GCP ACE overview

  • 200 practice questions
  • 32 flashcards
  • 5 study guide topics
  • 39 key terms

What you'll study

  • Setting up a cloud solution environment20%
  • Planning and configuring a cloud solution20%
  • Deploying and implementing a cloud solution20%
  • Ensuring successful operation of a cloud solution20%
  • Configuring access and security20%

The Google Cloud Associate Cloud Engineer certification is a foundational credential for practitioners who deploy and secure applications, services, and infrastructure on Google Cloud. Google frames the role as someone who monitors operations across multiple projects and maintains enterprise solutions to meet target performance metrics, drawing on experience with public cloud or on-premises environments. There are no formal prerequisites, but Google recommends at least six months of hands-on Google Cloud experience before attempting it. The standard exam lasts two hours, costs $125 plus tax where applicable, and contains 50–60 multiple choice and multiple select questions, taken either online with a remote proctor or at an onsite testing center. It is offered in English, Japanese, Spanish, and Portuguese, and the certification is valid for three years. Google publishes an exam guide, a Cloud Engineer learning path, a dedicated preparation course, and sample questions; a sensible order is to read the guide, book hands-on labs for the weak areas, then test yourself with the sample items. This page is independent study material and is not affiliated with, endorsed by, or sponsored by Google.

Common mistakes

Sets the target project once with `gcloud config set project` and then runs every later command without `--project`, including commands issued from another terminal, script, or CI step.

gcloud config is per-configuration and per-shell environment (CLOUD_SDK_CONFIG_DIR). A second terminal, an activated configuration, or a wrapper script silently targets a different project, and exam scenarios often describe multi-project layouts where this exact drift is the point of the question.

What to do instead

Pass --project explicitly on each command, or create and activate named configurations (gcloud config configurations create/activate), and confirm the active context with gcloud config list before acting on any resource.

Assumes a newly created project can immediately create resources, and never checks API enablement, billing linkage, or org policy constraints before diagnosing a failure as an IAM problem.

Labs and tutorials begin from an already-configured project, so the setup steps are invisible. On the exam the symptom is usually PERMISSION_DENIED or SERVICE_DISABLED, which look like missing roles but are actually unenabled services, an unlinked billing account, or a constraint such as restricting external IPs.

What to do instead

Treat gcloud services enable, billing linkage, and org policy constraints as distinct setup steps, and make them the first thing to verify whenever a resource creation fails, before touching IAM bindings.

Chooses a location flag without first classifying the resource as zonal, regional, or global - for example passing `--zone` when creating a regional GKE control plane, or expecting a Cloud Storage bucket or a global load balancer address to be regional.

Scope is hidden behind console defaults and the flag names differ per product (zone, region, location, multi-region) while sounding interchangeable, so candidates pattern-match on the word 'location' instead of on the resource type.

What to do instead

Memorize scope per resource: Compute instances, persistent disks and GKE nodes are zonal; subnets, Cloud SQL instances and the GKE control plane are regional; buckets, container images, global load balancers, service accounts and IAM policy are global. Let the flag name in the question be the hint.

Designs the network assuming the default VPC, its auto-mode subnets, and its permissive firewall rules exist in every region, and assumes VPC firewall rules also gate traffic to Cloud SQL, App Engine, or to Google's load balancer health checkers.

Hands-on practice usually happens in a project with the default network, which hides the need for custom-mode subnets and explicit ingress rules. Firewall rules attach to VM network interfaces only, and health check probes originate from Google-owned ranges rather than from the VPC.

What to do instead

In custom mode, create one subnet per region, write ingress rules scoped by target tags or service accounts, and allow 35.191.0.0/16 and 130.211.0.0/22 for health checks. Remember Cloud SQL reachability is controlled by authorized networks or private IP, not by VPC firewall rules.

Pushes or pulls container images without first running `gcloud auth configure-docker` for the correct registry host in the same environment, or mixes hostnames - using `gcr.io/<project>/...` when the image lives in Artifact Registry's `<location>-docker.pkg.dev/<project>/<repo>/...`.

Container Registry and Artifact Registry are separate services with different hostnames, IAM models, and credential helpers. A Docker daemon configured once for one host keeps working for that host, so the mismatch only surfaces when the registry or region changes.

What to do instead

Use the fully qualified Artifact Registry path <location>-docker.pkg.dev/<project>/<repo>/<image>, run gcloud auth configure-docker <location>-docker.pkg.dev on the machine that builds or pushes, and grant the runtime identity roles/artifactregistry.reader on the repository.

Deploys a containerized or source-based service and then treats the resulting 403 or build failure as a code bug - for example expecting an unauthenticated Cloud Run service by default, or assuming App Engine standard will build an arbitrary Dockerfile.

gcloud run deploy succeeds and requires authentication unless told otherwise, so the failure appears later at request time; App Engine standard supports a fixed set of runtimes while only the flexible environment consumes a Dockerfile, so the deploy fails at build time for a reason unrelated to the application.

What to do instead

Add --allow-unauthenticated for public Cloud Run services and call private ones with an identity token. Choose App Engine flexible or Cloud Run when a custom container is required, and standard only when the runtime is on the supported list.

Searches Cloud Logging for Data Access entries (for example a `storage.objects.get`) to prove who read an object, finds nothing, and concludes the read never happened or that logging is broken.

Admin Activity and System Event logs are always enabled, but Data Access logs are disabled by default for most services (BigQuery being the notable exception) and must be explicitly turned on, and they are billable - so the absence of a read entry is expected on a default project.

What to do instead

Know the four log types and their default states, enable Data Access logs explicitly at the project, folder, or org level before relying on them, and use Admin Activity logs to prove configuration changes rather than data reads.

Creates an alerting policy or uptime check and waits for a notification, or expects memory, disk, and process metrics from a Compute Engine instance without installing an agent.

An alerting policy with no notification channel attached fires silently, uptime checks need a reachable endpoint and a firewall path from the checker locations, and the platform-provided VM metrics cover CPU and network only - memory and disk require the Ops Agent.

What to do instead

Attach a notification channel to every alerting policy, confirm the uptime check's source ranges are allowed through the firewall, and install the Ops Agent whenever the scenario asks for memory, disk, or process-level metrics.

Grants a human `roles/editor` at project level and then expects them to deploy a VM, GKE node pool, or Cloud Function that runs as a custom service account.

Attaching a service account to a resource is an actAs operation requiring iam.serviceAccounts.actAs (contained in roles/iam.serviceAccountUser). Project-level edit does not imply it, and neither does holding a key for that service account, so the deploy fails with a permission error that looks unrelated to the role that was granted.

What to do instead

Grant roles/iam.serviceAccountUser on the service account (or at project level) to the deploying principal, and independently keep the runtime service account's own permissions scoped to what the workload actually calls.

Responds to a permission error by adding a broad primitive role at project level, or tries to remove an inherited grant by editing the project-level policy of the same principal.

IAM policies are additive and inherited down the resource hierarchy, so a grant made at the organization or folder cannot be revoked by a project-level change, and primitive roles such as editor include thousands of permissions that mask which one was actually required.

What to do instead

Identify the specific API permission being denied, select the least-privileged predefined or custom role, bind it at the highest resource in the hierarchy where it is needed, and remember that removing access means changing the binding at the level where it was granted.

Exam requirements can change. Confirm current details with the exam provider.

GCP ACE practice

200 questions. Each one lists the choices, the correct answer, and an explanation. With JavaScript, this list opens one question at a time. The checker at the top of this page saves progress in this browser when it is available.

  1. Question 1Setting up a cloud solution environment

    Which sequence lists Google Cloud resource hierarchy levels from broadest scope to narrowest scope?

    1. Folder, organization, project, resource
    2. Organization, folder, project, resource Correct answer.
    3. Project, folder, organization, resource
    4. Organization, project, folder, resource

    Explanation

    Google Cloud arranges resources in a tree whose root is the organization, followed by folders, then projects, then individual resources such as VMs or buckets. Policy applied higher in the tree is inherited by everything below it.

  2. Question 2Setting up a cloud solution environment

    A team has created a new project but has not linked any billing account to it. What is the result?

    1. The project cannot be created until a billing account is attached.
    2. All APIs in the project are permanently disabled.
    3. Resources run normally and charges go to the organization's default billing account.
    4. Billable services such as Compute Engine cannot be used until a billing account is linked. Correct answer.

    Explanation

    A project can exist without billing, and free operations such as creating folders or using certain always-free quotas still work, but any billable service requires a linked billing account. Google Cloud does not silently fall back to another billing account.

  3. Question 3Setting up a cloud solution environment

    You create a Cloud Billing budget of $500 with an alert threshold set at 100%. Actual spend reaches $650. What happens?

    1. Creation of new resources is blocked until the budget is increased.
    2. Compute Engine instances are stopped automatically.
    3. The billing account is closed and all services are suspended.
    4. Alert recipients are notified by email, and the resources keep running. Correct answer.

    Explanation

    Budgets in Cloud Billing are monitoring and alerting tools only; they send notifications to the configured recipients when thresholds are crossed. They never cap spending or disable resources unless you separately build automation such as Pub/Sub-triggered Cloud Functions.

  4. Question 4Setting up a cloud solution environment

    An IAM role is granted to a group at the folder level. Which resources does that grant affect?

    1. Only the projects directly inside the folder, not the resources within them.
    2. Only the folder itself.
    3. Only projects created after the grant was made.
    4. All projects and resources nested under that folder. Correct answer.

    Explanation

    IAM policies are inherited down the resource hierarchy, so a binding on a folder applies to every project and resource beneath it. A more specific policy can add permissions but cannot remove an inherited grant.

  5. Question 5Setting up a cloud solution environment

    Which command makes my-project the default project for subsequent gcloud commands in the active configuration?

    1. gcloud config set project my-project Correct answer.
    2. gcloud config set compute/project my-project
    3. gcloud init my-project
    4. gcloud projects set my-project

    Explanation

    The gcloud config set command writes a property into the active configuration, and the property for the default project is 'project'. The change applies to all later commands run with that configuration until it is changed again.

  6. Question 6Setting up a cloud solution environment

    Which statement about Cloud Shell is accurate?

    1. The home directory provides unlimited persistent storage.
    2. Cloud Shell can only be used after installing the Cloud SDK locally.
    3. The virtual machine is ephemeral and recreated per session, while files in the home directory persist within a 5 GB limit. Correct answer.
    4. Files written to the home directory are deleted when the session ends.

    Explanation

    Each Cloud Shell session runs on an ephemeral Compute Engine VM with the Cloud SDK, gcloud, and common tools preinstalled. A separate persistent disk keeps the home directory across sessions up to 5 GB.

  7. Question 7Setting up a cloud solution environment

    Which command enables the Compute Engine API on the currently configured project?

    1. gcloud services enable compute.googleapis.com Correct answer.
    2. gcloud services activate compute-googleapis-com
    3. gcloud api-services enable compute
    4. gcloud compute enable api

    Explanation

    Service enablement is handled by the services command group, and each API is referenced by its service name such as compute.googleapis.com. The command acts on the project in the active configuration unless --project is supplied.

  8. Question 8Setting up a cloud solution environment

    Which command creates a project named acme-prod inside the folder whose ID is 123456789012?

    1. gcloud projects create acme-prod --organization=123456789012
    2. gcloud resource-manager projects create acme-prod --folder=123456789012
    3. gcloud projects create acme-prod --parent-folder=123456789012
    4. gcloud projects create acme-prod --folder=123456789012 Correct answer.

    Explanation

    The projects create command accepts a --folder flag whose value is the numeric folder ID, which places the new project under that folder. Using --organization instead would place the project directly under the organization.

  9. Question 9Setting up a cloud solution environment

    An organization must ensure that no project can create resources outside approved European locations. Which control accomplishes this?

    1. Apply the constraints/gcp.resourceLocations organization policy with allowed values limited to European locations. Correct answer.
    2. Grant roles/editor only to users who are based in Europe.
    3. Create a Cloud Billing budget filtered to non-European regions.
    4. Set the default compute region to europe-west1 in every project.

    Explanation

    Organization policies are enforced by Google Cloud at the API layer and define which locations may be used for resource creation. Setting a default region only changes where new resources land when no region is specified, so it does not prevent someone from choosing another location.

  10. Question 10Setting up a cloud solution environment

    A deployment fails with a quota exceeded error for CPUs in us-central1. What is the appropriate response?

    1. Request a quota increase for the project in the Quotas page (or with gcloud) and wait for adjustment. Correct answer.
    2. Move the deployment to a different zone without checking available quota.
    3. Re-run the deployment several times so the quota adjusts automatically.
    4. Delete the project and recreate it, because new projects receive fresh quota.

    Explanation

    CPU quota is an allocation limit that must be raised through a quota increase request, which is then reviewed and granted for the project. Many quotas are per project per region, so changing zones or projects is not a substitute for a valid increase.

  11. Question 11Setting up a cloud solution environment

    Which statement correctly distinguishes labels from tags in Google Cloud?

    1. Labels are key-value pairs used for organizing and cost analysis, while tags drive behaviour such as firewall rules and conditional IAM. Correct answer.
    2. Tags are key-value pairs used for billing analysis, while labels are used for firewall rules.
    3. Labels can be attached only at the organization level.
    4. Tags appear in billing exports while labels never do.

    Explanation

    Labels are metadata key-value pairs useful for inventory, filtering, and splitting billing reports. Tags are named resources with their own IAM and are referenced by controls such as firewall rules and IAM conditions.

