SOA-C03 domain - 22% of the exam

Deployment, Provisioning, and Automation

Deployment, Provisioning, and Automation is 22% of the SOA-C03 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

82
Practice questions
7
Objectives
58
Medium
24
Hard

Where people go wrong

  • Swapping drift detection and change sets

    The wrong answers offer drift detection to preview what a new template will do, or offer a change set to reveal edits made directly in the console. Drift detection compares live resources against the template and reports the properties that differ, changing nothing. A change set compares a submitted template with the deployed stack and lists each add, modify and replace action before anything runs. A console edit with no template change produces no change set entries.

    See it in a sample question
  • Treating one CloudFormation protection as covering every case

    The wrong answers reach for termination protection, UpdateReplacePolicy or DeletionPolicy as though any of them guards a resource in every situation. Termination protection only blocks deleting the whole stack. UpdateReplacePolicy acts only when an update replaces a resource, not on stack deletion. DeletionPolicy Retain keeps a resource when the stack is deleted but does not stop an update replacing it; a stack policy that denies Update:Replace and Update:Delete covers updates. Neither DeletionPolicy nor termination protection stops the rollback of a failed create, while creating the stack with rollback disabled does.

    See it in a sample question
  • Mixing up StackSets, Resource Access Manager and Service Catalog

    The wrong answers use StackSets where one shared resource is wanted, which creates a separate copy in every account, or use Service Catalog or Resource Access Manager where a template must be deployed to every account. StackSets with service-managed permissions deploy a template to organisational units. Resource Access Manager shares an existing resource such as a Transit Gateway or subnets. Service Catalog gives developers controlled self-service launches of approved products.

    See it in a sample question
  • Picking the wrong Systems Manager capability

    The wrong answers swap Run Command, State Manager, Patch Manager and Session Manager. Run Command runs a document once, on demand, across tagged instances. State Manager binds a document to a tag target and reapplies it on a schedule to hold the desired state. Patch Manager adds a patch baseline, a maintenance window and compliance reporting. Session Manager opens an interactive shell on one instance and neither fans commands out nor enforces configuration.

    See it in a sample question
  • Putting a setting in the wrong Image Builder construct

    The wrong answers assign a job to a construct that cannot do it: customisation steps in the infrastructure configuration, the subnet and instance profile in the recipe, or sharing in a component. Components hold versioned build and test steps. The image recipe pins the base image and the ordered component versions. Infrastructure configuration sets the instance type, subnet, security group and instance profile. Distribution settings name the target Regions and accounts.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleDeployment, Provisioning, and Automationmedium

A stack was provisioned by CloudFormation, but an engineer later edited a security group rule directly in the console during an incident. The team now wants CloudFormation to report which managed resources no longer match their template definitions so the configuration can be reconciled. Which CloudFormation feature surfaces these out-of-band changes?

  • ARun drift detection on the stack, which compares each managed resource against its template definition and reports the properties that differ. Correct
  • BCreate a change set for the stack, which previews how a newly submitted template would change the deployed resources before you execute it.
  • CEnable a stack policy on the stack, which only governs which resources a subsequent stack update is permitted to modify and changes nothing else.
  • DTurn on rollback triggers on the stack, which monitor CloudWatch alarms during an update and roll the whole stack back automatically on failure.
Use CloudFormation drift detection to identify managed resources whose live configuration no longer matches the template. Drift detection inspects the actual configuration of each resource a stack manages and compares it to the values recorded in the template, then classifies each resource as in sync or drifted and lists the specific properties that differ. This is the only listed feature that examines live state for changes made outside CloudFormation, such as a security group rule edited directly in the console.

Why A is correct: Drift detection is purpose-built to compare live resource configuration with the expected template values and flag any property that was changed outside CloudFormation, which is exactly what reconciling a manual console edit needs.

Why B is wrong: A change set diffs a new template you submit against the deployed stack; it does not inspect the live resource state, so a manual console edit with no template change produces no entries.

Why C is wrong: A stack policy only restricts which resources an update may modify; it performs no comparison and generates no report of out-of-band changes made through the console.

