitted action succeeds; unpermitted action fails
Cleanup: remove test objects and bucket when finished
Review: billing and resource inventoryCheck the output
This is a lab plan, not an executed deployment. Actual charges and permission results depend on account configuration.Describe how you will verify cleanup. Do not assume deleting one visible resource deletes every related resource.
Open the AWS learning path on DevStudyHub · Play coding games on this blog
No comments:
Post a Comment
Please do not enter any spam link in the comment box.