  12. Question 12Setting up a cloud solution environment

    A developer wants a Python script running on their laptop to call a Google Cloud API with their own user identity, without downloading a service account key. Which command should they run first?

    1. gcloud auth login
    2. gcloud auth application-default login Correct answer.
    3. gcloud auth activate-service-account
    4. gcloud config set auth/impersonate_service_account

    Explanation

    Client libraries look for Application Default Credentials, which gcloud auth application-default login writes to a local well-known file using the developer's user identity. gcloud auth login only authenticates the gcloud CLI itself and does not create ADC for library code.

  13. Question 13Setting up a cloud solution environment

    A CI pipeline running outside Google Cloud must act as a service account, and the security team forbids long-lived key files. Which approach is recommended?

    1. Grant the pipeline identity roles/iam.serviceAccountTokenCreator on the service account and use impersonation. Correct answer.
    2. Add the pipeline identity to the service account's IAM policy with roles/owner.
    3. Attach the default Compute Engine service account to the pipeline runners.
    4. Generate a JSON key for the service account and store it as a pipeline secret.

    Explanation

    Service account impersonation produces short-lived credentials and requires the caller to hold the Token Creator role on the target service account. This removes the need to create and distribute key files that can be leaked or forgotten.

  14. Question 14Setting up a cloud solution environment

    You want to run detailed SQL analysis of daily cost by project and by label. What should you enable?

    1. The Cloud Billing API called with an API key.
    2. Cloud Billing export of cost data to a BigQuery dataset. Correct answer.
    3. A Cloud Logging sink that filters billing log entries.
    4. Cloud Billing budgets with Pub/Sub notifications.

    Explanation

    Billing export writes detailed usage and cost records, including label and project attribution, into a BigQuery dataset where they can be queried. Budgets and Pub/Sub notifications report thresholds rather than detailed line-item costs.

  15. Question 15Setting up a cloud solution environment

    What must exist before an organization resource for the domain example.com can be created in Google Cloud?

    1. At least one project already owned by the domain administrator.
    2. A billing account whose address matches the domain registration.
    3. A Cloud Identity or Google Workspace account that has verified ownership of the domain. Correct answer.
    4. A folder named after the domain.

    Explanation

    The organization resource is created automatically when a domain verified through Cloud Identity or Google Workspace is associated with Google Cloud. Without that verified domain there is no organization node, and projects exist in a 'No organization' state.

  16. Question 16Setting up a cloud solution environment

    Which predefined role gives full control over all resources in a project, including the ability to grant roles to other principals?

    1. roles/owner Correct answer.
    2. roles/viewer
    3. roles/editor
    4. roles/resourcemanager.projectCreator

    Explanation

    The basic Owner role includes all Editor permissions plus the ability to manage IAM policies and billing for the project. Editor can modify resources but cannot grant access to others.

  17. Question 17Setting up a cloud solution environment

    An administrator deletes a project that still contains data. What happens to that project?

    1. It remains fully active for 7 days and continues to be billed normally.
    2. Deletion is immediate and irreversible.
    3. It is shut down immediately and can be restored within a 30-day window before permanent deletion. Correct answer.
    4. It is moved into a Trash folder for 90 days.

    Explanation

    Deleting a project immediately shuts it down, so its services stop and API calls fail, but the project enters a pending-deletion period of roughly 30 days. During that window an authorized user can restore it; afterwards it is permanently removed.

  18. Question 18Setting up a cloud solution environment

    Which statement about a Google Cloud project ID is true?

    1. It must be unique only within its organization.
    2. It is a display name that may contain spaces.
    3. It can be renamed later from the project settings page.
    4. It is globally unique and cannot be changed after creation. Correct answer.

    Explanation

    The project ID is used in API calls and resource names, so it must be unique across all of Google Cloud and is immutable. The project name and labels can be changed freely, but the ID cannot.

  19. Question 19Setting up a cloud solution environment

    An engineer regularly switches between dev, staging, and production projects, each with a different account and default region. Which feature should be used?

    1. Setting the project before each command and leaving the account unchanged.
    2. Reinstalling the Cloud SDK whenever the environment changes.
    3. Running the Cloud SDK as a different operating system user for each environment.
    4. Named gcloud configurations that store account, project, and region per environment. Correct answer.

    Explanation

    A gcloud configuration bundles properties such as core/account, core/project, and compute/region together, and named configurations can be activated or invoked per command. This keeps credentials and defaults separate for each environment.

  20. Question 20Setting up a cloud solution environment

    A deny policy on a folder denies storage.objects.delete for a group, while a project-level allow policy grants the same group roles/storage.admin. What is the effective outcome?

    1. The group can delete objects because the project allow policy overrides the folder deny policy.
    2. The outcome depends on the order in which the bindings were created.
    3. The group can delete objects because deny policies apply only to service accounts.
    4. The group cannot delete objects, because deny policies take precedence over allow policies. Correct answer.

    Explanation

    IAM deny policies are evaluated before allow policies, so an explicit deny blocks the action even where an allow binding exists lower in the hierarchy. This makes deny policies a reliable guardrail against overly broad grants.

  21. Question 21Setting up a cloud solution environment

    Which command sets the default zone used by gcloud compute commands in the active configuration?

    1. gcloud config set compute/zone us-central1-a Correct answer.
    2. gcloud compute config set default-zone us-central1-a
    3. gcloud config set zone us-central1-a
    4. gcloud compute zones set us-central1-a

    Explanation

    Default location properties live under the compute section of the gcloud configuration, so the property name is compute/zone. Once set, compute commands that need a zone use this value unless --zone is supplied.

  22. Question 22Setting up a cloud solution environment

    Which command creates a folder named Prod directly under the organization with ID 123456789012?

    1. gcloud organizations create-folder Prod
    2. gcloud resource-manager folders create --display-name=Prod --organization=123456789012 Correct answer.
    3. gcloud resource-manager folders create --name=Prod --parent=123456789012
    4. gcloud projects create --folder=Prod

    Explanation

    Folders are created with the resource-manager folders create command, using --display-name for the human-readable name and --organization or --folder for the parent. The parent must be referenced by the numeric organization or folder ID.

  23. Question 23Setting up a cloud solution environment

    Security and billing notifications must reach a small set of named people rather than everyone holding a broad role. What should be configured?

    1. Cloud Audit Logs sinks that publish to Pub/Sub.
    2. An organization policy constraint restricting contact email domains.
    3. Essential Contacts with the relevant notification categories on the resource. Correct answer.
    4. A Google Group bound to roles/owner at the organization.

    Explanation

    Essential Contacts let you designate individuals per notification category, such as SECURITY, BILLING, and TECHNICAL, at the organization, folder, or project level. Google then sends those notifications to the designated contacts instead of to every principal with a broad role.

  24. Question 24Setting up a cloud solution environment

    An engineer must be able to associate new projects with an existing billing account but must not change payment methods or manage budgets. Which role should be granted?

    1. roles/billing.user Correct answer.
    2. roles/editor
    3. roles/billing.viewer
    4. roles/billing.admin

    Explanation

    Billing Account User allows a principal to link projects to a billing account without granting control over payment instruments or budgets. Administrator adds full control of the billing account, while Viewer is read-only and cannot link projects.

  25. Question 25Setting up a cloud solution environment

    Which statement describes the default Compute Engine service account created in every project and a recommended practice for it?

    1. It holds Owner on the organization but no project-level access.
    2. It has no permissions until an administrator binds a role to it.
    3. It can be used only from resources inside Google Cloud networks.
    4. It receives the Editor role on the project by default, and it should be replaced by a least-privilege service account. Correct answer.

    Explanation

    The default Compute Engine service account is granted Editor on its project, which is broader than most workloads require. Best practice is to create dedicated service accounts with narrow roles and to disable or restrict the default account.

  26. Question 26Setting up a cloud solution environment

    A team has just created a new Google Cloud project and wants to launch Compute Engine virtual machines in it. What must be associated with the project before billable resources can be created?

    1. A Cloud Identity domain
    2. A Shared VPC host project
    3. A Cloud Billing account Correct answer.
    4. An organization policy

    Explanation

    Paid Google Cloud services require the project to be linked to an active Cloud Billing account; without that link, resource creation attempts fail with a billing-related error. Identity domains, shared networking, and governance policies are optional and unrelated to the ability to incur charges.

  27. Question 27Setting up a cloud solution environment

    An engineer already has the Cloud SDK installed on a workstation and needs to add the kubectl command-line tool to that installation. Which command should be used?

    1. gcloud components install kubectl Correct answer.
    2. gcloud components update kubectl
    3. gcloud install kubectl
    4. gcloud sdk add kubectl

    Explanation

    Optional tools such as kubectl are distributed as Cloud SDK components and are added with the components install subcommand of gcloud. The components update subcommand only refreshes components that are already present.

  28. Question 28Setting up a cloud solution environment

    Which resource type in the Google Cloud resource hierarchy can directly contain both projects and additional resources of its own type?

    1. Service account
    2. Billing account
    3. Organization
    4. Folder Correct answer.

    Explanation

    Folders are grouping nodes that may contain projects as well as other folders, which is how multi-level hierarchies are built below the organization. The organization node sits above folders and cannot contain another organization, while billing and service accounts are not part of the resource containment hierarchy.

  29. Question 29Setting up a cloud solution environment

    Which project identifier is assigned by Google, is unique across all of Google Cloud, and is never reused, even if the project is deleted?

    1. Project name
    2. Project ID
    3. Project number Correct answer.
    4. Project label

    Explanation

    The project number is generated automatically when the project is created and is never reassigned, which makes it a stable identifier for billing and support purposes. The project ID is also unique but is chosen by the creator, while names and labels are user-controlled metadata that can be changed or duplicated.

  30. Question 30Setting up a cloud solution environment

    An administrator grants roles/storage.objectViewer to a group on a folder. A project inside that folder has no IAM binding mentioning the group. What access do group members have to Cloud Storage objects in that project?

    1. Read-only access only after an administrator copies the binding into the project
    2. Full control, because folder-level grants are expanded into basic roles at the project
    3. Read-only access, because allow policies are inherited down the resource hierarchy Correct answer.
    4. No access, because the project has no matching binding of its own

    Explanation

    IAM allow policies are inherited, so a role granted on a folder applies to every project and resource beneath it, and effective permissions are the union of all bindings along the ancestry path. Copying bindings into children is unnecessary and does not change the effective result.

  31. Question 31Setting up a cloud solution environment

    A team configures a Cloud Billing budget of $1,000 per month with alert thresholds at 50%, 90%, and 100%, scoped to one project. Actual spend for the month reaches $1,200. What is the outcome?

    1. Alert notifications are sent and resource usage continues unchanged Correct answer.
    2. The project is detached from its billing account once the threshold is crossed
    3. The budget amount is raised automatically to cover the overage
    4. Resources in the project are stopped until the next billing cycle begins

    Explanation

    A Cloud Billing budget is a monitoring construct: it triggers notifications when actual or forecast spend crosses configured thresholds but does not block, throttle, or modify resource usage. Stopping spend requires separate mechanisms such as quota limits or programmatic responses to the alerts.

  32. Question 32Setting up a cloud solution environment

    An organization must prevent any IAM policy in its hierarchy from granting access to identities outside its own domain. Which organization policy constraint should be enforced?

    1. constraints/iam.allowedPolicyMemberDomains Correct answer.
    2. constraints/iam.disableServiceAccountKeyCreation
    3. constraints/gcp.resourceLocations
    4. constraints/compute.vmExternalIpAccess

    Explanation

    The allowedPolicyMemberDomains constraint restricts the set of domains from which principals may appear in IAM allow policies, which is the supported way to enforce domain-restricted sharing. The other constraints govern key creation, external IP usage, and resource locations rather than principal identity.

  33. Question 33Setting up a cloud solution environment

    An engineer needs a machine-readable list of every project in the current organization whose project ID begins with prod-. Which command accomplishes this?

    1. gcloud projects list --format=json --prefix=prod-
    2. gcloud projects describe prod-* --format=json
    3. gcloud projects list --filter="projectId:prod-*" --format=json Correct answer.
    4. gcloud organizations list --filter="id:prod-*" --format=json

    Explanation

    The list verb accepts a --filter expression that can match a substring of the projectId field, and --format=json returns structured output suitable for scripts. describe operates on exactly one resource identifier and cannot expand a pattern.

  34. Question 34Setting up a cloud solution environment

    A security team needs a durable record of every configuration change made to resources in a project, available from the moment the project was created. Which log type satisfies this requirement by default?

    1. Admin Activity audit logs Correct answer.
    2. VPC Flow Logs
    3. Access Transparency logs
    4. Data Access audit logs

    Explanation

    Admin Activity audit logs record metadata-changing operations such as resource creation, updates, and deletions, and they are enabled for every project without any configuration. Data Access logs, by contrast, must be explicitly enabled and are typically higher volume.

  35. Question 35Setting up a cloud solution environment

    An engineer who can create Compute Engine instances receives PERMISSION_DENIED on iam.serviceAccounts.actAs when attaching an existing user-managed service account to a new VM. Which role should be granted on that service account?

    1. roles/iam.serviceAccountKeyAdmin
    2. roles/iam.serviceAccountTokenCreator
    3. roles/iam.serviceAccountUser Correct answer.
    4. roles/iam.workloadIdentityUser

    Explanation

    Attaching a service account to a resource is governed by the actAs permission, which is contained in the Service Account User role. Key administration and token creation grant different capabilities and do not permit attaching the identity to a VM.