Why D is wrong: Rollback triggers watch CloudWatch alarms during a stack operation to roll back on failure; they never inspect resource configuration for manual edits, so they cannot surface drift.

Free sampleDeployment, Provisioning, and Automationmedium

A CloudFormation template defines a production Amazon RDS database alongside other resources. The team needs the stack to keep the database intact if the stack is ever deleted, while still allowing the rest of the stack to be removed normally. Which template setting on the database resource achieves this outcome?

  • AAdd an UpdateReplacePolicy of Retain to the database so the underlying instance is preserved instead of being removed during a delete operation.
  • BAdd a CreationPolicy with a timeout to the database so the underlying instance is preserved instead of being removed during a delete operation.
  • CAdd a DependsOn attribute to the database so the underlying instance is preserved instead of being removed during a delete operation.
  • DAdd a DeletionPolicy of Retain to the database so the underlying instance is preserved instead of being removed during a delete operation. Correct
Set a DeletionPolicy of Retain on a stateful resource so CloudFormation preserves it when the stack is deleted. DeletionPolicy controls what CloudFormation does with a resource specifically when the stack is deleted or the resource is removed from the template. Setting it to Retain leaves the resource untouched in the account during a delete, which protects a stateful resource such as an RDS database while every other resource in the stack is removed as usual.

Why A is wrong: UpdateReplacePolicy only governs what happens to the old resource when an update would replace it; it does not control behaviour during a stack delete, so the database would still be removed.

Why B is wrong: A CreationPolicy only makes CloudFormation wait for success signals while a resource is created; it has no effect on stack deletion, so it cannot keep the database intact.

Why C is wrong: DependsOn only sequences the order in which resources are created or deleted; it does not exempt any resource from deletion, so the database would still be removed with the stack.

Why D is correct: A DeletionPolicy of Retain tells CloudFormation to leave the resource in place when the stack is deleted, so the database survives while the rest of the stack is removed normally.

Free sampleDeployment, Provisioning, and Automationhard

A central cloud team governs roughly sixty accounts through AWS Organizations and must deploy one identical baseline template that creates a guardrail IAM role and a logging bucket into every existing member account and into any account created later, without an engineer touching each account. They want the rollout and ongoing template updates driven from a single operation. Which approach meets this requirement with the least manual effort?

  • AShare the baseline role and bucket from the management account using AWS Resource Access Manager so every member account references the central resources directly.
  • BDeploy the template with CloudFormation StackSets using service-managed permissions and enable automatic deployment to the target organizational units. Correct
  • CPublish the baseline template as an AWS Service Catalog product and rely on each account team to launch the product so the role and bucket are created locally.
  • DRun a self-managed CloudFormation StackSet and add each member account number to the deployment targets whenever the organization onboards a new account.
Use service-managed CloudFormation StackSets with automatic deployment to provision a baseline into all current and future Organizations accounts from one operation. Service-managed StackSets integrate with AWS Organizations so you target organizational units rather than individual account numbers, and the automatic deployment option makes CloudFormation create the stack instance in any account that joins a targeted unit and remove it when an account leaves. That combination rolls the baseline out everywhere and keeps new accounts covered without per-account work, which self-managed mode and the sharing or self-service options cannot do.

Why A is wrong: Resource Access Manager shares an existing resource for cross-account use, but an IAM role and a bucket created per account are not shareable that way, so this neither provisions copies nor covers future accounts automatically.

Why B is correct: Service-managed StackSets target organizational units through Organizations, and automatic deployment creates a stack instance in every current member account and in any account added to the unit later, all from one administrator operation.

Why C is wrong: Service Catalog gives teams self-service launches, but it depends on someone in each account choosing to provision the product, so it does not guarantee identical deployment everywhere or reach new accounts on its own.

Why D is wrong: Self-managed StackSets work by explicit account lists and cross-account roles, so an engineer must edit the target list and trust roles for every new account, which fails the no-touch requirement for future accounts.

Free sampleDeployment, Provisioning, and Automationmedium

An operations team must run a one-off shell script that rotates a log file on roughly two hundred managed EC2 instances tagged Environment=prod, and they need the exit status and output captured centrally for each instance. They want to trigger the action on demand without opening SSH or building any new infrastructure. Which Systems Manager capability is designed for this ad hoc, fleet-wide command execution?

