Answer before reviewing
Choose the one answer that meets every requirement in the scenario. Then read why each other option falls short.
No sign-up needed
20 questions, each with a clear explanation.
Question review
Try the set above first, then use this list to review. Each answer and explanation stays closed until you open it.
Correct answer: C · The company, because in IaaS the customer runs and controls the guest OS
NIST SP 800-145 says an IaaS consumer can deploy and run arbitrary software, including operating systems, and has control over operating systems, storage, and deployed applications, while the provider manages the underlying cloud infrastructure. The guest operating system is therefore the company's to patch. Owning the physical hosts makes the provider responsible for the infrastructure beneath the VMs, not for the software inside them. Managed operating system updates describe a platform service, where the provider runs the operating system, not IaaS. The OS vendor releases the fix, but someone with control of each VM still has to apply it.
Look for this next time: In IaaS the customer controls the guest operating system, so patching it is the customer's job.
Correct answer: A · The 15-minute limit is the RPO, and the four-hour limit is the RTO
NIST's glossary, citing SP 800-34 Rev. 1, defines the recovery point objective (RPO) as the point in time to which data must be recovered after an outage, so a cap on how many minutes of orders can be lost is the RPO. It defines the recovery time objective (RTO) as how long system components can stay in the recovery phase before the mission or business is affected, which is the four-hour target. Swapping the labels reverses both definitions. Calling both targets RTOs ignores that one of them limits lost data, not downtime, and calling both RPOs ignores that four hours limits the time to restore service.
Look for this next time: RPO caps how much data you can lose; RTO caps how long recovery can take.
Correct answer: D · An application load balancer, with rules that match the URL path
AWS documents that an Application Load Balancer functions at the application layer, the seventh layer of the OSI model, and that its listener rules can use path conditions to route requests based on the URL, which AWS calls path-based routing. That is how /api/ requests reach one group while the rest go to the other. AWS describes a Network Load Balancer as working at the fourth layer, where it handles connections and ports rather than HTTP paths. DNS round robin hands out addresses before a request exists, so it never sees a path. A VPN gateway protects traffic in transit but does not route web requests by path.
Look for this next time: Routing by URL path needs a layer 7 load balancer that reads the HTTP request.
Correct answer: B · A persistent volume, which has a lifecycle independent of any one pod
Kubernetes documentation describes a PersistentVolume as cluster storage with a lifecycle independent of any individual pod that uses it, so the database files remain when the pod is replaced. It describes ephemeral volumes, including emptyDir, as following the pod's lifetime and being created and deleted along with the pod, which is the loss the team already sees. Files packed into a container image only give each new container the same starting content, and data written at runtime is not saved back into the image. Extra replicas do not preserve data by themselves; each pod still needs storage that outlives it.
Look for this next time: Data that must outlive a pod belongs on a persistent volume, not in ephemeral storage.
Correct answer: A · Spot instances, which sell spare capacity at a discount but can be reclaimed
AWS describes Spot Instances as spare EC2 capacity available for less than the On-Demand price and calls them a cost-effective choice if you can be flexible about when applications run and if they can be interrupted, naming batch jobs as an example. AWS can reclaim a Spot Instance with a two-minute warning, which this job tolerates because it saves its progress. A dedicated host pays for isolation, not a discount. On-demand pricing is the standard rate that Spot discounts are measured against. Reserved capacity rewards steady, round-the-clock use, and this job runs for only part of each night.
Look for this next time: Interruptible, restartable batch work is the classic fit for discounted spot capacity.
Correct answer: C · Community cloud, used only by organizations that have shared concerns
NIST SP 800-145 defines a community cloud as infrastructure provisioned for exclusive use by a specific community of consumers from organizations that have shared concerns, such as security requirements, policy, and compliance considerations, and says one or more of those organizations may own and manage it. Four hospitals with the same privacy rules and a joint committee fit that definition. NIST's private cloud serves a single organization, and its public cloud is provisioned for open use by the general public. A hybrid cloud is a composition of two or more distinct cloud infrastructures, and the hospitals are sharing one.
Look for this next time: Several organizations with shared requirements using one exclusive environment is a community cloud.
Correct answer: B · Canary, which sends a small share of users to it before a wider rollout
Martin Fowler's site describes a canary release as reducing the risk of a new version by rolling the change out slowly to a small subset of users before the whole infrastructure, then routing more users as confidence grows. The 5% share and the check after an hour match that pattern. The same site describes blue-green deployment as switching the router so all incoming requests go to the new environment at once. An in-place update changes the existing servers with no small test group. A rolling deployment is defined by batches of servers, while this release is staged by a measured share of users and a decision point.