  36. Question 36Setting up a cloud solution environment

    A contractor should hold roles/browser on a project only until the last day of the month, after which the access must lapse with no further administrator action. Which approach meets this requirement?

    1. Create a custom role with the same permissions and delete the custom role at month end
    2. Add an IAM condition with an expiration timestamp to the role binding Correct answer.
    3. Grant the role on the organization node and revoke it manually at month end
    4. Add the contractor to a Google Group and remove the group membership later

    Explanation

    IAM Conditions support time-based expressions, so a binding can be configured to stop granting the role after a specified timestamp without any follow-up work. The alternatives either grant broader access than needed or depend on a manual step that may be forgotten.

  37. Question 37Setting up a cloud solution environment

    Which command initializes a fresh Cloud SDK installation by authorizing the user, letting them choose or create a configuration, and setting a default project and region?

    1. gcloud components update
    2. gcloud init Correct answer.
    3. gcloud config set project
    4. gcloud auth login

    Explanation

    gcloud init is the guided setup that performs authorization, configuration selection or creation, and default property selection in one flow. The other commands perform only one narrow part of that setup and assume prerequisites already exist.

  38. Question 38Setting up a cloud solution environment

    An administrator deletes a folder that contains two projects and one nested folder. What is the result?

    1. The folder becomes read-only and its contents are preserved for 30 days
    2. The operation fails until every contained project is deleted first
    3. The folder is deleted and its projects are moved up to the organization root
    4. The folder is deleted, and the projects and nested folder inside it are also deleted Correct answer.

    Explanation

    Deleting a folder is a cascading operation: every folder and project beneath it is deleted along with it, and the affected projects enter pending deletion. Administrators must therefore confirm contents before removing a folder.

  39. Question 39Setting up a cloud solution environment

    An organization node enforces constraints/compute.vmExternalIpAccess with a deny-all list, and a project owner then adds that project to an allow list for the same constraint in a project-level policy. What is the effective behavior for the project?

    1. External IPs are allowed, because the project-level policy is more specific
    2. External IPs remain denied, because a child policy cannot loosen an enforced ancestor constraint Correct answer.
    3. External IPs are allowed, but only for instances created by the project owner
    4. The project-level policy is rejected when the administrator attempts to save it

    Explanation

    Organization policies are inherited, and a policy enforced on an ancestor resource continues to apply to all descendants; child resources can tighten restrictions but cannot relax what an ancestor enforces. The effective result is therefore the stricter of the policies in the ancestry path.

  40. Question 40Setting up a cloud solution environment

    A company needs a custom role that combines permissions from several services and that can be granted on any project in the organization. Where must the custom role be defined?

    1. At the organization level, so it can be granted on all descendant resources Correct answer.
    2. Separately in each project, because custom roles are always project-scoped
    3. On the billing account that pays for those projects
    4. Inside a folder, which automatically propagates the role to its projects

    Explanation

    Custom roles can be created at the organization level or at the project level; only an organization-level custom role can be granted on any project, folder, or other resource in that organization. Project-level definitions exist but restrict where the role may be bound.

  41. Question 41Planning and configuring a cloud solution

    Which resource is the top-level node of the Google Cloud resource hierarchy for an organization?

    1. A billing account
    2. A service account
    3. A project
    4. The organization resource Correct answer.

    Explanation

    The resource hierarchy is organization, then folders, then projects, then the resources inside projects. The organization node represents the company and is the root from which all folders and projects descend.

  42. Question 42Planning and configuring a cloud solution

    You need to group Compute Engine instances by cost center so the finance team can filter costs in billing reports. What should you use?

    1. Network tags on each instance
    2. A separate project for each instance
    3. Labels on each instance Correct answer.
    4. An instance name prefix per cost center

    Explanation

    Labels are key-value pairs that are exported with billing data, so costs can be filtered and grouped by label in billing reports and BigQuery exports. Tags are for networking, and names are not a supported cost filter.

  43. Question 43Planning and configuring a cloud solution

    Which statement about Compute Engine zones is accurate?

    1. Zones share one power source
    2. A region contains one zone
    3. Each zone spans three regions and one network
    4. Zones are isolated failure domains Correct answer.

    Explanation

    A region is a collection of independent zones with separate power, cooling, and networking, so an outage in one zone does not affect the others. Deploying across zones protects an application from a single-zone failure.

  44. Question 44Planning and configuring a cloud solution

    Which resource must be linked to a project before most billable Google Cloud services can be used?

    1. A Cloud DNS managed zone
    2. An organization policy
    3. A service account
    4. A billing account Correct answer.

    Explanation

    A project needs an active billing account attached before it can consume billable services such as Compute Engine or Cloud SQL. Projects without billing can only use services and quotas that are free of charge.

  45. Question 45Planning and configuring a cloud solution

    Virtual machine creation in a new region fails with a RESOURCE_EXHAUSTED error mentioning a limit. What is the most likely cause?

    1. The boot image has been deprecated
    2. The VPC has no firewall rule allowing SSH
    3. The service account lacks the Compute Admin role
    4. A regional CPU quota is too low Correct answer.

    Explanation

    Quota limits are enforced per project, per region, and per resource type, and exceeding them returns RESOURCE_EXHAUSTED. The fix is to request a quota increase in that region or choose another region.

  46. Question 46Planning and configuring a cloud solution

    Which Cloud Storage class is intended for data accessed less than once per month and has a 30-day minimum storage duration?

    1. Nearline Correct answer.
    2. Standard
    3. Archive
    4. Coldline

    Explanation

    Nearline is the low-frequency class with a 30-day minimum storage duration and lower storage cost than Standard. Coldline and Archive have longer minimum durations of 90 and 365 days respectively.

  47. Question 47Planning and configuring a cloud solution

    A batch job can tolerate interruptions and will finish within a few hours. Which Compute Engine purchasing option gives the lowest cost?

    1. Spot VMs Correct answer.
    2. Custom machine types
    3. Reserved sole-tenant nodes
    4. Committed use discounts

    Explanation

    Spot VMs use spare capacity at a large discount and can be reclaimed with 30 seconds' notice, which suits fault-tolerant batch work. Sole tenancy and committed use discounts target different requirements.

  48. Question 48Planning and configuring a cloud solution

    A service account must only read objects in a Cloud Storage bucket and change nothing. Which role should be granted?

    1. roles/storage.objectViewer Correct answer.
    2. roles/storage.objectCreator
    3. roles/storage.admin
    4. roles/iam.securityAdmin

    Explanation

    roles/storage.objectViewer grants read-only access to objects and their metadata, matching least privilege for a read-only consumer. Broader or write-oriented roles would exceed the stated requirement.

  49. Question 49Planning and configuring a cloud solution

    A central networking team must own and manage subnets while application teams deploy their own resources into those subnets. Which design fits?

    1. Shared VPC with a host project Correct answer.
    2. A separate VPC per team joined by Cloud VPN
    3. Cloud NAT deployed in each team project
    4. VPC Network Peering between the team projects

    Explanation

    In Shared VPC the host project owns the VPC network and subnets, and service projects attach to it, so the network team controls IP space while app teams create resources. This centralizes network administration without duplicating VPCs.

  50. Question 50Planning and configuring a cloud solution

    A relational workload must scale horizontally across regions and provide 99.999% availability for transactions. Which database should be chosen?

    1. Cloud Spanner Correct answer.
    2. Firestore in Datastore mode
    3. Cloud Bigtable
    4. Cloud SQL for PostgreSQL

    Explanation

    Cloud Spanner is a globally distributed relational database with horizontal scaling and a 99.999% SLA for multi-region configurations. Cloud SQL is regional and does not scale writes horizontally.

  51. Question 51Planning and configuring a cloud solution

    A multi-terabyte BigQuery table is almost always queried with a filter on its event_date column. What reduces both bytes scanned and query time?

    1. Exporting the table to Cloud Storage and querying the files
    2. Partitioning the table on the event_date column Correct answer.
    3. Creating a logical view over the table
    4. Adding a search index on every column

    Explanation

    Date-partitioned tables prune partitions that do not match the filter, so only the relevant partitions are scanned and billed. A view does not change the amount of data read from the base table.

  52. Question 52Planning and configuring a cloud solution

    You need a single global anycast IP that terminates TLS and distributes HTTP traffic across backends in several regions. Which product meets this?

    1. Global external Application Load Balancer Correct answer.
    2. Internal Application Load Balancer
    3. Regional external passthrough Network Load Balancer
    4. Proxy Network Load Balancer

    Explanation

    The global external Application Load Balancer provides a single anycast IP, TLS termination, and cross-region HTTP(S) load balancing with URL-based routing. Regional and internal load balancers are scope-limited.

  53. Question 53Planning and configuring a cloud solution

    You must ensure that no one in the organization can create Compute Engine instances with external IP addresses. What should you configure?

    1. An IAM deny policy on the Compute Engine API
    2. A hierarchical firewall rule that denies all egress traffic on every VPC
    3. The compute.vmExternalIpAccess organization policy constraint Correct answer.
    4. A Cloud NAT gateway in every region and subnet

    Explanation

    The compute.vmExternalIpAccess organization policy constraint denies creation of instances with external IPs across the selected organization, folder, or project scope. It is enforced by the platform rather than by network rules.

  54. Question 54Planning and configuring a cloud solution

    Finance wants an email alert at 80% of a project's monthly spend and a daily feed of line-item costs for analysis. What should be set up?

    1. An export of Cloud Logging entries to BigQuery
    2. Purchasing committed use discounts for all workloads
    3. A Cloud Monitoring uptime check on the billing API
    4. A budget with alert rules plus billing export Correct answer.

    Explanation

    A Cloud Billing budget with threshold rules sends alerts to recipients or Pub/Sub, and billing export writes detailed cost data to BigQuery for querying. Together they cover both notification and analysis.

  55. Question 55Planning and configuring a cloud solution

    Instances without external IP addresses must reach Google APIs and also download OS packages from public repositories. What is required?

    1. Private Google Access and Cloud NAT Correct answer.
    2. A VPC firewall rule that allows all egress traffic
    3. Private Google Access alone
    4. An external IP address on every VM

    Explanation

    Private Google Access lets instances without external IPs reach Google APIs, while Cloud NAT provides outbound internet connectivity for non-Google destinations such as OS package mirrors. Both are needed for the two traffic types.

  56. Question 56Planning and configuring a cloud solution

    A platform team wants identical projects, VPCs, and firewall rules deployed across dev, staging, and prod with an auditable change history. Which approach fits best?

    1. Saved Console pages bookmarked for each environment
    2. Cloud Shell scripts copied onto each engineer's laptop
    3. Terraform configurations with remote state Correct answer.
    4. Recreate the setup by hand with gcloud commands in each environment

    Explanation

    Terraform declaratively describes the target infrastructure, keeps it in version control for audit, and remote state keeps the deployed resources consistent across environments. Manual steps do not provide repeatability or drift detection.

  57. Question 57Planning and configuring a cloud solution

    An existing subnet hosting live VMs is running out of available IP addresses. What can be done without recreating the subnet or the VMs?

    1. Convert the primary range into a secondary range
    2. Expand the subnet's primary IP range Correct answer.
    3. Move the VMs to a different region
    4. Add a second primary range to the subnet

    Explanation

    A subnet's primary IPv4 range can be expanded to a larger prefix without downtime as long as the new range stays inside the VPC and does not overlap other subnets. Secondary ranges serve alias IP use cases and cannot replace the primary range.

  58. Question 58Planning and configuring a cloud solution

    A small team has stateless containerized HTTP services with sporadic traffic and wants to minimize operational overhead. Which platform should they choose?

    1. GKE Standard with node auto-provisioning
    2. Cloud Run Correct answer.
    3. A Compute Engine managed instance group
    4. App Engine flexible environment

    Explanation

    Cloud Run runs containers on a fully managed, request-driven platform that scales to zero and needs no cluster or node management. It is the lowest-overhead option for stateless HTTP services.

  59. Question 59Planning and configuring a cloud solution

    A company needs a dedicated 10 Gbps link with predictable latency between its data center and a VPC. Which connectivity option should be planned?

    1. HA VPN over the public internet
    2. Dedicated Interconnect Correct answer.
    3. VPC peering with the on-premises network
    4. Cloud NAT with a static IP

    Explanation

    Dedicated Interconnect provides a private physical connection at up to 10 Gbps or 100 Gbps per link with consistent latency. VPN options run over the public internet and cannot guarantee that predictability.

  60. Question 60Planning and configuring a cloud solution

    What is the recommended way to grant IAM roles to a large group of engineers who need the same access?

    1. Grant roles to a Google Group and manage its membership Correct answer.
    2. Grant roles to each engineer individually
    3. Grant roles/owner at the project level
    4. Distribute a shared service account key to the team

    Explanation

    Binding roles to a Google Group centralizes access management: membership changes take effect immediately without editing IAM policies. It also avoids the sprawl of per-user bindings.

  61. Question 61Planning and configuring a cloud solution

    In a Shared VPC, which grant lets a principal in a service project create instances that use a subnet owned by the host project?

    1. roles/compute.networkUser on the host project or subnet Correct answer.
    2. roles/viewer on the host project
    3. roles/compute.networkAdmin on the service project
    4. roles/browser on the host project

    Explanation

    Using a shared subnet requires the compute.networkUser role on the host project or on the specific subnet, which lets the principal attach instances to that subnet without managing the network. Host-level network administration is not required.

  62. Question 62Planning and configuring a cloud solution

    An application writes hundreds of thousands of time-series events per second and needs low-latency lookups by a single row key. Which data store should be planned?

