LevelUpWith

Explore / AWS Certified CloudOps Engineer - Associate (SOA-C03) Exam Prep / Practice questions

AWS Certified CloudOps Engineer - Associate (SOA-C03) practice questions

10 questions from the bank behind this course's timed mock. Free to read and to answer, no account needed, and each has its answer and an explanation. They belong to AWS Certified CloudOps Engineer - Associate (SOA-C03) Exam Prep.

Answer them one at a time →Read lesson 1 free

  1. Question 1

    A company runs a self-managed MySQL database on an Amazon EC2 instance. A CloudOps engineer is moving it to Amazon RDS for MySQL to reduce routine operational work. Under the AWS shared responsibility model, which task becomes AWS's responsibility after the migration?

    1. Managing which IAM users and roles can change the database
    2. Choosing whether the database storage is encrypted
    3. Patching the MySQL database engine
    4. Configuring security group rules that control access to the database
    Show the answer

    Answer: C. Patching the MySQL database engine

    The split of responsibility depends on the service: on RDS, AWS patches the database engine, while on EC2 the customer patches everything above the hypervisor. Security groups, encryption settings and IAM are security in the cloud, so they stay with the customer whichever service is used.

  2. Question 2

    A company has started using Amazon Bedrock in eu-west-2. Its compliance team needs a record of every model request and response, including the prompt text, that it can review in CloudWatch Logs. Nobody has changed any Bedrock logging settings yet. Which TWO actions should a CloudOps engineer take? (Select TWO.)

    More than one answer is right.

    1. Turn on Bedrock model invocation logging with that log group as the CloudWatch Logs destination and text data delivery enabled.
    2. Create the destination log group in a central logging account in us-east-1 so that logs from every Region end up in one place.
    3. Add a subscription filter to the log group that Bedrock creates by default for model invocations.
    4. Rely on CloudTrail Event history, which already keeps 90 days of management events at no cost.
    5. Create a CloudWatch Logs log group in the same account and Region, plus an IAM role that Bedrock can use to write to it, to act as the logging destination.
    Show the answer

    Answers: A. Turn on Bedrock model invocation logging with that log group as the CloudWatch Logs destination and text data delivery enabled.; E. Create a CloudWatch Logs log group in the same account and Region, plus an IAM role that Bedrock can use to write to it, to act as the logging destination.

    Bedrock model invocation logging is off by default. You first set up a destination in the same account and Region, then turn logging on to capture request and response data. There is no default Bedrock log group to subscribe to, and a destination in another account or Region does not meet the same-account, same-Region requirement.

  3. Question 3

    A CloudOps engineer manages a reporting server with a 200 GiB gp2 data volume. Each morning the reports run fast for about 20 minutes and then slow down sharply. CloudWatch shows BurstBalance falling to near 0 at that point, and VolumeIOPSExceededCheck stays at 0. The team wants steady performance throughout the run without paying for more storage than it needs. What should the engineer do?

    1. Change the volume to st1 to get higher sequential throughput at a lower price per GiB
    2. Use Elastic Volumes to change the volume to gp3 and set its IOPS explicitly to cover the workload's peak, without changing the volume size
    3. Increase the gp2 volume to 1,000 GiB so that its baseline IOPS covers the peak
    4. Move the instance to an instance type with more EBS bandwidth
    Show the answer

    Answer: B. Use Elastic Volumes to change the volume to gp3 and set its IOPS explicitly to cover the workload's peak, without changing the volume size

    When BurstBalance runs out, a gp2 volume falls back to its size-based baseline of 3 IOPS per GiB, which is 600 IOPS here. gp3 gives at least 3,000 IOPS whatever the size, and you can set IOPS separately from size, so the slowdown goes away without buying extra storage. Making the gp2 volume bigger would also raise IOPS, but you'd pay for capacity you don't need.

  4. Question 4

    A CloudOps engineer runs an Amazon DynamoDB table called Orders in provisioned mode, with auto scaling on the table's read and write capacity. Finance wants to keep provisioned mode so costs stay predictable. During a sustained rise in orders, CloudWatch shows write throttling on the table's global secondary index (GSI) ByCustomer, while the base table is not throttled. What should the engineer do?

    1. Switch the table from provisioned mode to on-demand mode
    2. Register the GSI's write capacity as its own scalable target and attach a target-tracking policy to it
    3. Raise the base table's target utilization from 70% to 90%
    4. Raise the base table's maximum write capacity in its auto scaling settings
    Show the answer

    Answer: B. Register the GSI's write capacity as its own scalable target and attach a target-tracking policy to it

    Each GSI has its own capacity, so the index needs its own auto scaling, and changing the base table's settings doesn't help it. Switching to on-demand would also stop the throttling, but it breaks the requirement to keep provisioned mode for predictable costs.

  5. Question 5

    A company is retiring an Amazon RDS for MySQL instance that has automated backups with a 14-day retention period. Compliance requires that the final state of the database can be restored at any time for the next 12 months after the instance is deleted. What should the CloudOps engineer do with the least ongoing effort?

    1. Back up the transaction logs to Amazon S3 every five minutes and delete the instance
    2. Raise the backup retention period to 35 days, then delete the instance
    3. Take a manual DB snapshot before deleting the instance, then delete the instance
    4. Delete the instance with the Retain automated backups option and depend on point-in-time restore
    Show the answer

    Answer: C. Take a manual DB snapshot before deleting the instance, then delete the instance

    Manual snapshots do not expire with the retention period and remain after the instance is deleted, so one snapshot keeps a known-good restore point for the full 12 months. Keeping automated backups is still limited by a retention period of no more than 35 days, which is far short of a year.

