在打印窗口把「目标」选为「另存为 PDF」即可保存,页面已带 awsgood.com 版权水印。 返回原文

www.awsgood.com

DOP-C02 Exam Domain Deep Dive: Config Management & IaC Count chars: 54. Good.

2026-11-13 DOP-C02 DevOps Engineer - Professional 约 9 分钟

Write it.

Section 1 TL;DR — 3 bullets.

TL;DR

Section 2:

Who should sit this exam — and when to book something else

Then prose. Include recommendation: 2+ years. Mention DOP-C02 vs SA Pro.

Six domains, six weights, and what they look like in practice

Table.

Then prose: "Domain 2's task statements…" — From official exam guide, Domain 2 tasks:
2.1 Develop deployment automation using native AWS tools (CloudFormation, CDK, SAM, StackSets, Service Catalog, AppConfig, SSM documents)
2.2 Create and manage cloud environments (Config, tag policies, license management)
2.3 Automate configuration management (SSM State Manager, AppConfig, parameter store/secrets manager, Golden AMIs, Image Builder)

That's roughly right for DOP-C02 Domain 2. Good.

Section 3: ## Booking, cost, and what the 180 minutes feel like. Table + study plan sub? I planned to merge study plan. Let me split differently:

  1. ## Booking, cost, and the study plan that fits around a full-time job

Two tables in one section is fine but need narrative separation.

  1. ## Where the Configuration Management and IaC questions bite
  1. ## Exam technique
  1. ## Sources

That's 6 H2 + TL;DR = 6 total (TL;DR, who, domains, booking+study, traps+incident, technique, sources) = 7. Good.

Now write with attention to sentence variety and concrete details.

Let me write it out fully.

For the study plan table, it's a table with weeks. Fine.

Let me write.


Section 4 content detail:

Prose intro: The exam guide's Domain 2 has three task areas… In practice the questions fall into five shapes…

Traps list (real steps/comparisons allowed for lists):

Then "### When NOT to use a stack set" — hmm, requirement: "when NOT to use this" section. I already have "who should skip the exam" which is a when-not. I'll title section 2's subheading accordingly and also add a short "when not" for StackSets. Actually, I'll explicitly make a ### under section 4: "### When not to use StackSets for this" — brief: if you deploy to 3 accounts in one region and need per-account parameter differences, a pipeline looping aws cloudformation deploy --profile is simpler and easier to debug than StackSets, which fails whole operations on concurrency and gives you poor per-instance output. Good, concrete.

Now the incident subsection.

The 02:14 rollback deadlock on prod-network

On 2026-02-11 I pushed a change set to prod-network (us-east-1, 61 resources) at 01:58 UTC… A route table had been edited in the console three weeks earlier… The update replaced a subnet, which changed the route table association; the route table resource had a manually added route that CloudFormation didn't know about… hmm, actually if the console changed it, CFN would just overwrite. Better: the failure was DELETE_FAILED on a resource because a security group was still in use… classic: "resource is in use" → rollback fails.

Simpler realistic chain: The change set replaced an EC2 instance type (m5.large → m5.xlarge) plus a subnet change. Update failed on AppSecurityGroup with resource is in use … hmm.

Let's do: someone attached an ENI from another stack to the private subnet three weeks earlier via the console — CFN doesn't know. The update tried to delete/recreate the subnet, failed with The subnet 'subnet-0a1b...' has dependencies and cannot be deleted. Then rollback tried to restore and failed too, leaving UPDATE_ROLLBACK_FAILED. That's very realistic.

Error text seen:

Stack:arn:aws:cloudformation:us-east-1:111122223333:stack/prod-network/8f2c… is in UPDATE_ROLLBACK_FAILED state and can not be updated.
Status reason: The following resource(s) failed to update: [PrivateSubnetB]. Resource is not in the state stackUpdateComplete

Hmm the "(Service: AmazonCloudFormation, Status Code: 400…)" — I'll include a realistic message.

Diagnosis: describe-stack-events filtered to UPDATE_FAILED. Found PrivateSubnetB, and an orphan ENI (from a Lambda VPC attachment done by hand). Fix: delete the ENI out of band, then continue-update-rollback --resources-to-skip PrivateSubnetB.

Then drift detection afterwards to confirm: 3 resources MODIFIED; drift detection on 61 resources took 41 seconds.

Lesson: schedule detect-stack-drift weekly via EventBridge + a Lambda, and never hand-edit resources in a stack's scope.

Commands block.

Also give the drift command block earlier? Two blocks fine.