    1. BigQuery with streaming inserts
    2. Cloud Storage with a JSON API
    3. Cloud Bigtable Correct answer.
    4. Cloud SQL with read replicas

    Explanation

    Bigtable is a wide-column store built for high write throughput and single-key reads at low latency, which matches time-series ingest patterns. Relational and analytical stores are not designed for this access profile at that volume.

  63. Question 63Planning and configuring a cloud solution

    A steady-state workload runs continuously for three years on Compute Engine with predictable capacity. Which purchasing choice minimizes cost?

    1. Preemptible VMs with autoscaling
    2. Spot VMs in a managed instance group
    3. Three-year committed use discounts Correct answer.
    4. Sustained use discounts alone

    Explanation

    Committed use discounts exchange a one- or three-year commitment for a large discount on the committed vCPU and memory, and the three-year term yields the deepest discount for predictable baseline load. Interruptible options are unsuitable for a workload that must not be reclaimed.

  64. Question 64Planning and configuring a cloud solution

    A hub VPC must forward traffic between two spoke VPCs that are each peered only with the hub. What must the design include?

    1. Shared VPC removes the need for any routing
    2. Routing through a hub VPC with an HA VPN or NCC hub Correct answer.
    3. VPC peering forwards traffic transitively by default
    4. Cloud NAT forwards traffic between the two VPCs

    Explanation

    VPC peering is non-transitive, so a spoke cannot reach another spoke through the hub by peering alone. Transit requires a routing-capable hub, such as Network Connectivity Center hub-and-spoke or VPN tunnels terminating in the hub VPC.

  65. Question 65Planning and configuring a cloud solution

    You must prevent data in a BigQuery dataset from being copied to projects outside an approved set, even by users with valid IAM permissions. Which control should be planned?

    1. VPC Service Controls Correct answer.
    2. Cloud Armor
    3. Private Google Access
    4. VPC firewall rules

    Explanation

    VPC Service Controls define a service perimeter that restricts API access to protected services such as BigQuery, blocking data movement to resources outside the perimeter regardless of IAM grants. It addresses exfiltration risk that IAM alone cannot.

  66. Question 66Planning and configuring a cloud solution

    A company needs every project owned by the finance department to inherit the same organization policies, and the projects must stay grouped for reporting. Where should the projects be placed?

    1. Create a folder for each department and place the projects inside it. Correct answer.
    2. Create a separate organization node for each department and move the projects under it.
    3. Grant each department a project-level role and let IAM inheritance handle the rest.
    4. Apply the same labels to every project and rely on label-based policy inheritance.

    Explanation

    Folders are the resource-hierarchy node used to group projects, and policies as well as IAM bindings attached to a folder are inherited by everything beneath it. Labels are metadata only and do not carry policy, and an organization has exactly one root node.

  67. Question 67Planning and configuring a cloud solution

    An analytics workload keeps its entire working set in RAM and needs the highest memory-to-vCPU ratio available on a single Compute Engine instance. Which machine family fits best?

    1. Memory-optimized M-series machine types. Correct answer.
    2. Accelerator-optimized A-series machine types.
    3. Compute-optimized C-series machine types.
    4. General-purpose E2 instances with a custom vCPU count.

    Explanation

    Machine families are organized by the ratio of memory to vCPU they offer: general purpose is balanced, compute optimized favors CPU, and memory optimized provides the largest memory per vCPU for in-memory workloads. Accelerator-optimized types exist to attach GPUs, not to maximize RAM.

  68. Question 68Planning and configuring a cloud solution

    Objects must remain readable during the loss of an entire region, and users are spread across the United States. The team does not want to select or maintain a specific pair of regions. Which bucket location should be planned?

    1. A single region with Object Versioning turned on.
    2. A dual-region composed of two regions the team selects.
    3. The US multi-region location. Correct answer.
    4. A region chosen for the lowest storage price.

    Explanation

    A multi-region location stores copies of objects in several regions of a geography automatically, which suits users spread across one country and removes the need to choose regions. Object Versioning keeps prior object generations but does not replicate data out of a single region.

  69. Question 69Planning and configuring a cloud solution

    A production Cloud SQL for MySQL instance must continue accepting writes if its primary zone becomes unavailable. What should be configured when the instance is planned?

    1. Point-in-time recovery with binary logging enabled for the instance.
    2. Read replicas in three regions promoted manually after a failure.
    3. Regional high availability with a standby in another zone. Correct answer.
    4. Automated daily backups retained for thirty days in a Cloud Storage bucket.

    Explanation

    Regional high availability keeps a synchronous standby in a second zone and fails over automatically, so writes resume without manual promotion. Backups and binary logs support recovery to a point in time but require a restore that takes much longer and loses recent writes.

  70. Question 70Planning and configuring a cloud solution

    You are planning the subnet that a GKE Standard cluster will use with VPC-native networking. What must the subnet provide?

    1. A secondary range reserved for the cluster control plane.
    2. Two secondary IP ranges for Pods and Services. Correct answer.
    3. A single primary range sized for the maximum number of Pods.
    4. One secondary range that is shared by Pods and Services.

    Explanation

    VPC-native clusters alias Pod and Service IPs from secondary ranges on the node subnet, and Pods and Services each need their own range so the two address spaces never collide. The control plane is managed by Google and is not addressed from a customer subnet range.

  71. Question 71Planning and configuring a cloud solution

    An ingress rule allows TCP 22 from 0.0.0.0/0 at priority 1000. You must block SSH from one specific /24 while still allowing SSH from every other source. What should you do?

    1. Delete the allow rule and add allow rules for every other source network.
    2. Add a deny rule at priority 2000 for that source range.
    3. Add a deny rule at priority 900 for that source range. Correct answer.
    4. Add a deny rule with the same priority and a higher target tag.

    Explanation

    VPC firewall rules are evaluated in priority order and the first matching rule decides the outcome, so the deny must carry a lower number than the existing allow rule. A deny at priority 2000 would be evaluated after the allow and never take effect.

  72. Question 72Planning and configuring a cloud solution

    A team must read Cloud Storage object metadata and restart a defined set of Compute Engine instances, and no predefined role matches exactly those permissions. What is the recommended IAM configuration?

    1. Grant the predefined Storage Object Viewer and Compute Instance Admin roles to the team.
    2. Create a custom role containing only the required permissions and grant it. Correct answer.
    3. Create a service account per team member and share its key with them.
    4. Grant the basic Editor role to the team at the project level.

    Explanation

    A custom role lets you grant exactly the permissions the task needs when no predefined role is a close match, which keeps the grant at least privilege. Basic roles and broad predefined roles would grant far more access than the task requires.

  73. Question 73Planning and configuring a cloud solution

    A CI/CD platform hosted with another cloud provider must call Google Cloud APIs as a specific service account, and the security team forbids long-lived credentials. What should be planned?

    1. Workload Identity Federation with the CI provider. Correct answer.
    2. OS Login with two-factor authentication enabled for every CI runner identity.
    3. A service account JSON key encrypted and stored in the CI system's secret manager.
    4. An organization policy constraint that disables service account key creation for the project.

    Explanation

    Workload Identity Federation lets an external identity provider vouch for the CI workload, which then obtains short-lived Google Cloud credentials by impersonating a service account. No downloadable key is ever created, satisfying the prohibition on long-lived credentials.

  74. Question 74Planning and configuring a cloud solution

    A batch tier in us-central1 streams results over the internet to client applications in Asia. Latency is unimportant but internet egress charges are a major cost. Which network service tier should the VPC use?

    1. Standard Tier, which costs less for that egress path. Correct answer.
    2. Either tier, since Google Cloud traffic never uses the public internet.
    3. Premium Tier, which routes traffic on Google's backbone.
    4. Premium Tier, which is required for Cloud Storage access.

    Explanation

    Standard Tier hands traffic to the internet as close to the destination region as possible and is priced lower for internet egress, which suits cost-sensitive, latency-tolerant workloads. Premium Tier routes traffic longer on Google's network and costs more for the same egress.

  75. Question 75Planning and configuring a cloud solution

    Instances in two separate VPC networks must resolve db.internal to private addresses, and the name must never be resolvable from the public internet. What should be configured?

    1. A public Cloud DNS zone named db.internal with DNSSEC enabled.
    2. A Cloud DNS forwarding zone that points both VPCs to 8.8.8.8.
    3. Custom entries added to each instance's /etc/hosts file.
    4. A private Cloud DNS zone authorized for both VPC networks. Correct answer.

    Explanation

    A private managed zone bound to specific VPC networks answers only queries from those networks and is never published to the internet. A public zone for the same name would expose the records globally, which the requirement forbids.

  76. Question 76Planning and configuring a cloud solution

    A stateless web tier must add and remove identical VMs automatically based on CPU utilization across three zones simultaneously. Which configuration meets the requirement?

    1. Unmanaged instance group with a scheduled start and stop script.
    2. A regional managed instance group with autoscaling. Correct answer.
    3. A single large VM with a custom machine type and a startup script.
    4. A zonal managed instance group with a fixed size of three instances.

    Explanation

    A regional managed instance group spreads identical instances built from an instance template across zones in a region and supports autoscaling policies such as CPU utilization. A fixed-size group or an unmanaged group cannot scale automatically.

  77. Question 77Planning and configuring a cloud solution

    A deployment into a new region fails because the project has no GPU quota there. The workload is already approved and funded. What should the plan include before deploying?

    1. Request a GPU quota increase in the target region. Correct answer.
    2. Enable the Compute Engine API again in the new region.
    3. Create the VMs with a smaller machine type and retry the deployment.
    4. Move the project into a folder that has no quota restrictions.

    Explanation

    Quotas are enforced per project and per region, and GPU quota commonly starts at zero, so a quota increase must be requested and approved for the target region before resources can be created. Folders do not carry or relax quota limits.

  78. Question 78Planning and configuring a cloud solution

    Security requires customer-managed encryption keys for BigQuery datasets and Cloud Storage buckets across many projects, with a central team owning key rotation. Which design should be planned?

    1. A central KMS project whose keys other projects use through IAM grants. Correct answer.
    2. Google-managed default encryption with Cloud External Key Manager enabled everywhere.
    3. An individual key ring created inside each project and rotated by that project's owners.
    4. A single symmetric key exported and stored in each project's Secret Manager.

    Explanation

    Cloud KMS key rings and keys can live in a dedicated project while resource service agents in other projects are granted the encrypt/decrypt role on those keys, which centralizes control and rotation. Per-project key rings would distribute ownership and defeat the central-team requirement.

  79. Question 79Planning and configuring a cloud solution

    Planning for a Cloud VPN connection to on-premises reveals several /16 ranges that exist in both the VPC and the on-premises network. What is the correct planning decision?

    1. Attach the on-premises network as an additional subnet of the VPC.
    2. Rely on Cloud VPN to translate addresses for the overlapping ranges.
    3. Renumber the overlapping subnets so no CIDR block exists in both. Correct answer.
    4. Use a Shared VPC host project, which removes the overlap automatically.

    Explanation

    VPN and Interconnect routing requires unique IP ranges on each side, because overlapping ranges create ambiguous routes that cannot be resolved by the tunnel. Renumbering one side before the connection is built is the standard remedy.

  80. Question 80Planning and configuring a cloud solution

    Regulators require that records be kept unaltered for seven years, and storage cost should drop sharply after the first year of active access. Which storage plan fits?

    1. Lifecycle to Archive storage plus Bucket Lock retention. Correct answer.
    2. Coldline storage with Object Versioning enabled and a customer-supplied key.
    3. Nearline storage with uniform bucket-level access and signed URLs.
    4. Standard storage with Object Versioning and a 90-day lifecycle rule.

    Explanation

    A lifecycle rule can transition objects from a warmer class to Archive after the first year, minimizing cost for rarely read data, while a Bucket Lock retention policy enforces that objects cannot be deleted or replaced before the retention period ends. Versioning alone does not prevent deletion of every generation.

  81. Question 81Deploying and implementing a cloud solution

    You are creating a Compute Engine VM named web-1 in zone us-central1-a from the command line. Which gcloud command performs this action?

    1. gcloud compute vms create web-1 --zone=us-central1-a
    2. gcloud compute instances new web-1 --zone=us-central1-a
    3. gcloud compute instances describe web-1 --zone=us-central1-a
    4. gcloud compute instances create web-1 --zone us-central1-a Correct answer.

    Explanation

    VM instances are managed under the `gcloud compute instances` command group, and `create` is the verb that provisions a new instance. The `--zone` flag places the instance in the named zone. `describe` only reads back properties of an existing instance.

  82. Question 82Deploying and implementing a cloud solution

    A deployment script fails while creating a Cloud Storage bucket named `prod-data` and the API returns HTTP 409. What is the most likely cause?

    1. The bucket name is already taken by another project. Correct answer.
    2. The project has reached the maximum number of buckets allowed.
    3. Bucket names must start with the owning project ID.
    4. The requested storage location is not enabled for the project.

    Explanation

    Bucket names share a single global namespace across all of Cloud Storage, so a name that already exists anywhere cannot be reused and the API returns a conflict. Prefixing names with the project ID is a convention that reduces collisions but is not enforced.

  83. Question 83Deploying and implementing a cloud solution

    Which metadata key lets you pass a startup script inline as a string when creating a Compute Engine instance?

    1. user-data
    2. startup-script Correct answer.
    3. init-script
    4. startup-script-url

    Explanation

    The Linux guest agent reads the `startup-script` metadata key and executes its value as a script at boot. The companion key `startup-script-url` fetches the script from a Cloud Storage object instead of accepting it inline.

