Practice exams / CGE-AZ CGE-AZ practice exam 20 questions across the 7 exam domains, weighted like the real exam. Untimed. Submit to see your score by domain and an explanation for every answer. The real exam is 50 questions in 75 minutes and you pass at 70%. These questions are for practice and do not appear on the graded exam.
Question 1 of 20 · Azure Foundations A subscription is moved from mg-sandbox to mg-prod. A deny initiative is assigned at mg-prod, and the subscription already holds a few resources that violate it. What happens after the move? The move is blocked until every violating resource in the subscription has been fixed or deleted first The subscription keeps the mg-sandbox assignments until someone reassigns policy by hand It inherits the mg-prod assignments; existing violators are flagged non-compliant, not removed It inherits the mg-prod assignments, and the deny effect deletes the existing violators
Question 2 of 20 · Azure Foundations An access review export contains this row:
Role: User Access Administrator
Principal: dana@contoso.example (User)
Scope: /providers/Microsoft.Management/managementGroups/mg-grc
How should the reviewer rate this assignment? Highly privileged: she can grant any role, including Owner, on every subscription under mg-grc Low risk: the role only lets her read role assignments across the subscriptions under mg-grc Out of scope: this is an Entra directory role, so it belongs in the directory role review instead Low risk: a role at a management group covers the group object, not the subscriptions below it
Question 3 of 20 · Azure Foundations A Function App is deleted and redeployed from code several times a month. The team wants its identity and that identity's role assignments to stay stable across every redeploy, with no stored secret. Which principal type fits? A system-assigned managed identity, because it is tied directly to the Function App resource itself A service principal with a client secret kept in the Function App's settings A dedicated user account whose password is rotated by the platform team A user-assigned managed identity, created as its own resource and attached to the app
Question 4 of 20 · The Azure GRC Toolkit A control requires every Key Vault to send its resource logs to the central Log Analytics workspace. The diagnostic setting is a separate child resource, and vaults that lack one must be fixed without a human touching them. Which policy effect fits? modify, because it can rewrite properties and tags on the vault at request time deployIfNotExists, because it deploys the missing child resource when absent deny, because it stops any vault from being created until the logs are routed audit, because it records which vaults lack the setting and corrects them on the next scan
Question 5 of 20 · The Azure GRC Toolkit An engineer saves this query in the central workspace:
AzureActivity
| where OperationNameValue =~ 'Microsoft.Authorization/roleAssignments/write'
| where ActivityStatusValue == 'Success'
| summarize grants = count() by Caller
Which auditor question does it answer directly? Which principals hold a role in the subscription today, and at what scope Which principals signed in to the tenant during the retained period, and how often each did Which role assignments have gone unused and are candidates for removal Who granted access during the retained period, and how many times each caller did it
Question 6 of 20 · The Azure GRC Toolkit A subscription has never had a paid Defender plan turned on. What does Defender for Cloud still give the GRC engineer on that subscription? Foundational CSPM: security recommendations on existing resources and a secure score Nothing at all, because assessments are produced only after a paid plan is enabled Workload protections such as malware scanning on storage, limited to one resource group A full set of threat alerts, but no configuration recommendations and no secure score
Question 7 of 20 · Terraform on Azure A stage's backend is configured like this:
backend "azurerm" {
resource_group_name = "rg-tfstate"
storage_account_name = "sttfstate01"
container_name = "tfstate"
key = "foundation.tfstate"
use_azuread_auth = true
}
What must the identity running terraform init hold for state access to work? Contributor on the storage account, so the backend can list the account keys A blob data-plane role such as Storage Blob Data Contributor on the state container Reader on the resource group, since init only downloads the current state file A SAS token for the state container, passed to the backend as an environment variable
Question 8 of 20 · Terraform on Azure A plan job triggered by a pull request fails at the Azure login step:
AADSTS70021: No matching federated identity record found for presented assertion.