Look for this next time: A small, measured share of users first, then a wider release if metrics hold, is a canary.
Correct answer: D · Replatform, because the move adds a targeted optimization along the way
AWS Prescriptive Guidance describes replatforming, also called lift and reshape, as moving an application to the cloud while introducing some level of optimization, and its example is replatforming a SQL Server database to a managed relational database service. Rehosting, or lift and shift, moves applications without making any changes, but this move changes the database platform. Refactoring modifies the architecture to take full advantage of cloud-native features, which the stem rules out. Retaining keeps an application in the source environment, and here the database moves to the cloud.
Look for this next time: Moving with one targeted change, such as a managed database, is replatforming.
Correct answer: C · Drift, a live configuration that differs from its definition in code
AWS CloudFormation documentation explains that drift detection finds whether a stack's actual configuration differs, or has drifted, from its expected configuration, and that drift comes from changes made outside the IaC tool, including ones made to respond to time-sensitive operational events. A port opened in the console during an incident is that case; the fix is to update the template or reverse the manual change so the two agree. Versioning tracks changes stored in the repository, and this change never reached it. Repeatability is about the code producing the same environment each time. Testing checks code before it deploys and would not catch a later console change.
Look for this next time: A manual change that leaves live resources different from the IaC definition is drift.
Correct answer: A · Distributed tracing, which follows one request through each service it calls
OpenTelemetry's documentation says traces give the big picture of what happens when a request is made to an application and are essential to understanding the full path a request takes, whether the application is a monolith or a mesh of services. Timing one checkout request across six services is that job. Longer log retention keeps more history but does not link one request's steps across services. More CPU alerts would repeat what the metrics already show, and the metrics look normal. A larger instance size adds capacity without showing where the time goes.
Look for this next time: To see where one request spends its time across services, use distributed tracing.
Correct answer: B · Scheduled scaling, so new servers are running before each Friday sale
AWS's EC2 Auto Scaling documentation describes scheduled scaling as automatic scaling based on predictable load changes, using scheduled actions that increase or decrease capacity at specific times. A sale that starts at the same hour every week is predictable, so capacity can be in place before shoppers arrive. Waiting for alerts or a CPU trigger is the delay that hurt the first wave, and a higher threshold would act even later. Vertical scaling changes the size of servers rather than their number, and it would still need a schedule to be ready on time.
Look for this next time: Predictable peaks call for scheduled scaling that adds capacity before the load arrives.
Correct answer: D · Differential, which captures every change made since the last full backup
NIST SP 800-34 Rev. 1 describes a differential backup as one that stores files created or modified since the last full backup, so a restore needs only the full backup media and the last differential. That matches Sunday's full backup plus Wednesday's backup. NIST describes an incremental backup as capturing changes since the last backup of any type, so restoring from incrementals would also require Monday's and Tuesday's media. With a full backup every night, Wednesday's copy alone would do. Replication copies changes to another location as they happen and is not restored as a full-plus-one pair.
Look for this next time: Full plus the latest differential restores everything; incrementals need every backup since the full.
Correct answer: A · PCI DSS, because the log bucket now stores payment card account data
The PCI Security Standards Council says PCI DSS applies to entities that store, process, or transmit cardholder data or sensitive authentication data, and that it sets technical and operational requirements to protect payment account data. A bucket holding full card numbers stores that data, so it falls under the same requirements as the checkout service. A SOC 2 report, an ISO 27001 certificate, and a CSA STAR listing can each help a company show its security practices, but storing card numbers is not what makes any of them apply. The obligation that follows the card data comes from PCI DSS.
Look for this next time: Any system that stores, processes, or transmits card data falls under PCI DSS.
Correct answer: C · Use a role with short-lived credentials limited to that one bucket
AWS's IAM best practices tell teams to require workloads to use temporary credentials with IAM roles and to apply least-privilege permissions, granting only the permissions required to perform a task. A role limited to one bucket, with short-lived credentials, keeps the job running while shrinking what a leaked credential could reach. Rotating the administrator keys every 90 days still leaves full administrative access exposed between rotations. A dedicated user with administrator permissions keeps the same excess rights. A hidden file is still a long-term secret on disk with the same reach.
Look for this next time: Give workloads short-lived credentials scoped to only the resources they need.
Correct answer: D · Rotate the password, then load it at runtime from a secrets manager
The OWASP Secrets Management Cheat Sheet warns that many organizations hardcode secrets in source code, and its incident guidance calls for revoking an exposed secret, rotating in a new one, and removing it from code, noting that the old value can remain in commit history. A password stored in a repository for three months should be treated as exposed, so it needs a new value, and the script should fetch it from a secrets manager at runtime. Deleting the line in a new commit leaves the old value in history. Making the repository private limits future readers only. Encrypting the file protects only later versions.