  84. Question 84Deploying and implementing a cloud solution

    Which gcloud command deploys a container image as a new revision of a Cloud Run service?

    1. gcloud app deploy
    2. gcloud run deploy Correct answer.
    3. gcloud builds submit
    4. gcloud functions deploy

    Explanation

    `gcloud run deploy` creates or updates a Cloud Run service from a container image (or from source) and rolls out a new revision. The other commands create different resource types or run a build without deploying a serving resource.

  85. Question 85Deploying and implementing a cloud solution

    An application running on Compute Engine must connect to a Cloud SQL instance using IAM authentication over an encrypted channel, without distributing SSL certificates or database passwords to the VM. Which component provides this?

    1. A Cloud SQL read replica
    2. Cloud SQL Auth Proxy Correct answer.
    3. A downloaded service account key file
    4. A Cloud VPN tunnel to the instance

    Explanation

    The Cloud SQL Auth Proxy handles IAM-based authorization and TLS encryption on behalf of the client, so the application connects to a local socket or port. It relies on the identity of the environment it runs in and removes the need for long-lived credentials on disk.

  86. Question 86Deploying and implementing a cloud solution

    Which flag on `gcloud container clusters create` creates a GKE cluster whose control plane and default node pool are spread across multiple zones in us-central1?

    1. --zone=us-central1-a
    2. --node-locations=us-central1-a
    3. --locations=us-central1-a
    4. --region=us-central1 Correct answer.

    Explanation

    Passing `--region` creates a regional cluster, which replicates the control plane across zones of that region and spreads the default node pool. `--zone` creates a single-zone cluster whose control plane lives in one zone.

  87. Question 87Deploying and implementing a cloud solution

    You are writing a first-generation Cloud Functions function that runs each time a new object is written to a Cloud Storage bucket. Which event type should the function subscribe to?

    1. providers/cloud.pubsub/eventTypes/topic.publish
    2. providers/cloud.firestore/eventTypes/document.create
    3. google.storage.object.delete
    4. google.storage.object.finalize Correct answer.

    Explanation

    The `google.storage.object.finalize` event fires when an object has been written and is immutable, which is the correct trigger for reacting to new uploads. The `delete` event corresponds to object removal.

  88. Question 88Deploying and implementing a cloud solution

    You need to create many identical VMs from the same machine type, boot image, and startup script, and keep them in sync over time. Which resource encapsulates that reusable definition?

    1. A persistent disk snapshot
    2. A custom image
    3. An instance template Correct answer.
    4. An instance schedule

    Explanation

    An instance template stores machine type, image, disk, network, and metadata settings so identical VMs can be created repeatedly, including by a managed instance group. A custom image captures only the boot disk contents.

  89. Question 89Deploying and implementing a cloud solution

    A managed instance group must never drop below 3 instances or exceed 10 instances, and it should scale based on average CPU utilization. Which configuration meets these requirements?

    1. Set a health check that restarts instances below 10 percent CPU
    2. Set the autoscaler target CPU utilization to 30 percent
    3. Set the autoscaler min replicas to 3 and max replicas to 10 Correct answer.
    4. Set the instance template to create three instances per zone

    Explanation

    The autoscaler's minimum and maximum replica settings are hard bounds on group size, and the CPU utilization target decides when the group grows or shrinks within those bounds. A template describes how each instance is built, not how many exist.

  90. Question 90Deploying and implementing a cloud solution

    You need to roll out a new instance template to an existing managed instance group with no downtime, replacing instances gradually while keeping the group available. Which approach should you use?

    1. SSH into each instance and replace the application binary manually
    2. Delete the managed instance group and recreate it from the new instance template
    3. Detach every instance from the group and reattach them one by one
    4. Run rolling-action start-update with type proactive on the group Correct answer.

    Explanation

    The rolling-action start-update command performs a controlled, gradual replacement of instances using the new template, honoring surge and unavailability settings so capacity is maintained. Deleting the group removes all capacity at once.

  91. Question 91Deploying and implementing a cloud solution

    You need one global anycast IP address that terminates SSL and sends users to the nearest healthy backend across multiple regions. Which load balancer provides this?

    1. Internal TCP/UDP load balancer
    2. Regional external proxy network load balancer
    3. Regional external passthrough network load balancer
    4. Global external Application Load Balancer Correct answer.

    Explanation

    The global external Application Load Balancer gives a single global anycast VIP, terminates TLS at the edge, and routes to the closest healthy backend based on the URL map and backend health. Passthrough and internal load balancers do not provide a global HTTPS frontend.

  92. Question 92Deploying and implementing a cloud solution

    You want a managed instance group to automatically recreate any VM whose application stops responding to an HTTP check. What must you configure?

    1. An uptime check in Cloud Monitoring attached to the group
    2. A startup script that restarts the application on failure
    3. An autoscaling policy based on CPU utilization
    4. An autohealing policy that references a health check Correct answer.

    Explanation

    An autohealing policy points at a health check, and the group recreates instances that fail it, which repairs application-level failures rather than just boot problems. Uptime checks can alert you but cannot act on the group.

  93. Question 93Deploying and implementing a cloud solution

    A bucket's objects must be readable by every authenticated user in the project, and permissions must be managed only at the bucket level so object ACLs cannot override them. Which setting enforces this?

    1. Uniform bucket-level access Correct answer.
    2. A Bucket Lock retention policy
    3. Object versioning
    4. Fine-grained ACLs with a public read grant

    Explanation

    Uniform bucket-level access disables object ACLs and evaluates access solely through IAM policies on the bucket and project, which satisfies the requirement. Versioning and retention policies affect object lifecycle, not permission evaluation.

  94. Question 94Deploying and implementing a cloud solution

    An application using Cloud SQL for MySQL must keep serving writes if the instance's zone fails, with failover happening without manual intervention. What should you configure?

    1. A read replica in a different zone
    2. Automated backups with point-in-time recovery
    3. High availability with a regional Cloud SQL instance Correct answer.
    4. A larger machine tier with more vCPUs and memory

    Explanation

    A regional (high availability) Cloud SQL instance keeps a standby in a second zone and promotes it automatically when the primary zone fails. Read replicas and backups do not provide automatic promotion of the primary.

  95. Question 95Deploying and implementing a cloud solution

    A Cloud Run service experiences cold starts during traffic bursts. You want at least two instances kept warm at all times. Which deploy flag achieves this?

    1. --max-instances=2
    2. --cpu-throttling=false
    3. --concurrency=2
    4. --min-instances=2 Correct answer.

    Explanation

    `--min-instances` keeps the specified number of instances running and ready even with no traffic, which removes cold starts for that baseline. `--max-instances` only caps how far the service can scale out.

  96. Question 96Deploying and implementing a cloud solution

    Your team wants Google to manage node provisioning, scaling, and patching, with billing based on the pods that actually run rather than on provisioned nodes. Which cluster type should you choose?

    1. GKE Standard with the cluster autoscaler enabled
    2. GKE Autopilot Correct answer.
    3. GKE Standard with node auto-provisioning enabled
    4. A managed instance group of VMs running kubeadm

    Explanation

    GKE Autopilot provisions and scales nodes for you and bills per pod resource request, removing node-level operations from the team. Standard clusters still expose nodes that you size, patch, and pay for.

  97. Question 97Deploying and implementing a cloud solution

    A Cloud Build trigger uses a dedicated service account to push container images to an Artifact Registry repository in the same project. Which role grants only the permissions needed to push?

    1. roles/artifactregistry.reader
    2. roles/artifactregistry.writer Correct answer.
    3. roles/storage.admin
    4. roles/cloudbuild.builds.editor

    Explanation

    The Artifact Registry Writer role allows uploading and deleting artifacts in a repository, which covers pushing images. The Reader role allows only downloads, so a push would be rejected.

  98. Question 98Deploying and implementing a cloud solution

    A Compute Engine VM must read objects from a Cloud Storage bucket, and no credential files may be stored on the VM. What should you do?

    1. Grant the allUsers principal the Storage Object Viewer role
    2. Enable uniform bucket-level access on the bucket
    3. Download a service account key and set GOOGLE_APPLICATION_CREDENTIALS
    4. Attach a user-managed service account with the Storage Object Viewer role Correct answer.

    Explanation

    Instances obtain identity from the service account attached to them and receive short-lived tokens from the metadata server, so no key file is needed. The IAM role must then grant the required object read permission.

  99. Question 99Deploying and implementing a cloud solution

    Instances carrying the network tag `web` must accept HTTPS from any internet client but accept SSH only from 10.0.0.0/8. Which pair of VPC ingress firewall rules implements this?

    1. One rule allowing tcp:443 from 0.0.0.0/0 and one allowing tcp:22 from 10.0.0.0/8 Correct answer.
    2. One rule allowing tcp:22 from 0.0.0.0/0 and one rule allowing tcp:443 from 10.0.0.0/8
    3. One rule allowing all protocols from 0.0.0.0/0 with the target tag web
    4. One rule allowing tcp:443 from 10.0.0.0/8 with no target tags applied

    Explanation

    Firewall rules match on protocol, port, source range, and target tags, so separate rules are needed for the two ports with their distinct sources. Using the `web` target tag limits both rules to the intended instances.

  100. Question 100Deploying and implementing a cloud solution

    A Cloud Run service needs a database password at runtime, and the value must not appear in the container image or in plain text in the service configuration. What should you do?

    1. Store the password in Secret Manager and reference it with --set-secrets Correct answer.
    2. Bake the password into the image during the Cloud Build step
    3. Pass the password as a literal value to --set-env-vars at deploy time
    4. Declare the password with an ENV instruction in the Dockerfile

    Explanation

    Referencing a Secret Manager secret with `--set-secrets` injects the value at runtime without persisting it in the image or the revision's plain configuration. The service account running the revision needs permission to access the secret version.

  101. Question 101Deploying and implementing a cloud solution

    Within a single managed instance group, you want a small fraction of the instances to run a new instance template before the full rollout happens. Which feature should you use?

    1. Create a second managed instance group and point the load balancer at both
    2. Run rolling-action start-update with a canary version set to 10 percent Correct answer.
    3. Reduce the autoscaler maximum replicas to one tenth of the current size
    4. Enable scale-in controls and let the autoscaler replace slow instances

    Explanation

    The rolling update command accepts a canary version, which applies the new template to a specified percentage of instances while the rest keep the current template. The canary can then be promoted to a full rollout once validated.

  102. Question 102Deploying and implementing a cloud solution

    A GKE node pool runs expensive GPU nodes that only a few workloads should use. Which mechanism keeps all other pods off those nodes?

    1. Add a node taint to the pool and a matching toleration to the pods Correct answer.
    2. Add a pod disruption budget to the namespace running the pods
    3. Set the node pool's autoscaling minimum to one and maximum to three
    4. Add a node label to the pool and a matching node selector to every pod

    Explanation

    A taint on the node repels pods that do not declare the matching toleration, so only the intended workloads can be scheduled there. Labels and selectors only attract pods that opt in and do not exclude others.

  103. Question 103Deploying and implementing a cloud solution

    A Shared VPC host project provides subnets to several service projects. Where should a team create a new Compute Engine instance, and which subnet should it use?

    1. In the host project, using a subnet from the service project
    2. In the service project, using a subnet shared from the host project Correct answer.
    3. In the service project, using a subnet created inside the service project
    4. In the host project, using an automatically created default subnet

    Explanation

    In Shared VPC the network and subnets live in the host project, while workloads run in service projects and consume those shared subnets. Creating a separate subnet in the service project puts the instance on a different network entirely.

  104. Question 104Deploying and implementing a cloud solution

    A service on GKE sits behind a global external Application Load Balancer. You want exactly 5 percent of production traffic to reach a new version so it can be validated before cutover. What should you implement?

    1. Expose the new version with its own external Service and use DNS round robin between them
    2. Add a Cloud Armor rule that samples five percent of incoming requests
    3. Register both versions as NEG backends of one backend service with weighted traffic split Correct answer.
    4. Use a canary annotation on the Deployment so kube-proxy splits traffic per node

    Explanation

    The load balancer splits traffic at the backend service level, where multiple network endpoint groups can be assigned weights that determine the share of requests each receives. Traffic splitting happens in the proxy, not in the cluster's Service layer.

  105. Question 105Deploying and implementing a cloud solution

    Cloud Build triggers currently deploy to production using a broadly privileged default service account. Which change best limits the impact of a compromised build?

    1. Enable the default Compute Engine service account for all builds in the project
    2. Attach a dedicated user-managed service account with minimal roles to the trigger Correct answer.
    3. Grant the Cloud Build service account the Owner role at the project level
    4. Store a downloaded JSON key for the default service account in a Cloud Storage bucket

    Explanation

    Each trigger can run as a user-specified service account, so giving it only the roles the pipeline needs confines what a compromised build can reach. Broad predefined roles such as Owner defeat least privilege because they span the whole project.

  106. Question 106Deploying and implementing a cloud solution

    You need to create a Cloud Storage bucket named acme-logs whose data is stored in the europe-west4 region. Which command accomplishes this?

    1. gcloud storage buckets create acme-logs --location=europe-west4
    2. gcloud storage buckets create gs://acme-logs --location=europe-west4 Correct answer.
    3. gcloud storage buckets create gs://acme-logs --region=europe-west4
    4. gcloud storage buckets update gs://acme-logs --location=europe-west4

    Explanation

    Bucket names are passed as gs:// URIs, and the physical location is set at creation time with the --location flag. The bucket location is fixed once the bucket exists, so it must be supplied on the create command.

