AWS CloudOps Engineer Practice Exam (Free, 2026)

Illustration of a stack of pastel cards with a tick mark beside a gauge

This free AWS CloudOps Engineer practice exam has 18 questions written to the current SOA-C03 blueprint, the exam that replaced AWS Certified SysOps Administrator – Associate, covering all five domains, and every answer explains why each option is right or wrong. Choose your answers, then open “Show answer” to check. The real SOA-C03 has 65 questions in 130 minutes, and you need a scaled score of 720 to pass (facts verified November 2026 against the AWS exam page).

See the full blueprint in our SOA-C03 exam guide.

Domain 1: Monitoring, Logging, Analysis, Remediation, and Performance Optimization (questions 1–4)

This domain is 22% of the scored exam.

1. A web application runs stateless web servers on three Amazon EC2 instances behind a load balancer. Users complain about performance. Amazon CloudWatch shows one instance at about 90% CPU while the other two are at about 20%. How can a SysOps administrator prevent this from happening again?

  1. Manually provision a fourth EC2 instance and add it to the load balancer.
  2. Review the performance of the database server that supports the web application, then choose a more appropriate instance size and apply the change.
  3. Disable sticky sessions on the load balancer and check CloudWatch to confirm that the load is now even.
  4. Review the number of sessions from the load balancer to each web server and restart any server with significantly more sessions.
Show answer

Answer: C. One busy instance and two quiet ones behind a load balancer means traffic isn’t being spread evenly. With sticky sessions turned on, each user keeps going to the same instance, so a few heavy users can pin one server. The servers are stateless, so they don’t need stickiness at all. Turning it off lets the load balancer distribute requests evenly, and CloudWatch confirms the fix.

Why not the others: A fourth instance adds capacity but doesn’t fix the uneven routing, so the same pattern would return. The database isn’t the instance showing high CPU. Restarting the busiest server moves the problem around for a while without removing its cause.

Domain: Monitoring, Logging, Analysis, Remediation, and Performance Optimization

2. An operations team runs a web application on EC2 instances registered with an Application Load Balancer. They want an alert within about a minute whenever the load balancer fails to open connections to its targets, and they don’t want to write any code. What should they set up?

  1. A CloudWatch alarm on HTTPCode_Target_5XX_Count above zero that notifies an Amazon SNS topic
  2. A CloudWatch alarm on TargetConnectionErrorCount that notifies an Amazon SNS topic
  3. ALB access logs delivered to Amazon S3, parsed by an AWS Lambda function that emails the team
  4. A CloudWatch alarm on HealthyHostCount equal to zero that notifies an Amazon SNS topic
Show answer

Answer: B. TargetConnectionErrorCount is the ALB metric that counts connections the load balancer failed to open to its targets, which is exactly the problem described. A CloudWatch alarm on it, sending to an SNS topic, is near-real time and needs no code.

Why not the others: HTTPCode_Target_5XX_Count counts errors returned by the application after a connection succeeded, which is a different problem. HealthyHostCount equal to zero only alerts when every target is down, far too late. Processing access logs with Lambda works, but it’s custom code and slower, which breaks the “no code” requirement.

Domain: Monitoring, Logging, Analysis, Remediation, and Performance Optimization

3. You notice that the CPU credit balance of some burstable EC2 instances keeps dropping to 0. What can help? (Select TWO.)

  1. Switch the instances to unlimited mode
  2. Change to a higher instance type
  3. Use t1.micro instances
  4. Use EBS optimized instances
Show answer

Answer: A and B. Burstable instances (the T family) earn CPU credits while they’re below their baseline and spend them to burst above it. When the balance hits zero in standard mode, the instance is held to its baseline, and the application slows. Unlimited mode lets it keep bursting, with extra charges if average usage stays above baseline. A larger instance type, or a non-burstable family, gives the workload more CPU permanently.

Why not the others: t1.micro is an old, small previous-generation type, so it would make the shortage worse. EBS optimization improves storage throughput; it does nothing for CPU credits.

Domain: Monitoring, Logging, Analysis, Remediation, and Performance Optimization

4. A memory-hungry application runs on a fleet of Amazon Linux EC2 instances. The team needs one memory utilization data point per instance every minute in Amazon CloudWatch. Which TWO steps achieve this? (Select TWO.)