  6. Question 6

    A CloudOps engineer manages a production CDK app. Under the change-control policy, a second engineer must review exactly what CloudFormation will change before anything is modified, and the deployment then runs from that reviewed change. Which approach meets this requirement with the least effort?

    1. Run `cdk deploy --require-approval never` and have the reviewer check the stack events afterwards
    2. Run `cdk deploy --hotswap` so that only the changed resources are updated directly
    3. Run `cdk deploy --method=prepare-change-set`, have the reviewer inspect the change set, then execute it
    4. Run `cdk synth` and email the generated template to the reviewer before running `cdk deploy`
    Show the answer

    Answer: C. Run `cdk deploy --method=prepare-change-set`, have the reviewer inspect the change set, then execute it

    `--method=prepare-change-set` creates the CloudFormation change set but does not execute it, so the reviewer approves the exact set of changes before anything happens. Emailing the `cdk synth` output only shows the new template. It doesn't show what will change against the deployed stack, and the deploy that follows isn't tied to what was reviewed.

  7. Question 7

    A CloudOps engineer runs Terraform with an S3 backend that has use_lockfile = true. A faulty script overwrote the production state file with an empty one. The next terraform plan now proposes creating every resource again, even though they all still exist in AWS. What is the fastest way to recover with no change to the infrastructure?

    1. Run terraform apply so Terraform rebuilds the state from the resources it creates
    2. Run terraform force-unlock to release the lock and then run terraform plan again
    3. Restore the previous version of the state object, which works only if S3 Versioning is enabled on the state bucket
    4. Switch the backend to a DynamoDB lock table so the state cannot be overwritten again
    Show the answer

    Answer: C. Restore the previous version of the state object, which works only if S3 Versioning is enabled on the state bucket

    Bucket versioning keeps earlier copies of the state object, so restoring the last good version puts Terraform back in step with real AWS. Locking, whether S3-native or DynamoDB, only stops two runs writing at the same time. It can't undo a bad write that already happened.

  8. Question 8

    A CloudOps engineer in account 111111111111 runs an application under the IAM role app-reader. The role must read objects from a bucket in account 222222222222. The bucket owner added a bucket policy that allows s3:GetObject for the role's ARN. The role's identity policy only allows DynamoDB actions. No SCPs, boundaries or session policies apply. GetObject calls still return AccessDenied. What is the correct fix?

    1. Pass a session policy that allows s3:GetObject when the application assumes app-reader
    2. Add an identity-based policy to app-reader in account 111111111111 that allows s3:GetObject on the bucket's objects
    3. Change the bucket policy in account 222222222222 to name the account root principal instead of the role ARN
    4. Attach a permissions boundary to app-reader that allows s3:GetObject on the bucket
    Show the answer

    Answer: B. Add an identity-based policy to app-reader in account 111111111111 that allows s3:GetObject on the bucket's objects

    In cross-account access both sides must allow the request, so the bucket policy alone is not enough and the role's own account must also grant s3:GetObject. A permissions boundary or a session policy only limits permissions and never grants them, so neither one can supply the missing Allow.

  9. Question 9

    A CloudOps engineer turns on scheduled rotation for a Secrets Manager secret that holds credentials for an Amazon RDS for PostgreSQL database. The rotation Lambda function runs in private subnets that have no internet access, and its security group can reach the database port. Every rotation attempt times out, and the function logs show calls to Secrets Manager that hang until the timeout. The database password never changes. What should the engineer do so that rotation succeeds?

    1. Re-encrypt the secret with a customer managed KMS key instead of the aws/secretsmanager key
    2. Increase the Lambda function timeout and set a longer rotation window Duration in the rotation rules
    3. Give the function a network path to Secrets Manager by adding a Secrets Manager interface VPC endpoint or a route through a NAT gateway
    4. Switch the secret from the single-user rotation strategy to the alternating-users rotation strategy
    Show the answer

    Answer: C. Give the function a network path to Secrets Manager by adding a Secrets Manager interface VPC endpoint or a route through a NAT gateway

    A rotation function in a VPC must reach both the database and the Secrets Manager endpoint, and here the hanging Secrets Manager calls show that the second path is missing. A longer timeout or rotation window only makes the function wait longer for a connection it can never open.

  10. Question 10

    A CloudOps engineer manages a shared-services VPC A that has a NAT gateway and an S3 gateway endpoint. Application VPC B is peered with VPC A, the connection is active, and both sides have routes for each other's CIDRs. To avoid running extra infrastructure, the application team adds a 0.0.0.0/0 route in VPC B that targets the peering connection, expecting VPC B's private instances to reach the internet and Amazon S3 through VPC A. The traffic fails. What should the engineer do to meet the requirement?

    1. Attach a virtual private gateway to VPC A and turn on route propagation in VPC B's route tables
    2. Add a route in VPC A's private route table that sends traffic arriving from the peering connection to the NAT gateway
    3. Create a NAT gateway and an S3 gateway endpoint in VPC B and route VPC B's private subnets to them
    4. Enable transitive routing on the VPC peering connection so that VPC B can use VPC A's gateways
    Show the answer

    Answer: C. Create a NAT gateway and an S3 gateway endpoint in VPC B and route VPC B's private subnets to them

    VPC peering doesn't support edge-to-edge routing, so VPC B can never use VPC A's NAT gateway or gateway endpoint, and VPC B needs its own. Editing VPC A's route tables doesn't help, because the peering connection won't forward VPC B's traffic through VPC A's gateways, and 'transitive peering' isn't a setting you can turn on.

The full timed mock

65 questions in 130 minutes, with a pass mark of 72%. Each paper is drawn from a bank of questions written for the mock alone, and these 10 are left out of it. Full access opens it, with every level of the course.

See the course →See the plans

LevelUpWith is independent of the organisation that sets this exam. These questions were written for this course; they are not taken from the exam.