  107. Question 107Deploying and implementing a cloud solution

    An existing Pub/Sub topic named orders must feed a new pull subscription named orders-sub, where each message is delivered to a single consumer. Which command creates the required resource?

    1. gcloud pubsub topics create orders-sub --topic=orders
    2. gcloud pubsub subscriptions pull orders-sub --topic=orders
    3. gcloud pubsub subscriptions create orders --topic=orders-sub
    4. gcloud pubsub subscriptions create orders-sub --topic=orders Correct answer.

    Explanation

    Subscriptions are created against an already existing topic using gcloud pubsub subscriptions create, with the topic named by --topic. A pull subscription holds messages until a subscriber acknowledges them, so each message goes to one consumer at a time.

  108. Question 108Deploying and implementing a cloud solution

    A container image has already been pushed to us-central1-docker.pkg.dev/acme/app/api:1.4. You must run it as a new Cloud Run service named api in us-central1. Which command does this?

    1. gcloud run jobs deploy api --image=us-central1-docker.pkg.dev/acme/app/api:1.4 --region=us-central1
    2. gcloud run deploy api --image=us-central1-docker.pkg.dev/acme/app/api:1.4 --region=us-central1 Correct answer.
    3. gcloud run deploy api --source=. --region=us-central1
    4. gcloud run services create api --image=us-central1-docker.pkg.dev/acme/app/api:1.4 --region=us-central1

    Explanation

    gcloud run deploy creates the service if it does not exist and points it at a container image supplied with --image. Supplying --region on the command line avoids the interactive prompt that asks which platform and region to use.

  109. Question 109Deploying and implementing a cloud solution

    The file app.yaml in the current directory defines an App Engine service named web. Which command releases it to App Engine?

    1. gcloud app create app.yaml
    2. gcloud app services deploy web --config app.yaml
    3. gcloud builds submit app.yaml
    4. gcloud app deploy app.yaml Correct answer.

    Explanation

    gcloud app deploy reads the app.yaml descriptor, builds the runtime, and creates a new version of the service named in the file. The service name comes from the descriptor, not from a command-line argument.

  110. Question 110Deploying and implementing a cloud solution

    Creating a VPC-native GKE Standard cluster in an existing custom-mode subnet fails with an error about missing secondary ranges. Which change makes cluster creation succeed?

    1. Attach a Cloud NAT gateway to the subnet before creating the cluster.
    2. Add one secondary IP range to the subnet and share it between Pods and Services.
    3. Disable alias IPs on the node pool so pods use the node's primary address.
    4. Add two secondary IP ranges to the subnet, one for Pods and one for Services. Correct answer.

    Explanation

    VPC-native clusters give every pod and service an alias IP drawn from secondary ranges on the subnet, and the two purposes require separate ranges. Without both ranges the cluster cannot allocate the pod and service CIDRs it needs.

  111. Question 111Deploying and implementing a cloud solution

    A pod on a private GKE cluster must reach a Cloud SQL for PostgreSQL instance that has only a private IP in the same VPC. The nodes have no external IP addresses. What is the supported way to connect?

    1. Run the Cloud SQL Auth Proxy as a sidecar container and connect the app to 127.0.0.1. Correct answer.
    2. Create an internal passthrough load balancer whose backend is the instance's private IP.
    3. Give the instance a public IP and authorize the cluster's node CIDR range.
    4. Copy the instance's server certificate into the pod and connect to its private IP on port 5432.

    Explanation

    The Cloud SQL Auth Proxy handles IAM-based authorization and TLS for you, and as a sidecar it is reachable by the application over the pod's loopback address. This works without giving the instance or the nodes public connectivity.

  112. Question 112Deploying and implementing a cloud solution

    You are deploying a 2nd gen Cloud Function that must execute whenever a message is published to the existing Pub/Sub topic orders. Which deploy flag expresses this trigger?

    1. --trigger-http
    2. --trigger-topic=orders Correct answer.
    3. --trigger-bucket=orders
    4. --trigger-resource=orders

    Explanation

    --trigger-topic creates the Pub/Sub trigger for the named topic and, on 2nd gen functions, provisions the backing Eventarc trigger automatically. The function then receives the message payload as its event argument.

  113. Question 113Deploying and implementing a cloud solution

    A Docker repository team/app exists in Artifact Registry in us-east1. Before running docker push to us-east1-docker.pkg.dev, which command configures the local Docker client to authenticate with that registry?

    1. docker login -u oauth2accesstoken us-east1-docker.pkg.dev
    2. gcloud auth configure-docker us-east1-docker.pkg.dev Correct answer.
    3. gcloud artifacts repositories describe team/app --location=us-east1
    4. gcloud auth application-default login

    Explanation

    gcloud auth configure-docker writes a credential helper entry for the named host into the Docker configuration, so later docker push and pull calls use your gcloud credentials. The registry hostname must match the Artifact Registry location prefix.

  114. Question 114Deploying and implementing a cloud solution

    A Cloud Run service must be callable only by a specific service account in another project. Unauthenticated users must never reach it. Which approach meets the requirement?

    1. Deploy with --no-allow-unauthenticated and grant roles/run.developer to the calling service account.
    2. Deploy with --no-allow-unauthenticated and grant roles/run.invoker on the service to that service account. Correct answer.
    3. Deploy with --allow-unauthenticated and rely on a VPC firewall rule to block other callers.
    4. Deploy with --ingress=internal and grant roles/viewer to the calling service account.

    Explanation

    Request authentication on Cloud Run is controlled by the IAM invoker binding plus the unauthenticated-access setting. roles/run.invoker lets the caller invoke the service, and --no-allow-unauthenticated rejects callers that present no identity token.

  115. Question 115Deploying and implementing a cloud solution

    A BigQuery table analytics.events with a DATE column named event_date must be partitioned on that column, and every query against it must include a partition filter. Which command creates it that way?

    1. bq load --autodetect --partition_by=event_date analytics.events gs://acme/events/*.json
    2. bq mk --table --clustered_fields=event_date analytics.events schema.json
    3. bq mk --table --time_partitioning_field=event_date --require_partition_filter=true analytics.events schema.json Correct answer.
    4. bq mk --table --time_partitioning_type=DAY analytics.events schema.json

    Explanation

    --time_partitioning_field partitions on a DATE or TIMESTAMP column in the schema, and --require_partition_filter=true makes queries that omit a filter fail. Both settings belong on the table-creation command.

  116. Question 116Deploying and implementing a cloud solution

    Backend VMs behind an external Application Load Balancer pass an HTTP health check only when the firewall allows the probes. Which source ranges must the ingress rule permit?

    1. 130.211.0.0/22 and 35.191.0.0/16 Correct answer.
    2. 0.0.0.0/0 on the backend service port
    3. The load balancer's frontend IP address as a /32
    4. 10.128.0.0/9, the default VPC subnet range

    Explanation

    Google's load balancer probes originate from the fixed ranges 130.211.0.0/22 and 35.191.0.0/16, so the backend firewall rule must allow both. Restricting to any other range leaves the health check unable to reach the instances.

  117. Question 117Deploying and implementing a cloud solution

    You need to create a VM that always uses the newest Debian 12 image without pinning a specific image name. Which command does this?

    1. gcloud compute instances create vm1 --image=debian-12
    2. gcloud compute instances create vm1 --boot-disk=debian-12-latest
    3. gcloud compute instances create vm1 --image-family=debian-12
    4. gcloud compute instances create vm1 --image-family=debian-12 --image-project=debian-cloud Correct answer.

    Explanation

    An image family resolves to the most recent non-deprecated image in that family, so the boot disk tracks new releases. The family name must be paired with the project that publishes it, here debian-cloud.

  118. Question 118Deploying and implementing a cloud solution

    A Cloud SQL for PostgreSQL primary in us-central1 is lost in a regional outage. A cross-region read replica in us-east1 is current. Which sequence restores write capability with the least data loss?

    1. Restart the us-central1 primary and wait for the region to recover.
    2. Promote the us-east1 replica to a standalone primary, then repoint the application at it. Correct answer.
    3. Change the replica's preferred zone to us-central1 so it becomes the primary.
    4. Restore the replica's latest backup into a new instance in us-central1.

    Explanation

    A cross-region replica can be promoted, which stops replication and turns it into a read/write primary with the data it has already applied. Promoting keeps the most recent replicated state, so it loses less data than restoring a backup.

  119. Question 119Deploying and implementing a cloud solution

    A three-node distributed database on GKE requires each pod to keep the same DNS name and its own persistent disk across reschedules. Which configuration provides this?

    1. A DaemonSet with hostPath volumes and a NodePort Service
    2. A Deployment with pod anti-affinity and one ReadWriteMany PersistentVolume mounted by every pod
    3. A StatefulSet with a headless Service and volumeClaimTemplates Correct answer.
    4. A Deployment with three replicas, a ClusterIP Service, and emptyDir volumes

    Explanation

    StatefulSets assign each pod a stable ordinal identity, which combined with a headless Service yields a per-pod DNS name that survives restarts. volumeClaimTemplates gives each ordinal its own PersistentVolumeClaim, so storage follows the identity.

  120. Question 120Deploying and implementing a cloud solution

    A Deployment runs 10 replicas and must keep at least 9 pods available at all times while rolling out a new image, halting if new pods fail their readiness probe. Which configuration achieves this?

    1. Set maxUnavailable to 1, maxSurge to 0, and add a readiness probe to the container. Correct answer.
    2. Set maxSurge to 5 and maxUnavailable to 5 for a faster rollout.
    3. Keep the default rollout parameters and set imagePullPolicy to Always.
    4. Set the strategy type to Recreate so all pods are replaced in one step.

    Explanation

    maxUnavailable caps how many pods may be down during the rollout, so a value of 1 keeps at least 9 of 10 available. The readiness probe keeps new pods out of the available count until they actually serve traffic, which stalls the rollout when they fail.

  121. Question 121Ensuring successful operation of a cloud solution

    You need Cloud Monitoring to probe a public HTTPS endpoint from multiple Google regions every minute and notify you when it stops returning HTTP 200. Which feature should you configure?

    1. A Cloud Trace span report
    2. A CPU utilization alerting policy
    3. A log-based metric
    4. An uptime check Correct answer.

    Explanation

    Uptime checks issue synthetic requests to a URL or resource from several geographic locations at a configurable frequency and publish success, latency, and status code metrics that alerting policies can watch. Log-based metrics and trace reports only describe activity that has already happened.

  122. Question 122Ensuring successful operation of a cloud solution

    Cloud Logging stores log entries in the _Default bucket unless you configure otherwise. What is the default retention period of that bucket?

    1. 30 days Correct answer.
    2. 7 days
    3. 365 days
    4. 3650 days

    Explanation

    The _Default log bucket retains entries for 30 days out of the box, and you can raise that retention period on the bucket itself if you need longer storage. User-defined buckets can be created with their own retention settings.

  123. Question 123Ensuring successful operation of a cloud solution

    A microservice throws exceptions in production. Which service groups these exceptions into distinct issues, counts their occurrences, and shows representative stack traces?

    1. Cloud Profiler
    2. Cloud Audit Logs
    3. Cloud Trace
    4. Error Reporting Correct answer.

    Explanation

    Error Reporting ingests application exceptions and groups them into issues, tracking occurrence counts and showing stack traces so you can prioritize fixes. It can also open incidents and send notifications when a new issue appears.

  124. Question 124Ensuring successful operation of a cloud solution

    You configure a Cloud Billing budget of $1,000 with an alert threshold at 100% of the amount. What is the effect when actual spend reaches $1,000?

    1. Compute Engine instances in the project are stopped automatically.
    2. The project's billing account is closed automatically.
    3. Recipients receive an alert and resources keep running. Correct answer.
    4. A hard quota is applied to all billable APIs.

    Explanation

    Billing budgets are monitoring constructs: crossing a threshold sends notifications to the configured users and, optionally, a Pub/Sub message. They never stop, throttle, or disable resources on their own; automation must be built separately.

  125. Question 125Ensuring successful operation of a cloud solution

    Which log stream records administrative operations such as creating a Compute Engine instance, and is always enabled for every project?

    1. Data Access audit logs
    2. Admin Activity audit logs Correct answer.
    3. VPC Flow Logs
    4. Firewall Rules Logging

    Explanation

    Admin Activity audit logs capture configuration and metadata writes, are enabled by default, and cannot be disabled or excluded by log sinks. Data Access audit logs record reads and writes of user data and must be enabled explicitly.

  126. Question 126Ensuring successful operation of a cloud solution

    You want to see how CPU utilization of one Compute Engine instance varied over the past six hours. Which action is the most direct?

    1. Read the serial port output of the instance.
    2. Enable VPC Flow Logs on the subnet.
    3. Run gcloud compute instances describe to read the metric history.
    4. View the instance CPU metric in Metrics Explorer. Correct answer.

    Explanation

    Metrics Explorer charts any time series collected by Cloud Monitoring, including the built-in agentless CPU metric for Compute Engine instances, over a selectable time window. The other options expose boot messages, network flows, or configuration metadata rather than historical metrics.

  127. Question 127Ensuring successful operation of a cloud solution

    Which Google Cloud tool continuously samples a running production application to show which functions consume the most CPU time and heap allocation?