Look for this next time: Treat a secret found in source control as exposed, rotate it, and serve the new value from a secrets manager.
Correct answer: B · Cryptojacking, where attackers use the VMs' CPUs to mine cryptocurrency
CISA describes cryptojacking as malicious actors hijacking the processing power of victim systems to run cryptomining software, and lists degraded performance from CPU resources monopolized by mining and increased power consumption among its effects. Constant high CPU on VMs that normally idle, steady outbound connections, and a jump in the compute bill fit that pattern. A DDoS attack floods targets with incoming traffic, while this pattern is outbound and CPU-bound. Phishing may be how an attacker first got in, but it does not describe what the VMs are doing now. Ransomware would show encrypted data and a demand, which the stem does not report.
Look for this next time: Unexplained, sustained CPU use with outbound connections and a rising bill points to cryptojacking.
Correct answer: C · Continuous deployment, which ships every passing change to production
Martin Fowler's description of continuous delivery separates it from continuous deployment: in continuous deployment every change goes through the pipeline and is automatically put into production, while continuous delivery means a team is able to deploy frequently but may choose not to. A pipeline that releases each passing change with no approval step is continuous deployment. Continuous integration covers merging, building, and testing, and stops short of release. Blue-green deployment is a technique for switching traffic during a release, and it says nothing about whether a person approves each change.
Look for this next time: Every passing change released automatically, with no approval gate, is continuous deployment.
Correct answer: D · Ansible, which applies YAML playbooks over SSH without agents
Ansible's documentation describes an agent-less architecture that avoids installing additional software across the infrastructure, playbooks in YAML that declare the desired state of a system, and access to remote machines over SSH with existing operating system credentials. Each requirement in the stem matches. HashiCorp describes Terraform as an infrastructure as code tool for building and changing cloud and on-premises resources, which suits creating infrastructure more than reconfiguring 300 existing servers over SSH. Jenkins is an automation server for building, testing, and delivering software, and Grafana queries and visualizes metrics, logs, and traces. Neither is the agentless YAML configuration tool the stem describes.
Look for this next time: Agentless configuration in YAML playbooks, pushed over SSH, describes Ansible.
Correct answer: A · API rate limiting; slow the loop and retry after the interval the API gives
RFC 6585 defines HTTP 429 Too Many Requests as the status code for a client that has sent too many requests in a given amount of time, which it calls rate limiting, and says the response may include a Retry-After header saying how long to wait. Slowing the loop and retrying after that interval fixes the cause. An expired credential is an authentication problem, and a 429 says the calls came too fast, not who made them. A regional outage would not single out one fast-calling script with a status code about request volume. A malformed request would have failed from the first call, not after dozens of successes.
Look for this next time: HTTP 429 means rate limiting: back off and retry instead of calling the API faster.
Correct answer: B · The CIDR blocks overlap, since all of 10.20.128.0/17 is in 10.20.0.0/16
AWS documentation says you cannot create a VPC peering connection between VPCs that have matching or overlapping IPv4 or IPv6 CIDR blocks. 10.20.0.0/16 covers every address from 10.20.0.0 to 10.20.255.255, so 10.20.128.0/17, which runs from 10.20.128.0 to 10.20.255.255, sits entirely inside it. One VPC needs a non-overlapping range before peering can work. AWS also documents inter-Region peering and says the accepter VPC can belong to you or to another AWS account, so peering is not limited to one region or one account. Subnet size limits the addresses available to instances, not routes. Security groups filter traffic once a connection exists; they do not decide whether peering can be created.
Look for this next time: Networks you want to peer need address ranges that do not overlap.
Use the sample well
A short set can show you what deserves another look. A strong run shows you can read a cloud requirement, an error, or a monitoring clue and pick the step a cloud engineer would take next. Twenty questions are a small sample, so the result does not predict a 750 on the 100–900 scale.
Choose the one answer that meets every requirement in the scenario. Then read why each other option falls short.
Confirm why the best option fits, even when you were right.
Use recurring misses—not one score—to choose your next focus.
Full CompTIA Cloud+ prep
Want more than 20 questions?
At launch: 800 exam-style CompTIA Cloud+ questions and 500 flashcards, practiced one domain at a time.
Explore Full CompTIA Cloud+ PrepFree CompTIA Cloud+ resources
Open the exam overview, build a study plan, or browse the CompTIA Cloud+ guides.
Review the exam format, official sources, and preparation options.
Open exam hub →Practical planning aidTurn the time you have into a clear study rhythm.
Build my outline →Source-backed answersGo straight to the question already on your mind.
Browse guides →