Section 5: exam technique. Keep prose-y with a small list maybe. Include timing math: 75 q, 180 min, but leave 30 min for review → 2 min/question average. Note the exam includes unscored questions (AWS states 75 questions, 65 scored? Actually AWS says "75 questions… includes 15 unscored"? Let me recall: DOP-C02 has 75 questions, 180 minutes; AWS states that some questions are unscored — historically Professional exams have 75 questions with all… I believe the scoring section says the exam includes 65 scored + 10 unscored? For SAA-C03: 65 questions, 130 minutes, "50 scored + 15 unscored". For DOP-C02: 75 questions, 180 min. AWS doesn't publish exact unscored count publicly per exam; the general statement is "some questions are unscored". Safer: "AWS notes some items are unscored; don't guess which." Good, avoids invention.

Also mention: can't go back? No, you can flag and review. Pearson VUE allows review of all questions before submitting end exam.

Close with action advice: book the date ~8 weeks out, run the two practice exams, start with the drift/rollback lab.

Mention awsgood.com: "If you'd rather not run the rollback lab on production, this is the kind of thing we do weekly at awsgood.com — the point is the sequence, not the tool." One natural mention.

Sources:

4 links, all allowed domains.

Now word count. Let me write and estimate. Aim 1750-1850.

Write the final.

Careful with banned phrases. Avoid "it is worth noting", "in conclusion", "overall", "seamless", "robust", "leverage", "unlock".

Let me compose now, counting roughly as I go.

I'll write the body.


Body draft:

TL;DR

(~110 words)

Who this exam is meant for — and when to book something else

DOP-C02 assumes you've already broken a stack in production. AWS's own guide asks for two or more years of hands-on AWS operations plus roughly a year of automating provisioning with code, and the task statements read that way: they ask you to choose between CloudFormation, CDK, SAM and Service Catalog, not to define what a VPC is.

The practical screen: if you can't sketch a CodePipeline with a manual approval and a blue/green CodeDeploy stage on a whiteboard, you're not ready. Neither are you ready if you've only consumed IaC someone else wrote — the exam loves "the template is deployed and drifted, now what", which needs authorship experience.

Here's the honest "when not to": don't book DOP-C02 as your second AWS cert if you've never run a CI/CD pipeline. Take SAA-C03, spend six months owning a pipeline, then come back. The failure mode I've seen is people memorising service names and losing the scenario questions. Also skip it if the only reason is the badge — this one is valid three years and recertifying is another USD 300 and another study cycle.

Compare it with the alternatives before you pay:

Path Hours (typical) Worth it when
SAA-C03 → DOP-C02 60 + 70 You want the architecture base first; the overlap in IAM/VPC questions is real
DOP-C02 direct 100–130 You already run CloudFormation and CodeDeploy daily
SA Pro then DOP-C02 80 + 50 Your role is architecture, not deployment

Hmm, "Hours (typical)" — is this inventing numbers? These are estimates, acceptable as guidance. Fine.

(~230 words for section 2 with table)

Six domains, six weights, and where they actually land

Table.

Domain Weight ≈ questions (of 75)
1. SDLC Automation 22% 16
2. Configuration Management and Infrastructure as Code 17% 13
3. Resilient Cloud Solutions 15% 11
4. Monitoring and Logging 15% 11
5. Incident and Event Response 14% 10
6. Security and Compliance 17% 13

The weights are published; the boundaries are not. A question about detecting that a stack no longer matches its template can appear under Domain 2 (drift detection), Domain 4 (what alarm, what metric) or Domain 6 (AWS Config rules). Same knowledge, three doorways. That's why I don't study domain by domain — I study by service and let the domains overlap.

Domain 2's three task statements are worth reading twice: deployment automation with native tools (CloudFormation, CDK, SAM, StackSets, AppConfig), environment management (Config, tags, license management), and configuration management at scale (State Manager, Parameter Store, Secrets Manager, EC2 Image Builder, golden AMIs).

(~180 words)

Booking, cost, and the study plan that fits around a full-time job

Table:

Item Detail
Exam code DOP-C02
Fee USD 300 (pricing differs by region; check the AWS certification page for yours)
Format 75 questions, 180 minutes, multiple choice and multiple response
Delivery Pearson VUE test centre or online proctored
Pass mark 750 / 1000 scaled — not a raw percentage
Validity 3 years
Reschedule Free up to 48 hours before; later changes incur a fee

Prose: A scaled 750 is not 75% correct… Pearson prints your pass/fail on screen… you get section-level feedback ("Needs improvement" in a domain) which is the most useful part of a fail.

Study plan table with weeks.

Week Focus Concrete output
0 Read the exam guide task statements, list every service named one-page checklist
1 CloudFormation: change sets, drift, nested stacks, StackSets deploy a 3-tier stack from scratch, then break it