    1. Cloud Profiler Correct answer.
    2. Cloud Debugger
    3. Error Reporting
    4. Cloud Trace

    Explanation

    Cloud Profiler collects statistical samples of CPU and heap usage from live applications with low overhead and attributes them to functions, producing flame graphs of hot code paths. It answers 'where in the code is time or memory going' rather than 'which request was slow' or 'what failed'.

  128. Question 128Ensuring successful operation of a cloud solution

    Requests to your application cross four microservices and are sometimes slow. Which service records per-request timing across those services so you can identify the slowest hop?

    1. Error Reporting
    2. Cloud Logging
    3. Cloud Trace Correct answer.
    4. Cloud Monitoring uptime checks

    Explanation

    Cloud Trace collects latency spans for a request as it propagates between services and assembles them into a trace, making the slowest hop visible. Uptime checks measure only end-to-end availability from outside, and logs and error reports do not produce span-level timing.

  129. Question 129Ensuring successful operation of a cloud solution

    Your security team must keep all VPC Flow Logs for seven years and run ad hoc SQL queries over them. Which Cloud Logging sink destination fits best?

    1. A BigQuery dataset Correct answer.
    2. A Pub/Sub topic with a pull subscription
    3. A Cloud Storage bucket with an object lifecycle rule
    4. A Cloud Logging bucket in a second project

    Explanation

    A BigQuery sink writes log entries into partitioned tables that remain queryable with SQL for as long as the dataset's table expiration allows, which fits the analysis requirement. Cloud Storage is cheaper for pure archival but requires external tooling to query.

  130. Question 130Ensuring successful operation of a cloud solution

    You want an alert to fire only when CPU utilization on an instance stays above 80% for ten minutes, so short spikes do not page anyone. Which alerting policy setting achieves this?

    1. The alignment period of the metric only
    2. The condition's 'for' duration Correct answer.
    3. The alerting policy's documentation field
    4. The notification channel's rate limit

    Explanation

    The 'for' duration on a condition keeps the condition open for the specified window before an incident is opened, filtering out transient violations. The alignment period controls how samples are aggregated into points, not how long a violation must persist.

  131. Question 131Ensuring successful operation of a cloud solution

    You want a metric that counts how many log entries contain the string 'OutOfMemory' so you can chart it and alert on it. What should you create?

    1. A log exclusion
    2. An uptime check
    3. A log-based counter metric Correct answer.
    4. A custom metric written from your application

    Explanation

    A log-based metric applies a filter to the log stream and exposes the matching entry count as a Cloud Monitoring metric that can be charted and used in alerting policies, with no application changes. An exclusion filter removes entries instead of counting them.

  132. Question 132Ensuring successful operation of a cloud solution

    Your alerting policy fires correctly, but nobody is notified. You want incidents to reach an on-call rotation in PagerDuty and a Slack channel. What must be attached to the alerting policy?

    1. A monitoring dashboard
    2. A log sink
    3. An uptime check
    4. Notification channels Correct answer.

    Explanation

    An alerting policy opens incidents only through the notification channels listed on it, such as email, SMS, Slack, webhooks, or PagerDuty. Without at least one channel the policy evaluates conditions but delivers nothing.

  133. Question 133Ensuring successful operation of a cloud solution

    A developer must be able to view dashboards and inspect alerting policies in Cloud Monitoring but must not be able to create or change them. Which predefined role should you grant?

    1. roles/monitoring.editor
    2. roles/monitoring.viewer Correct answer.
    3. roles/logging.admin
    4. roles/owner

    Explanation

    The predefined Monitoring Viewer role grants read-only access to monitoring resources including dashboards, alerting policies, and metric data, while Monitoring Editor additionally permits creating and modifying them. Following least privilege, the viewer role satisfies the requirement.

  134. Question 134Ensuring successful operation of a cloud solution

    You need memory and disk-usage metrics plus application log files from a Compute Engine VM to appear in Cloud Monitoring and Cloud Logging. What should you install on the VM?

    1. An uptime check
    2. The Ops Agent Correct answer.
    3. The Cloud Profiler agent
    4. Only the legacy Cloud Logging agent

    Explanation

    The Ops Agent is the current single agent that collects both system metrics (including memory and disk, which are not available from the hypervisor) and logs and sends them to Cloud Monitoring and Cloud Logging. Legacy agents split these duties and are no longer the recommended path.

  135. Question 135Ensuring successful operation of a cloud solution

    You want an automated script to react when a project's spend crosses 90% of its budget. What should you configure?

    1. A Cloud Scheduler job that calls the Billing API
    2. A Cloud Audit Logs sink to Pub/Sub
    3. An uptime check on the billing console
    4. Budget notifications sent to a Pub/Sub topic Correct answer.

    Explanation

    Cloud Billing budgets can publish programmatic notifications to a Pub/Sub topic when thresholds are crossed, and a subscriber can then trigger Cloud Functions or other automation. Audit log sinks carry API activity records and do not carry budget events.

  136. Question 136Ensuring successful operation of a cloud solution

    You need to restore a Cloud SQL for MySQL instance to an arbitrary point within the last seven days after a bad migration. Which configuration is required?

    1. On-demand backups taken before the migration
    2. Daily automated backups alone
    3. Read replicas in a second region
    4. Automated backups plus point-in-time recovery Correct answer.

    Explanation

    Point-in-time recovery keeps binary logs alongside automated backups so the instance can be recovered to any moment inside the retention window, rather than only to the moment a backup finished. Read replicas improve availability and read scale but do not provide rewinding.

  137. Question 137Ensuring successful operation of a cloud solution

    You want Compute Engine to snapshot a persistent disk every hour and keep each snapshot for 30 days, with no manual steps. What should you use?

    1. A startup script on the instance that runs gcloud snapshots
    2. A Cloud Scheduler job that calls the Compute API
    3. A snapshot schedule resource policy attached to the disk Correct answer.
    4. An image created daily from the disk

    Explanation

    Snapshot schedules are implemented as resource policies that you attach to persistent disks; they create snapshots on the defined cadence and apply a retention policy that deletes expired snapshots. This keeps the automation inside Compute Engine with no custom code.

  138. Question 138Ensuring successful operation of a cloud solution

    An instance in a managed instance group becomes unresponsive while the VM itself is still running. You want the group to replace it automatically when the application stops answering health checks. What should you configure?

    1. An autoscaler with a CPU utilization target
    2. A rolling update with maxSurge and maxUnavailable
    3. A backend service with connection draining
    4. Autohealing with an application health check Correct answer.

    Explanation

    Autohealing continuously probes instances with the configured health check and recreates any instance that fails, which handles hung applications that the hypervisor still reports as running. Autoscaling changes the number of instances based on load but does not repair unhealthy ones.

  139. Question 139Ensuring successful operation of a cloud solution

    Creating 200 VMs in us-central1 fails with a quota error for CPUs. What is the appropriate way to proceed?

    1. Use preemptible instances to bypass the limit
    2. Retry the request until capacity becomes available
    3. Create the remaining instances in a different region without checking
    4. Request a quota increase for CPUs in us-central1 Correct answer.

    Explanation

    Compute Engine quotas are per project and per region, and the supported path to exceed a default limit is submitting a quota increase request through the quotas page or support, which is reviewed and usually granted quickly for standard CPU quota. Retrying does not change the limit.

  140. Question 140Ensuring successful operation of a cloud solution

    Your project ingests about 5 TB of DEBUG-level log entries per month that nobody reads, and you want to cut Cloud Logging ingest charges while keeping all WARNING and ERROR entries. What should you do?

    1. Shorten the retention period of the _Default bucket
    2. Add an exclusion filter for DEBUG severity entries Correct answer.
    3. Lower the project's budget amount
    4. Disable the Cloud Logging API for the project

    Explanation

    Exclusion filters are evaluated before entries are stored, so matching entries are never ingested and are not billed. Changing retention only affects how long stored entries are kept and does not reduce ingest volume.

  141. Question 141Ensuring successful operation of a cloud solution

    Your service writes one log line per request with an http_status field. You want to page when more than 5% of requests return a 5xx status over any 10-minute window. Which design meets the requirement?

    1. Create log-based metrics for 5xx and total requests, then alert on their ratio exceeding 0.05 Correct answer.
    2. Create a single log-based counter for 5xx entries and alert when its raw count exceeds 5 per minute
    3. Create an uptime check and alert when its failure count exceeds 5%
    4. Rely on Error Reporting to page when new error groups appear

    Explanation

    A percentage threshold requires a ratio, so two log-based metrics (5xx count and total count) are combined into a ratio time series in Cloud Monitoring and the alerting policy compares that ratio against 0.05 over the alignment window. Raw counts alone cannot express a proportion and vary with traffic volume.

  142. Question 142Ensuring successful operation of a cloud solution

    Users report slow responses. Cloud Trace shows that one downstream RPC accounts for 80% of request latency, but that service's CPU utilization and error rate look normal. Which tool best shows which code paths inside the service are consuming the time?

    1. Error Reporting issue details
    2. A Cloud Debugger snapshot at the RPC entry point
    3. Cloud Profiler wall-clock profiling of the service Correct answer.
    4. An uptime check on the upstream endpoint

    Explanation

    Once Trace localizes the problem to one service, Profiler attributes time or allocations to individual functions, and wall-clock profiling exposes time spent waiting rather than burning CPU, which matches a latency problem with normal CPU. Trace tells you which hop is slow but not which function inside it.

  143. Question 143Ensuring successful operation of a cloud solution

    Your application writes a custom metric each minute reporting how many messages it processed since the previous write, and you want Monitoring to sum those values over the alignment period. Which metric kind should the descriptor use?

    1. DELTA Correct answer.
    2. CUMULATIVE
    3. GAUGE with a distribution value type
    4. GAUGE

    Explanation

    A DELTA metric records the change in a value between consecutive writes, so Monitoring sums the deltas across an alignment period to produce the total for that window. GAUGE describes a point-in-time value and CUMULATIVE describes a running total measured from a fixed start time.

  144. Question 144Ensuring successful operation of a cloud solution

    An organization has 40 projects. All Admin Activity logs except those from a designated test folder must be routed to a single BigQuery dataset owned by the central logging project. Which configuration achieves this with least effort?

    1. A log sink in each project with an exclusion for the test folder
    2. A log bucket in the organization with a 400-day retention policy
    3. An organization-level aggregated sink with an exclusion filter Correct answer.
    4. A log-based metric defined at the organization level

    Explanation

    An aggregated sink created at the organization or folder level includes log entries from every project beneath it, so one sink plus an exclusion filter covers all 40 projects and omits the test folder. Per-project sinks would require 40 separate configurations that drift over time.

  145. Question 145Ensuring successful operation of a cloud solution

    You want to release a new Cloud Run revision to 5% of traffic, watch error rate and latency in Cloud Monitoring, and then shift more traffic while keeping an instant rollback path. Which approach should you take?

    1. Deploy to a second Cloud Run service and split traffic at an external load balancer
    2. Deploy the revision with no traffic and move 100% of traffic to it after one day
    3. Redeploy the previous image if error rates increase after a full cutover
    4. Split traffic by percentage across revisions and increase the new share gradually Correct answer.

    Explanation

    Cloud Run supports assigning percentage splits across revisions of the same service, which is exactly a canary release: a small share goes to the new revision, metrics are compared, and the split is adjusted or reverted to the previous revision instantly. A full cutover removes the gradual exposure and the immediate rollback benefit.

  146. Question 146Ensuring successful operation of a cloud solution

    You want to be alerted when your public web application in us-central1 stops responding to HTTP requests, and you want false positives caused by a single network path to be unlikely. What should you configure?

    1. A load balancer health check on the backend service with a one-second interval and no alerting policy.
    2. A Cloud Logging sink that exports every 5xx response and an alerting policy that fires when the exported log volume drops.
    3. An autoscaling policy on the managed instance group that scales on HTTP load balancing utilization.
    4. A public uptime check that probes the URL from several regions, plus an alerting policy on the uptime check metric. Correct answer.

    Explanation

    Uptime checks run from multiple Google-managed locations and record results in the uptime_check/check_passed metric, which alerting policies can watch, so a failure from one vantage point does not necessarily page anyone. Load balancer health checks only control traffic routing and do not notify you.

  147. Question 147Ensuring successful operation of a cloud solution

    Which statement about Cloud Audit Logs Admin Activity entries is correct?

    1. They are enabled by default and are retained for 30 days.
    2. They are always enabled and are retained for 400 days. Correct answer.
    3. They are written only when Data Access logs are also enabled.
    4. They are disabled by default and are retained for 30 days.

    Explanation

    Admin Activity audit logs record configuration and metadata changes, are written automatically for every project, cannot be disabled, and are retained for 400 days in Cloud Logging. Data Access logs are the category that is off by default and must be explicitly enabled.

  148. Question 148Ensuring successful operation of a cloud solution

    An incident affected one Compute Engine VM named web-1 in zone us-east1-b. You need to see only that VM's ERROR-severity log entries from the last hour. What is the most direct approach?

    1. Run gcloud logging read with a freshness filter for the last hour and then grep the output for the instance name.
    2. Export the entries to BigQuery with a log sink and then query the table with a severity filter over the last hour.
    3. Use a Logs Explorer query with resource.type="gce_instance", the instance_id label, severity=ERROR, and a one-hour range. Correct answer.
    4. Create a log-based metric for ERROR severity on that instance and chart the result in Cloud Monitoring.

    Explanation

    Logs Explorer supports a query language that filters on monitored resource type, resource labels, severity, and time range, which returns exactly the matching entries interactively. Exporting or aggregating first adds delay and extra components without narrowing the result as directly.

