# Free SOA-C03 practice questions

> 10 free SOA-C03 practice questions with the correct answer and an explanation.

Page: https://devopscodelab.com/exams/aws-soa-c03/free-practice-questions/
Practice exams: https://devopscodelab.com/exams/aws-soa-c03/

### Question 1 of 10

An S3 bucket receives application logs that are read frequently for 30 days, occasionally for the next 60 days, and must be retained for 7 years for compliance but are almost never read after 90 days. Which lifecycle configuration is MOST cost-effective?

- A. Transition to S3 Glacier Instant Retrieval on day 0 and expire the objects after 7 years
- B. Keep objects in S3 Standard for 90 days, then One Zone-IA until expiry after 7 years
- C. Use S3 Intelligent-Tiering from day 0 with the archive tiers enabled, then expire after 7 years
- D. Standard-IA after 30 days, S3 Glacier Deep Archive after 90 days, expire after 7 years

**Answer:** D

**Explanation:** The access pattern is known, so lifecycle transitions to Standard-IA and then Deep Archive minimize cost while honoring the retention period. Keeping data in Standard for 90 days and then One Zone-IA for years costs far more than Deep Archive for data that is almost never read, and stores compliance data in a single AZ. Intelligent-Tiering adds per-object monitoring charges and is best for unknown patterns. Glacier Instant Retrieval on day 0 incurs retrieval charges during the frequent-access period.

### Question 2 of 10

Development EC2 instances that carry the tag Env=dev must be stopped at 20:00 and started at 07:00 on weekdays. The team wants a serverless, low-maintenance solution using AWS-managed automation. What should be implemented?

- A. AWS Compute Optimizer idle-resource recommendations configured to automatically stop and start the tagged instances on weekdays
- B. Auto Scaling scheduled actions that set desired capacity to 0 at 20:00 and back at 07:00, attached to each standalone dev instance
- C. EventBridge Scheduler schedules that run the AWS-StopEC2Instance and AWS-StartEC2Instance Automation runbooks against the tagged instances
- D. A CloudWatch alarm on low CPUUtilization for each dev instance with an EC2 stop action, plus a second alarm to start them in the morning

**Answer:** C

**Explanation:** EventBridge Scheduler can invoke Systems Manager Automation runbooks on cron schedules, and the AWS-provided start and stop runbooks can target instances by tag, with no servers to manage. Scheduled actions apply to Auto Scaling groups, not standalone instances. CloudWatch alarm EC2 actions can stop, terminate, reboot or recover an instance but cannot start one, and low CPU is not a time schedule. Compute Optimizer only recommends; it does not stop or start instances on a schedule.

### Question 3 of 10

A target group's health check uses the path /health, which queries the database. When the database is briefly slow, all targets fail health checks at once and the ALB stops sending traffic everywhere. What should the CloudOps engineer recommend?

- A. Disable health checks on the target group and rely on Auto Scaling EC2 status checks instead
- B. Change the health check to TCP on port 22 so it does not touch the database
- C. Use a shallow health check that verifies the application itself, and monitor dependencies separately
- D. Set the healthy threshold to 1 and the unhealthy threshold to 10 and keep the deep check

**Answer:** C

**Explanation:** Deep health checks that depend on shared dependencies can mark every target unhealthy at once; a shallow check tests the instance's own ability to serve while dependencies are alarmed on separately. Target group health checks cannot be disabled for instance targets, and EC2 status checks do not detect application failures. Tweaking thresholds only delays the same correlated failure. Checking SSH does not prove the application is serving.

### Question 4 of 10

An auditor asks the company to prove that its backups can actually be restored and to measure restore duration regularly, without manual restore exercises. Which AWS Backup feature should be used?

- A. AWS Backup Audit Manager frameworks only
- B. Amazon Data Lifecycle Manager policies
- C. AWS Elastic Disaster Recovery drills
- D. AWS Backup restore testing

**Answer:** D

**Explanation:** Restore testing runs scheduled restores of selected recovery points, records restore times and outcomes, and can clean up the restored resources. Audit Manager frameworks check that backup practices comply with controls but do not perform restores. Data Lifecycle Manager creates and retains snapshots without testing restores. Elastic Disaster Recovery drills test server replication for DR, not AWS Backup recovery points.

### Question 5 of 10

A CloudFormation stack contains an RDS DB instance. The team wants a final snapshot taken automatically if the stack is deleted or if an update requires replacing the DB instance. Which template settings should be used?

- A. A stack policy that denies Update:Replace and Update:Delete on the DB instance
- B. DeletionPolicy: Snapshot and UpdateReplacePolicy: Snapshot on the DB instance
- C. DeletionPolicy: Retain on the DB instance so it survives stack deletion
- D. DependsOn on the DB instance pointing to the DB subnet group and parameter group

**Answer:** B

**Explanation:** DeletionPolicy Snapshot creates a final snapshot when the resource is deleted with the stack, and UpdateReplacePolicy Snapshot does the same when an update replaces it. Retain keeps the live instance rather than snapshotting and does not cover replacement. DependsOn only controls ordering. A stack policy can block replacement but does not take snapshots.

### Question 6 of 10

An EC2 launch template contains user data that should run every time the instance boots, but the script only ran on the first launch. Why?