  1. Turn on EC2 detailed monitoring for the instances
  2. Send the memory figures to Amazon EventBridge (CloudWatch Events)
  3. Install and configure the CloudWatch agent to collect memory metrics
  4. Write the memory figures to Amazon CloudWatch Logs
  5. Set the agent’s metrics_collection_interval to 60 seconds
Show answer

Answer: C and E. EC2 sees an instance from the outside, so its built-in metrics don’t include memory, which only the operating system knows. The CloudWatch agent runs inside the instance and publishes memory metrics. Its metrics_collection_interval setting controls how often, and 60 seconds gives the one-minute resolution required.

Why not the others: Detailed monitoring makes the built-in metrics arrive every minute instead of every five, but it still doesn’t add memory. CloudWatch Events (now Amazon EventBridge) routes events rather than storing metrics. CloudWatch Logs stores log data; publishing a metric there isn’t how memory monitoring works.

Domain: Monitoring, Logging, Analysis, Remediation, and Performance Optimization

More on monitoring: CloudWatch vs CloudTrail.

Domain 2: Reliability and Business Continuity (questions 5–8)

This domain is 22% of the scored exam.

5. Someone deleted an important file from the data volume of an EC2 instance. An EBS snapshot of that volume was taken last night. How can an administrator get that single file back?

  1. Create a new EBS volume from the snapshot in the instance’s Availability Zone, attach and mount it, then copy the file across
  2. Use the file-restore option for the snapshot in the EC2 console
  3. Open the snapshot in the Amazon S3 console and download the file
  4. Attach the snapshot directly to the instance as a second disk and copy the file
Show answer

Answer: A. An EBS snapshot is a block-level backup stored by AWS, not a disk you can open. To get one file back, you create a new volume from the snapshot in the instance’s Availability Zone, attach it to the instance, mount it, and copy the file to its original location.

Why not the others: The EC2 console has no file-level restore for snapshots. Snapshots are stored by AWS behind the scenes and don’t appear as browsable objects in the S3 console. They can’t be attached to an instance directly either; only volumes can.

Domain: Reliability and Business Continuity

6. A production Amazon RDS for MySQL instance runs in a single Availability Zone, and several applications depend on it. The operations lead wants a database failure to cause as little downtime as possible, and any change must not affect customers during the working day. What should be done?

  1. Convert the instance to a Multi-AZ deployment, applying the change outside working hours
  2. Ask AWS Support to pre-warm the database
  3. Add a read replica outside working hours
  4. Launch a new Multi-AZ instance, migrate the data, then delete the old instance
Show answer

Answer: A. A Multi-AZ deployment keeps a synchronous standby in another Availability Zone and fails over to it automatically, so applications reconnect to the same endpoint within minutes. You can convert an existing instance in place, and scheduling the change outside business hours keeps any performance impact away from customers.

Why not the others: “Pre-warming” isn’t an RDS feature. A read replica uses asynchronous replication and doesn’t fail over automatically, so it adds read capacity rather than availability. Building a new instance and migrating the data is more disruptive and riskier than modifying the existing one.

Domain: Reliability and Business Continuity

7. An application runs on EC2 web servers behind a load balancer, with an Amazon RDS relational database and a NAT gateway for outbound internet traffic. Which changes increase the scalability and availability of the application? (Select TWO.)

  1. Launch the EC2 instances as part of an Auto Scaling group
  2. Enable Multi-AZ for the database
  3. Use a NAT instance instead of a NAT gateway
  4. Use Route 53 instead of the load balancer
Show answer

Answer: A and B. An Auto Scaling group adds web servers as demand grows and replaces unhealthy ones, which helps both scalability and availability. Multi-AZ gives the database a standby in another Availability Zone that takes over automatically if the primary fails.

Why not the others: A NAT instance is a single EC2 instance you manage yourself, so it’s less available and less scalable than a managed NAT gateway. Route 53 can route between endpoints, but it can’t replace a load balancer for spreading requests across a fleet and checking target health per request.

Domain: Reliability and Business Continuity

8. A media company stores video uploaded by reporters in Amazon S3, using a new bucket each month such as example-bucket-july. The buckets have versioning enabled. At the start of each month, the previous month’s bucket and all its data must be deleted. Which approach will work?

