What's changed: In-scope coverage: added App Runner, App2Container, AWS Copilot, Serverless Application Repository, CloudShell, AWS CLI/SDK, CodeStar to s3
1.2Deployment Strategies
Understand low-risk deployments—CodeDeploy, in-place/blue-green, rolling/canary/linear. Release while minimizing downtime and blast radius.
Deployment is about switching over safely. CodeDeploy offers multiple strategies for EC2/ECS/Lambda.
1.2.1Choosing a deployment method
- In-place: update existing instances gradually (rolling); low extra cost but mixed versions during rollout.
- Blue-green: build a parallel environment and switch all at once; instant rollback on issues (minimal downtime).
- Canary: send a small portion first → shift the rest if healthy.
- Linear: shift a fixed percentage at intervals. All-at-once shifts everything together.
Common on DOP-C02: parallel env + all-at-once switch + instant rollback = blue-green, validate from a small portion = canary, fixed percentage steps = linear, update existing gradually = in-place/rolling, plus automatic rollback on CloudWatch alarms. Choose blue-green or canary to minimize downtime/risk.
CodeDeploy integrates with CloudWatch alarms to auto-rollback when thresholds are breached mid-deploy. For Lambda/ECS, pre/post hooks can run smoke tests.
DOP-C02 probes which strategies apply per target (EC2/on-prem, ECS, Lambda) and their safety mechanisms. For EC2/on-prem, define hooks in AppSpec (appspec.yml) (BeforeInstall/AfterInstall/ApplicationStart/ValidateService, etc.), running scripts at each hook. Blue-green launches a new Auto Scaling group or ECS task set in parallel, shifts the load balancer (ALB/NLB) target group, and keeps the old environment briefly before termination—making instant rollback easy. For Lambda/ECS, choose predefined traffic-shifting configs like Canary10Percent5Minutes (10% for 5 minutes, then the rest) or Linear10PercentEvery1Minute, and run smoke tests via Lambda validation functions in BeforeAllowTraffic/AfterAllowTraffic hooks. Automated rollback centers on CloudWatch alarm integration—if an alarm enters ALARM mid-deploy, it reverts to the prior version automatically. For Lambda alone, alias weighted routing also achieves gradual shift. These are options to limit downtime and blast radius.
| Strategy | How it switches | Best for |
|---|---|---|
| In-place/rolling | Update existing gradually | Low extra cost; brief mixed versions OK |
| Blue-green | Switch all at once to a parallel env | Minimal downtime; instant rollback required |
| Canary | Small portion, then the rest | Quick validation in production |
| Linear | Fixed percentage at intervals | Gradual, observed migration |
Scenario: Roll out a new Lambda version safely rather than all at once, with auto-revert if error rate rises. → Choose CodeDeploy Canary10Percent5Minutes and run a light check in BeforeAllowTraffic. Set CloudWatch alarms on production metrics (error rate/latency); if they hit ALARM mid-deploy, auto-rollback fires. This caps blast radius at 10% and reverts within minutes if something breaks.
FAQ: Blue-green vs canary? If you want to switch wholesale to a new environment and revert entirely on issues, use blue-green (you pay for the parallel environment). If you want to validate on a slice of production traffic before widening, use canary/linear (shift traffic percentages within the same environment). Decide by whether the priority is minimal downtime or staged validation.
Exam trap: Claiming "auto-rollback is configured" does nothing if no CloudWatch alarm is associated—rollback triggers are alarms or deployment-failure events. Also, EC2 blue-green needs temporary extra resources (cost) for the new environment—under a "minimize cost" requirement, in-place/rolling may be the intended choice.
1.2.2Section summary
- Low-risk switch = blue-green / canary / linear
- CodeDeploy = EC2/ECS/Lambda + auto-rollback on CloudWatch alarms
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. You build a parallel environment, switch all at once, and roll back instantly on issues. Which strategy?
Q2. You send 10% of traffic to the new version first, then shift the rest if healthy. Which method?
Q3. You want auto-rollback if error rate exceeds a threshold mid-deploy. CodeDeploy integrates with what?