- A. Launch templates pass user data only at launch and not to restarted instances
- B. The instance profile lacks the permissions needed to rerun user data on boot
- C. User data is processed only once on Nitro instances because of the metadata service
- D. By default cloud-init runs user data scripts only on an instance's first boot

**Answer:** D

**Explanation:** By default, user data shell scripts run only at first boot; running on every boot requires cloud-init configuration changes or another mechanism such as State Manager. IAM permissions do not affect whether user data executes. Nitro instances behave like any other instance here. The limitation is cloud-init behavior, not the launch template, and the user data remains available to the instance after restarts.

### Question 7 of 10

An application uses KMS to encrypt records and must ensure that ciphertext from one tenant cannot be decrypted as if it belonged to another tenant. The decryption call should fail if the tenant ID does not match. What should be used?

- A. A multi-Region KMS key with one replica per tenant, where each replica's key policy allows decrypt only for that tenant's role
- B. Automatic rotation of the KMS key every 90 days, so each tenant's data ends up under a different backing key version
- C. An encryption context with the tenant ID, which must match on decrypt and can be required by key policy conditions
- D. A separate AWS Region per tenant, with each tenant's records encrypted under a KMS key created in its own Region

**Answer:** C

**Explanation:** Encryption context is additional authenticated data that must be supplied identically on decrypt, binding the ciphertext to the tenant ID; key policy conditions can require it. Rotation changes key material on a schedule and does not bind data to tenants. Multi-Region replicas share the same key material and are meant for cross-Region use, not tenant binding. Separate Regions isolate keys but do not make a decrypt call fail when the supplied tenant ID is wrong.

### Question 8 of 10

An S3 bucket must accept requests only through a specific VPC gateway endpoint, and the endpoint itself should only allow access to that company's buckets. Which combination achieves this?

- A. A network ACL that allows only the S3 prefix list, plus a bucket policy condition on aws:SourceIp matching the VPC's private CIDR
- B. A security group on the gateway endpoint that allows only the S3 prefix list, plus a bucket policy condition on aws:SourceVpce
- C. A bucket policy denying requests whose aws:SourceVpc is not the VPC ID, plus IAM policies allowing s3:* only on the company's buckets for the instance roles
- D. A bucket policy denying requests whose aws:SourceVpce is not the endpoint ID, plus an endpoint policy allowing only the company's bucket ARNs

**Answer:** D

**Explanation:** The aws:SourceVpce condition restricts the bucket to one endpoint, and the endpoint policy limits which buckets can be reached through it. Gateway endpoints do not have security groups. aws:SourceIp does not carry private VPC addresses for requests through a gateway endpoint. aws:SourceVpc allows any endpoint in the VPC rather than the specific one, and IAM policies on some roles do not stop the endpoint from reaching other companies' buckets.

### Question 9 of 10

Development VPCs and production VPCs are attached to the same Transit Gateway. They must both reach a shared-services VPC, but development and production must not reach each other. How should this be configured?

- A. Use one route table with all attachments associated and propagating, then add blackhole routes for the prod CIDR ranges so traffic toward production is dropped
- B. Enable appliance mode on the shared-services attachment so that traffic between dev and prod is forced through the shared-services VPC and dropped there
- C. Keep the default route table with every attachment propagating, and use security groups in the dev and prod VPCs to reject each other's CIDR ranges
- D. Separate Transit Gateway route tables: dev and prod attachments use tables with only shared-services routes, and the shared-services table has routes to both

**Answer:** D

**Explanation:** Transit Gateway route tables provide segmentation: the dev and prod route tables contain routes to shared services but not to each other, while the shared-services route table reaches both. Security groups alone leave every route open and depend on per-VPC rules staying correct. Appliance mode only keeps flows symmetric through one attachment; it does not filter traffic. Blackhole routes in a single shared route table would also stop the shared-services VPC from reaching production, because every attachment uses the same table.

### Question 10 of 10

Instances in a VPC must use the company's own DNS servers (10.1.0.10 and 10.1.0.11) and a custom domain suffix automatically at boot, instead of the Amazon-provided DNS. What should be configured?

- A. User data on each AMI that writes the company DNS servers and search domain into /etc/resolv.conf
- B. A Route 53 private hosted zone for the custom domain associated with the VPC, with records pointing at the company DNS servers
- C. A custom DHCP options set with the DNS servers and domain name, associated with the VPC
- D. A Route 53 Resolver outbound endpoint with a forwarding rule that sends all queries to 10.1.0.10 and 10.1.0.11

**Answer:** C

**Explanation:** DHCP options sets define the DNS servers and domain name handed to instances at boot, and associating a custom set with the VPC applies them to every instance. A private hosted zone holds records but does not change which resolver instances use. An outbound forwarding rule sends queries to the company servers but does not set the domain suffix on instances. Editing resolv.conf in user data is overwritten by the DHCP client and must be repeated for every AMI.

Ready for a timed run? Take the free 15-question SOA-C03 mini exam: https://devopscodelab.com/app/exams/aws-soa-c03/mini/

DevOpsCode Lab is not affiliated with AWS. SOA-C03 and related names are trademarks of their owners. We write every question ourselves; these are not real exam questions.