  1. Delete bucket example-bucket-july programmatically using AWS SDK.
  2. Delete bucket using command “$ aws s3 rb s3://example-bucket-july –force”
  3. Delete bucket using command “$ aws s3 rb s3:// example-bucket-july”
  4. Delete bucket using command “$ aws s3 rm s3:// example-bucket-july –recursive”
Show answer

Answer: A. S3 only deletes a bucket once it’s completely empty. In a versioned bucket, that means every object version and every delete marker must be gone, not just the current objects. A script using the AWS SDK can list and delete all versions, then delete the bucket. The S3 console’s Empty bucket action or a lifecycle rule can also clear the versions.

Why not the others: aws s3 rb --force and aws s3 rm --recursive remove current objects only. Older versions stay behind, so the bucket isn’t empty and deleting it fails. Plain aws s3 rb without emptying the bucket fails straight away.

Domain: Reliability and Business Continuity

Domain 3: Deployment, Provisioning, and Automation (questions 9–12)

This domain is 22% of the scored exam.

9. A team’s large AWS CloudFormation template is being broken up into one template per service. Every service uses the same Amazon S3 bucket. How should the bucket be defined so that all the new templates can refer to it?

  1. Deploy the bucket with CloudFormation StackSets
  2. Declare the bucket as a resource in every service template
  3. Define the bucket in its own stack and export its name for the other stacks to import
  4. Add the bucket to the Mappings section of each template
Show answer

Answer: C. Put the shared bucket in its own stack and export its name from the Outputs section. Each service template then reads it with Fn::ImportValue. CloudFormation also stops you from deleting the exporting stack while other stacks still import the value.

Why not the others: StackSets deploy the same stack to multiple accounts or Regions; they don’t share one resource between stacks. Declaring the bucket in every template would try to create it several times, and bucket names are globally unique. Mappings hold fixed lookup values, not live resources.

Domain: Deployment, Provisioning, and Automation

10. You are creating AWS CloudFormation stacks for the Operations department. If stack creation fails, you want the resources left in place so you can find and fix the errors manually. How can you achieve this?

  1. Set the DeleteAPI for the stack to None
  2. Set the OnFailure to DO_NOTHING when creating the stack
  3. Enable Termination protection on the stack
  4. Set the ForceAPI for the stack as DO_NOT_DELETE
Show answer

Answer: B. By default, a failed stack creation rolls back and deletes what it created. Setting OnFailure to DO_NOTHING (in the console, choosing to preserve successfully provisioned resources) leaves those resources in place, so you can inspect them and work out what went wrong.

Why not the others: Termination protection stops someone from deleting a stack; it doesn’t change what happens when creation fails. “DeleteAPI” and “ForceAPI” aren’t CloudFormation settings at all, which makes them easy to rule out.

Domain: Deployment, Provisioning, and Automation

11. An Auto Scaling group spread across three Availability Zones was created by an AWS CloudFormation stack. A new, patched AMI is ready. Using CloudFormation, how can the instances be moved to the new AMI while a minimum number always stays in service?

  1. Change the AMI ID in the template, update the stack, then reboot all the instances
  2. Build a second stack with the new AMI and switch traffic over with Amazon Route 53
  3. Add an UpdatePolicy with AutoScalingRollingUpdate (including MinInstancesInService), then update the stack with the new AMI ID
  4. Run aws cloudformation update-stack with the –rollback-configuration option
Show answer

Answer: C. An UpdatePolicy with AutoScalingRollingUpdate tells CloudFormation to replace instances in batches when the launch configuration or template changes. MinInstancesInService keeps enough instances running throughout, which meets the “minimum number always stays in service” requirement.

Why not the others: Rebooting doesn’t change an instance’s AMI; only replacement does. A second stack with a DNS cutover can work, but it’s a blue/green rebuild, which is more effort than the built-in rolling update. A rollback configuration sets alarms that trigger a rollback; it doesn’t control how instances are replaced.

Domain: Deployment, Provisioning, and Automation

12. A development team runs an application in Docker containers on premises and wants to move it to AWS without managing the underlying infrastructure. Which service should you use to provision the environment?

  1. AWS Elastic Beanstalk
  2. Amazon EC2
  3. AWS CloudFormation
  4. AWS Systems Manager
Show answer

Answer: A. Elastic Beanstalk’s Docker platform takes the team’s containers and handles capacity provisioning, load balancing, scaling and health monitoring for them. The team deploys the application, and AWS runs the environment. Of the options listed, it’s the only one that removes infrastructure management.