Assertion Subject: repo:acme/grc-pipeline:pull_request
The app registration has one federated credential, with subject repo:acme/grc-pipeline:ref:refs/heads/main. What is the right fix? Create a client secret on the app registration and store it as an encrypted repository secret Add id-token: write to the workflow permissions so GitHub is able to issue a token Add a federated credential whose subject matches the pull_request token for this repo Run az login on the runner first so the provider can reuse the cached session
Question 9 of 20 · Terraform on Azure A custom azurerm_policy_definition is created with no management_group_id, so it is stored in the sandbox subscription. The initiative that references it is then assigned at mg-grc-sandbox, and the apply fails. Why? A definition must be stored at or above its assignment scope, so move it to the management group Custom definitions cannot be placed in initiatives, so it has to be assigned individually Management group assignments accept built-in definitions only, so the custom one must be removed The definition needs a managed identity of its own before any assignment is able to reference it
Question 10 of 20 · The Evidence Plane An engineer runs this in Azure Resource Graph:
securityresources
| where type == 'microsoft.security/assessments'
| where properties.status.code == 'Unhealthy'
| summarize count() by tostring(properties.metadata.severity)
What does the result represent? A 90-day history of failed assessments, grouped by the severity each one carried at the time it failed The findings written by the last collector sweep, grouped by severity and pinned to its run A list of resources that were denied at deployment by policy, grouped by severity A count of currently unhealthy Defender assessments by severity, as of right now, with no history
Question 11 of 20 · The Evidence Plane The reports container carries a time-based retention policy. Tonight the generator uploads a new, dated artifact to a path that does not exist yet. What does the platform do? Rejects the upload, because a container under retention is read-only for all callers Accepts the new blob, then refuses changes or deletion of it until retention expires Accepts the new blob, and lets holders of a blob data role overwrite it afterward Queues the upload until an administrator lifts the policy for the write window
Question 12 of 20 · The Evidence Plane The evidence Cosmos account has key-based authentication turned off. To let the collector write documents, a teammate grants its managed identity the Azure RBAC role DocumentDB Account Contributor on the account. Writes still fail with 403. What is the cause? Role grants on Cosmos accounts take about a day to apply, so the writes will begin to succeed The collector must authenticate with the account's primary key, which the role now lets it read That role manages the account; document writes need a Cosmos data-plane role assignment The role was granted at account scope, and Cosmos only honors roles set at the subscription
Question 13 of 20 · The Evidence Plane An assessor asks how many storage accounts were failing the public network access check on a specific date three weeks ago. Which source can answer that, and why does the answer hold up? The assessments container, filtered to that date's run: stored, timestamped documents back the count The Defender for Cloud portal, because each recommendation keeps a dated history per resource Azure Resource Graph, because the Resources table can be queried as of any earlier date The Terraform state for the evidence stage, because state versioning records each resource's posture over time
Question 14 of 20 · Reporting & Narrative An engineering lead asks: which open findings are past their due date, and who owns each one? Which pipeline deliverable is built to answer that question? The SAR The POA&M The OSCAL SSP The weekly digest
Question 15 of 20 · Reporting & Narrative The digest prompt template in the repo reads:
Summarize the attached report JSON for leadership. Cite every figure from the input. Add nothing that is not in the input.
A pull request appends: If a count looks stale, call the Defender API and use the latest number instead. How should the reviewer respond? Approve it, since fresher figures make the leadership summary more accurate Approve it, provided the digest footnotes which of its figures came from the live Defender call Reject it, because the model deployment is unable to make outbound API calls Reject it, since the digest may only restate finished report JSON, never fetch live data
Question 16 of 20 · Reporting & Narrative Last Monday's SAR reports 18 unhealthy findings for its run. Today an executive opens the Defender portal, sees 11, and says the SAR is wrong. The engineer reruns the SAR's stored query pinned to that run and gets 18. What is the correct conclusion? The SAR overcounted, so it should be corrected in place to match the portal's figure The portal is lagging, so its figure should be ignored until it catches up with the store Both are right: the SAR is pinned to its sweep, and the portal shows the state as of today The store is corrupted, so the collector should resweep and then regenerate last Monday's SAR
Question 17 of 20 · Enforcement & Operations A pull request check ends with:
conftest test plan.json --policy policy/
Error: open plan.json: no such file or directory
Process completed with exit code 1.
A developer proposes appending || true to the command so that tooling errors stop blocking merges. How should the proposal be judged? Reject it: a gate that passes when it cannot evaluate the plan is not a control, so fix the plan step Accept it: the Azure deny policy still blocks bad resources, so the gate only adds duplicate friction here Accept it for tooling errors, as long as real rule violations are still reported in the job log Reject it: conftest cannot read JSON plans, so the check should be replaced with terraform validate
Question 18 of 20 · Enforcement & Operations Last night's scheduled terraform plan was clean for every stage. This morning an engineer finds a storage account that someone created by hand in the portal yesterday, in a resource group the pipeline governs. Why did the scheduled plan miss it, and what catches it? The plan ran before the account existed; scheduling the plan every hour would have caught it The plan skips storage accounts by default; a lifecycle block is needed to include them The plan was reading cached state; a terraform refresh would have surfaced the account The plan only compares what is in state; the Activity Log query on other callers catches it
Question 19 of 20 · Capstone A capstone repo's activation stage declares one azurerm_security_center_subscription_pricing resource per Defender plan, each hardcoded to Standard, with no reference to any discovery output. What is the main weakness of this design? Pricing resources cannot be created by Terraform, so the stage will fail on its first apply Activation ignores the measured gap, so it collides with or overrides what a real estate already has Hardcoded tiers stop the collector from reading assessments for the plans being enabled It makes the plan output longer, which slows down review without changing behavior
Question 20 of 20 · Capstone A candidate wants to show that the repo's CI gate really blocks a non-compliant change. Which artifact is the strongest proof? The Rego rule files committed under the policy directory of the repo A passing gate run on the main branch from the day of submission A closed test pull request whose gate run failed on a deliberate violation A README section describing which resource settings the gate will refuse
Submit answers 0 of 20 answered