  • AUse State Manager to associate the script with the tagged instances so it runs on the fleet and records the output centrally for each instance.
  • BUse Patch Manager with a patch baseline so the script is applied to the tagged instances on the fleet and records the output centrally for each instance.
  • CUse Session Manager to open a shell on the tagged instances so the script runs across the fleet and records the output centrally for each instance.
  • DUse Run Command targeting instances by the Environment tag so the script runs across the fleet and records the output centrally for each instance. Correct
Use Systems Manager Run Command to execute an on-demand action across tag-selected managed instances and capture per-instance output centrally. Run Command is the Systems Manager feature for ad hoc operational actions: an operator picks a command document, targets managed instances by tag or resource group, and the SSM agent runs it on each instance and reports status and output back to the console and to optional S3 or CloudWatch Logs. It needs no inbound SSH and no new servers, which is why it fits a one-off log-rotation across a tagged fleet better than the scheduled or patch-specific features.

Why A is wrong: State Manager enforces a configuration on a schedule to keep instances compliant over time; it is built for recurring desired-state runs, not a single on-demand action triggered when the operator decides to rotate the logs.

Why B is wrong: Patch Manager only scans for and installs operating system and application patches against a baseline; it has no mechanism to run an arbitrary log-rotation script, so it cannot perform this task.

Why C is wrong: Session Manager gives an interactive shell to one instance at a time for troubleshooting; it does not fan a command out across a tagged fleet or aggregate per-instance exit status automatically.

Why D is correct: Run Command executes a chosen document against targets selected by tag on demand, returns per-instance status and output through the agent, and needs no SSH or extra infrastructure, which matches the one-off fleet task exactly.

Free sampleDeployment, Provisioning, and Automationmedium

An EC2 Image Builder pipeline builds a golden AMI in eu-west-1, and the resulting image must be made available, copied and shareable to two further Regions and to three other accounts in the organisation automatically as part of each pipeline run. Which Image Builder configuration achieves this distribution without a separate post-build copy step?

  • AAdd a final build component to the recipe that runs the AWS CLI copy-image and modify-image-attribute commands on the build instance to replicate the AMI to the other Regions and accounts.
  • BAttach a distribution settings configuration to the pipeline that lists the target Regions and the account IDs to share with, so Image Builder copies and grants launch permission on each run. Correct
  • CEnable cross-Region replication on the AMI the way it is enabled on an Amazon ECR repository so each new image version is mirrored to the additional Regions and shared accounts on creation.
  • DCreate an EventBridge rule on the image-build-succeeded event that triggers a Lambda function to copy the AMI to each Region and call ModifyImageAttribute to share it with the three accounts.
Use Image Builder distribution settings to copy a built AMI to target Regions and share it with named accounts automatically on each pipeline run. An EC2 Image Builder distribution settings resource is attached to a pipeline and declares, per Region, the AMI distribution targets and the launch permissions that grant named account IDs the right to launch the image. During each successful run Image Builder copies the AMI to every listed Region and applies those launch permissions, so cross-Region and cross-account availability happen as part of the pipeline rather than through a CLI build step, a post-build Lambda, or an ECR replication feature that does not apply to AMIs.

Why A is wrong: Build components run inside the build instance before the AMI exists, so they cannot copy the finished image or grant launch permissions, and this hand-rolled step duplicates what distribution settings provide.

Why B is correct: Distribution settings let the pipeline name target Regions and accounts, and Image Builder copies the AMI and sets the launch permissions during the run, achieving cross-Region and cross-account sharing automatically.

Why C is wrong: Cross-Region replication is an ECR repository feature for container images and does not exist for AMIs, so it cannot be enabled on the machine image this pipeline produces.

Why D is wrong: A Lambda triggered after the build can copy and share the AMI but adds a separate post-build workflow to maintain, which the pipeline's own distribution settings make unnecessary.

Other domains in this exam

See also the SOA-C03 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Amazon Web Services. Original, blueprint-aligned practice material only.