Why not the others: Amazon EC2 means launching and managing the servers yourself. CloudFormation describes infrastructure as code, but you still own and operate what it creates. Systems Manager helps operate existing instances; it doesn’t host applications. Outside these options, Amazon ECS on AWS Fargate is another common way to run containers without servers.

More on containers: ECS vs EKS.

Domain: Deployment, Provisioning, and Automation

Domain 4: Security and Compliance (questions 13–15)

This domain is 16% of the scored exam.

13. A company must scan its EC2 instances against the Center for Internet Security (CIS) Benchmarks to find security weaknesses. Which service can do this?

  1. Use the AWS Trusted Advisor
  2. Use the AWS Inspector service
  3. Use the AWS Config service
  4. Use the AWS GuardDuty service
Show answer

Answer: B. Amazon Inspector scans EC2 instances for software vulnerabilities and can also run CIS Benchmark scans of the operating system, reporting which checks pass and fail.

Why not the others: Trusted Advisor gives account-level best-practice recommendations; it doesn’t scan inside instances. AWS Config evaluates resource settings, such as whether a security group allows open SSH, not the OS hardening that CIS Benchmarks cover. GuardDuty detects threats from activity such as CloudTrail events, VPC Flow Logs and DNS logs; it doesn’t audit configurations.

Domain: Security and Compliance

14. Several security groups in a VPC allow SSH (port 22) and RDP (port 3389) from 0.0.0.0/0. The security team wants any such rule to be found and removed automatically, now and in the future. What should the administrator configure?

  1. AWS Trusted Advisor, set to remove the open rules
  2. AWS Systems Manager configuration compliance, set to remove the open rules
  3. An AWS Config rule that detects the open ports, with automatic remediation that runs an AWS Systems Manager Automation document
  4. An IAM policy that denies ports 22 and 3389 in every security group
Show answer

Answer: C. An AWS Config rule, such as the managed rule that checks for unrestricted SSH, flags any security group with the ports open to the world. A remediation action then runs a Systems Manager Automation document that removes the offending rule automatically. Detection and fix, with nobody logging in, is exactly the pattern SOA-C03 looks for.

Why not the others: Trusted Advisor can report open ports, but it doesn’t fix them. Systems Manager configuration compliance tracks patch and association compliance on managed instances, not security group rules. IAM policies control who can call AWS APIs; they can’t close network ports on existing security groups.

Domain: Security and Compliance

15. Engineers call AWS from the CLI with IAM user access keys. A new rule says any access key older than 90 days must be switched off automatically. Which approach meets the rule?

  1. AWS Trusted Advisor, set to disable old keys
  2. A CloudWatch alarm on access key age that invokes an AWS Lambda function
  3. The AWS Config managed rule for access key rotation, with automatic remediation that runs an AWS Systems Manager Automation document to deactivate noncompliant keys
  4. An account password policy with a 90-day expiry
Show answer

Answer: C. The AWS Config managed rule access-keys-rotated marks access keys older than a set number of days, such as 90, as noncompliant. Attaching a remediation action that runs a Systems Manager Automation document disables those keys automatically. It’s the same detect-and-fix pattern as the previous question.

Why not the others: Trusted Advisor can’t disable keys. CloudWatch has no built-in metric for access key age, so there’s nothing for an alarm to watch. A password policy controls console passwords, not access keys.

Domain: Security and Compliance

Domain 5: Networking and Content Delivery (questions 16–18)

This domain is 18% of the scored exam.

16. Your company will run a web application on EC2 instances and needs a service to spread traffic across them that can scale to millions of requests per second. What should you use?

  1. An ECS cluster
  2. A Network Load Balancer
  3. A Classic Load Balancer
  4. An SQS queue
Show answer

Answer: B. A Network Load Balancer works at layer 4 (TCP and UDP) and is built to handle millions of requests per second with very low latency. It can also have a static IP address per Availability Zone. “Millions of requests per second” is the classic cue for an NLB.

Why not the others: An ECS cluster runs containers; it doesn’t distribute traffic to EC2 instances. The Classic Load Balancer is a previous-generation service that AWS recommends replacing. An SQS queue holds messages for asynchronous processing, not live web traffic.

Domain: Networking and Content Delivery

17. A startup has three VPCs for testing. VPC A peers with VPC B and with VPC C, and VPC B and VPC C use the same CIDR block. VPC B has one instance and VPC C has more than 100. An instance in VPC A must reach the instance in VPC B and all instances in VPC C. Which routing meets this requirement?

  1. Add a more specific routing entry for Instance in VPC A in routing table of VPC B pointing to target as VPC Peering connection between VPC A & VPC B.
  2. Add a more specific routing entry for Instance in VPC B in routing table of VPC A pointing to target as VPC Peering connection between VPC A & VPC B.
  3. Create a new VPC with separate subnets between VPC A & VPC B.
  4. Add a less specific routing entry for Instance in VPC A in routing table of VPC B pointing to target as VPC Peering connection between VPC A & VPC B.
Show answer

Answer: B. VPC route tables always use the most specific matching route (longest prefix match). In VPC A’s route table, a /32 route for the single VPC B instance points to the A–B peering connection, while the broader route for the shared CIDR points to the A–C peering connection. Traffic to that one address goes to VPC B, and everything else in the range goes to VPC C. AWS documents this overlapping-CIDR peering pattern.

Why not the others: The problem is about where traffic leaving VPC A goes, so the fix belongs in VPC A’s route table, not VPC B’s. A “less specific” route would lose to the broader routes already there. A new VPC doesn’t solve the overlap and adds complexity.

Domain: Networking and Content Delivery

18. Your company serves files from an S3 bucket through an Amazon CloudFront distribution. Some requests to CloudFront return 4xx errors. What could cause these errors? (Select TWO.)

  1. The user does not have access to the underlying bucket
  2. The CloudFront service is unavailable
  3. The object which the user is requesting is not present
  4. There is an internal server error
Show answer

Answer: A and C. 4xx status codes mean a problem with the request itself. A 403 Forbidden appears when the request isn’t allowed to read the object, for example when the bucket policy doesn’t grant access to CloudFront’s origin access control. A 404 Not Found appears when the requested object doesn’t exist. See CloudFront’s error response troubleshooting.

Why not the others: An unavailable service or an internal server error are server-side problems, which produce 5xx codes, not 4xx.

Domain: Networking and Content Delivery

Your score by domain

Count your correct answers in each domain. Multiple-response questions count only if you picked every correct option and nothing else.

DomainQuestionsYour score
Monitoring, Logging, Analysis, Remediation, and Performance Optimization1–4__ / 4
Reliability and Business Continuity5–8__ / 4
Deployment, Provisioning, and Automation9–12__ / 4
Security and Compliance13–15__ / 3
Networking and Content Delivery16–18__ / 3

A short set shows where you’re weak; it can’t predict your real exam result. SOA-C03 rewards automated fixes over manual ones, so for every miss, ask whether the right answer removed a person from the loop.

Ready for a full, timed exam?

The real SOA-C03 gives you 130 minutes for 65 questions. Our SOA-C03 practice exams follow the current CloudOps blueprint, with every option explained.

Free mock exam

Take SOA-C03 Mock Exam 1, free

A full-length, timed mock in the real exam format, with every answer explained so you know exactly where you stand.

65

questions

130

minutes

720

to pass

Free account required.

If you score 80% or higher on every mock exam in the course, sit the real exam within 30 days of buying and don’t pass, we’ll refund the course in full. Terms apply.

Weak on a topic? Read these next

AWS CloudOps Engineer practice exam FAQ

Is this a SysOps Administrator practice exam?

It’s written for SOA-C03, the CloudOps Engineer – Associate exam that replaced SysOps Administrator (SOA-C02) on 30 September 2025. Most SysOps topics still apply, with more weight on reliability and automation.

Are these real exam questions?

No. AWS doesn’t publish live exam questions, and sharing them breaks the AWS Certification agreement. These are written to the official SOA-C03 exam guide.

What is the SOA-C03 passing score?

720 on a scaled score from 100 to 1000, with compensatory scoring across domains.

Does SOA-C03 have labs?

No. It has multiple-choice and multiple-response questions only.

Is this practice exam free?

Yes, with no sign-up. Mock Exam 1 in our course is also free; it needs a free account.


AWSBoy is not affiliated with Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates.

Sources (checked November 2026): AWS Certified CloudOps Engineer – Associate exam page and SOA-C03 exam guide · AWS service documentation linked in each answer.

Shopping Cart
Scroll to Top