  149. Question 149Ensuring successful operation of a cloud solution

    Your application on GKE writes exceptions with stack traces to Cloud Logging. You want new error types to be grouped and counted automatically without writing any code. Which service should you use?

    1. Error Reporting Correct answer.
    2. Cloud Profiler
    3. Cloud Trace
    4. Cloud Monitoring uptime checks

    Explanation

    Error Reporting automatically detects exceptions and stack traces in logs from GKE, Compute Engine, App Engine, Cloud Run, and Cloud Functions, then groups them into error types with occurrence counts and first/last seen times. It requires no application changes beyond logging the stack trace.

  150. Question 150Ensuring successful operation of a cloud solution

    A service has a 99.9% availability SLO defined in Cloud Monitoring. You want to page quickly when errors are consuming the error budget much faster than the SLO permits, while avoiding pages for brief blips. What should you configure?

    1. A log-based metric that counts 5xx entries, charted on a dashboard reviewed by the team every day.
    2. An alerting policy on the service's 99th percentile request latency with a five-minute threshold.
    3. An alerting policy on the uptime check failure count that fires when failures are recorded in the last 24 hours.
    4. A burn-rate alerting policy on the SLO with a short lookback window and a high multiplier threshold. Correct answer.

    Explanation

    Burn-rate alerting compares the observed error rate against the error budget consumption rate, so a high multiplier over a short lookback window pages only when the budget is being spent fast enough to threaten the SLO. This ties the alert to user-facing reliability rather than to raw infrastructure signals.

  151. Question 151Ensuring successful operation of a cloud solution

    Your company has separate prod, staging, and dev projects. The SRE team needs one place to view dashboards and manage alerting policies covering all three projects. What should you configure?

    1. A metrics scope in a host monitoring project with the other three projects added as monitored projects. Correct answer.
    2. A log sink in each project that routes monitoring data to the host project's log bucket.
    3. An identical alerting policy recreated in each project that uses the same notification channels.
    4. Monitoring Viewer granted to the SRE team separately in each of the three projects.

    Explanation

    Cloud Monitoring uses a metrics scope: a host project whose dashboards and alerting policies can query the time-series data of up to 375 monitored projects. Adding the three projects to one scope gives a single pane of glass without duplicating configuration.

  152. Question 152Ensuring successful operation of a cloud solution

    A platform team maintains 30 projects and must deploy identical alerting policies and dashboards to each one, with every change reviewed and versioned. Which approach fits best?

    1. Recreate the policies in each project through the Cloud Console and record the settings in a shared spreadsheet.
    2. Define the policies and dashboards as code with the Google Cloud Terraform provider and apply them from a CI pipeline. Correct answer.
    3. Use gcloud commands saved in a shared document that engineers paste into a terminal when needed.
    4. Rely on Cloud Monitoring's recommended alerts in each project and adjust them manually as needs change.

    Explanation

    Infrastructure as code keeps the desired state in version control, makes changes reviewable through pull requests, and lets the same module be applied uniformly across all 30 projects. Console clicks and ad hoc commands drift over time and leave no audit trail of intent.

  153. Question 153Ensuring successful operation of a cloud solution

    A new image was rolled out to a GKE Deployment and is throwing errors. You must return to the previous revision with minimal effort. What should you do?

    1. Run kubectl rollout undo on the Deployment to return to the previous ReplicaSet. Correct answer.
    2. Edit the image tag directly on the running Pods with kubectl edit pod.
    3. Delete the Deployment and recreate it from the original manifest file.
    4. Scale the Deployment to zero replicas and then scale it back up.

    Explanation

    Deployments keep prior ReplicaSets according to revisionHistoryLimit, and kubectl rollout undo switches the Deployment back to the previous revision with a controlled rollout. Deleting and recreating loses history and causes an outage window.

  154. Question 154Ensuring successful operation of a cloud solution

    A managed instance group behind an external HTTP(S) load balancer keeps marking healthy instances as unhealthy, although the application listens on port 8080 and responds locally. What should you verify?

    1. That the instances have external IP addresses so the load balancer can reach them directly.
    2. That Cloud CDN is enabled on the backend service so cached responses absorb the health check probes.
    3. A firewall rule allowing the Google health check ranges to port 8080, and a health check path that returns 200. Correct answer.
    4. That the health check is configured on port 80 to match the load balancer frontend port.

    Explanation

    Health check probes originate from the Google ranges 35.191.0.0/16 and 130.211.0.0/22 and are subject to VPC firewall rules, so a rule allowing those sources to the serving port is required. The probe path must also return an HTTP 2xx response for the backend to be considered healthy.

  155. Question 155Ensuring successful operation of a cloud solution

    You must be paged when a Compute Engine instance stops reporting its custom heartbeat metric for more than five minutes, even though the VM itself may still exist. What should you use?

    1. An alerting policy with a metric-absence condition on the heartbeat metric and a five-minute duration. Correct answer.
    2. An alerting policy with a threshold that fires when the heartbeat metric value drops below 1.
    3. A log-based alert that fires when no log entries arrive from the instance.
    4. An uptime check configured against the instance's internal IP address.

    Explanation

    Cloud Monitoring alerting policies support an absent-metric condition, which fires when no data points match the filter for a configured duration. A threshold on the metric value never evaluates when data stops arriving, so it cannot detect silence.

  156. Question 156Ensuring successful operation of a cloud solution

    You want Cloud Logging to classify your application's JSON log lines by their own severity field and let you filter on fields such as user_id in Logs Explorer. What should you do?

    1. Write the fields to a local file on each VM and install the Ops Agent to parse them with a regular expression.
    2. Write each log line as a plain text string that begins with the severity level.
    3. Write structured JSON to stdout with a severity field and custom fields at the top level of the JSON object. Correct answer.
    4. Encode the fields in the log name so that they appear as indexed labels.

    Explanation

    When an entry's jsonPayload contains a recognized severity field, Cloud Logging uses it for the entry's severity, and top-level JSON fields become searchable in Logs Explorer queries. Structured output avoids custom parsing and keeps field types intact.

  157. Question 157Ensuring successful operation of a cloud solution

    Finance needs the monthly cloud bill broken down by team and by environment. What should you do?

    1. Enable billing export to BigQuery and rely on the default project-level grouping when writing the reports.
    2. Enable the Cloud Billing budgets API and poll it hourly to compute per-team spend.
    3. Apply consistent team and environment labels to resources and analyze the detailed billing export in BigQuery using those labels. Correct answer.
    4. Create a separate Cloud Billing account for each team and each environment so spend separates automatically.

    Explanation

    Resource labels propagate into the detailed billing export, so grouping by label columns in BigQuery attributes cost to team and environment even when many teams share a project. The export is the authoritative source for granular cost reporting.

  158. Question 158Ensuring successful operation of a cloud solution

    An engineer accidentally overwrote 2,000 objects in a Cloud Storage bucket. Object Versioning is disabled on the bucket, but soft delete is enabled with its default settings. How should you recover the overwritten objects?

    1. Apply an Object Lifecycle Management rule that reverts objects to their previous generations.
    2. Wait for the daily Google-managed bucket backup and restore the entire bucket from it.
    3. Run gsutil rsync from a replica bucket that holds the original copies of the objects.
    4. Restore the affected objects from their soft-deleted copies within the retention window. Correct answer.

    Explanation

    Soft delete retains deleted and overwritten object generations for a retention period (seven days by default) and they can be restored by name and generation while the window is open. Without versioning or a separate backup bucket, this is the recovery path available.

  159. Question 159Ensuring successful operation of a cloud solution

    An external auditing vendor must receive all Admin Activity logs from your project but must not be granted any IAM role inside that project. What should you configure?

    1. Grant roles/storage.objectViewer on the project's default log bucket to the vendor's service account.
    2. Enable Data Access logs so the vendor can pull all entries through the Cloud Logging API.
    3. Grant roles/logging.viewer on the project to the vendor's service account so it can read the logs directly.
    4. Create a log sink to a Pub/Sub topic in the vendor's project and grant the vendor subscriber access on that topic. Correct answer.

    Explanation

    A log sink can route matching entries to a destination in another project, and the vendor only needs pubsub.subscriber on that topic in its own project. This shares the required audit data without granting any permission on resources in your project.

  160. Question 160Ensuring successful operation of a cloud solution

    You updated a managed instance group to a new instance template and the new instances are failing their startup checks. You want every instance back on the previous template with minimal disruption. What should you do?

    1. Delete the managed instance group and recreate it from the previous instance template.
    2. Start a rolling update on the managed instance group that sets the instance template back to the previous version. Correct answer.
    3. Create a second managed instance group from the previous template and delete the failing group once it serves traffic.
    4. Stop all instances in the group and start them again so they boot from the previous template.

    Explanation

    A managed instance group remembers its prior template, and initiating a new rolling update with the previous template replaces instances gradually while respecting the update policy's surge and unavailable settings. This restores the known-good configuration without taking the whole group down.

  161. Question 161Configuring access and security

    Which IAM role grants full control over all resources in a project, including managing billing and granting roles to other principals?

    1. roles/viewer
    2. roles/editor
    3. roles/owner Correct answer.
    4. roles/browser

    Explanation

    The basic roles are Owner, Editor, and Viewer. Owner includes every Editor permission plus the ability to manage access policies and billing, so it is the broadest basic role available at project level.

  162. Question 162Configuring access and security

    An operator should only be able to start and stop Compute Engine instances. Which role choice best follows the principle of least privilege?

    1. A predefined role for Compute Engine instance administration Correct answer.
    2. The Owner role granted at the folder level so it is inherited
    3. The basic Viewer role granted at the project level
    4. The basic Editor role granted at the project level

    Explanation

    Predefined roles bundle a curated set of permissions for a specific service, which is far narrower than the basic roles. Basic roles such as Editor and Owner span nearly all services and should be avoided for day-to-day operational access.

  163. Question 163Configuring access and security

    What is the recommended way for an application on a Compute Engine instance to obtain credentials for Google Cloud APIs?

    1. Use the attached service account's metadata-server credentials Correct answer.
    2. Store a downloaded JSON service account key on the instance's persistent disk
    3. Embed the key in the application's source code
    4. Download the key from the metadata server at startup and save it

    Explanation

    When a service account is attached to an instance, the metadata server issues short-lived access tokens automatically. This avoids long-lived key files that can be copied, committed to source control, or forgotten during rotation.

  164. Question 164Configuring access and security

    Which mechanism allows or denies inbound and outbound traffic for Compute Engine instances in a VPC network?

    1. Cloud NAT
    2. IAM roles on the instance's service account
    3. Cloud Armor security policies
    4. VPC firewall rules Correct answer.

    Explanation

    VPC firewall rules are stateful and apply to instances based on network tags or service accounts, controlling which connections are permitted. They operate at layers 3 and 4 and are separate from identity and application-layer controls.

  165. Question 165Configuring access and security

    Which Cloud Audit Logs stream is always enabled and records administrative configuration changes such as creating a VM or altering an IAM policy?

    1. Admin Activity logs Correct answer.
    2. System Event logs
    3. Policy Denied logs
    4. Data Access logs

    Explanation

    Admin Activity logs capture writes and configuration changes made through the API and are enabled for every project with no charge. Data Access logs, by contrast, record reads and writes of user data and must be turned on explicitly.

  166. Question 166Configuring access and security

    What is the primary purpose of OS Login on Compute Engine?

    1. It controls SSH access to VMs using IAM rather than metadata keys Correct answer.
    2. It stores the application secrets that the operating system reads at boot time
    3. It replaces VPC firewall rules for SSH traffic
    4. It encrypts the boot disk of each Compute Engine instance with CMEK

    Explanation

    OS Login ties Linux logins to a Google identity and to IAM roles such as roles/compute.osLogin, so administrators no longer distribute and maintain per-instance SSH keys in metadata. Firewall rules still determine whether SSH traffic can reach the instance.

  167. Question 167Configuring access and security

    A folder-level IAM policy grants a group roles/storage.objectViewer. A project inside that folder has no storage-related bindings. What access do members of the group have to buckets in that project?

    1. Read access only after an explicit project-level grant
    2. No access unless an organization policy allows it
    3. Read access, because policies are inherited from the folder Correct answer.
    4. No access, because grants must be made at the project level

    Explanation

    IAM policies are hierarchical: the effective policy on a resource is the union of allow bindings set on that resource and on all of its ancestors. A grant at the folder therefore applies to every project and resource beneath it.

  168. Question 168Configuring access and security

    Which role must a user have to reach a web application published through Identity-Aware Proxy over HTTPS?

    1. roles/iap.tunnelResourceAccessor
    2. roles/compute.networkUser
    3. roles/iap.httpsResourceAccessor Correct answer.
    4. roles/viewer

    Explanation

    IAP checks for the IAP-secured Web App User role on the resource or its parent before letting a request through to a web application. The tunnel role is the equivalent for TCP forwarding, such as SSH or RDP.

  169. Question 169Configuring access and security

    You need a custom IAM role in a project that grants only get and list permissions on Cloud Storage buckets. Which command creates the role?

    1. gcloud iam service-accounts create
    2. gcloud auth application-default login
    3. gcloud iam roles create Correct answer.
    4. gcloud projects add-iam-policy-binding

    Explanation

    Custom roles are defined with the gcloud iam roles create command, which takes a title, a permission list, and a scope such as a project or organization. The resulting role can then be bound to principals like any predefined role.

  170. Question 170Configuring access and security

    A workload running as service account A must call Google Cloud APIs as service account B. Which grant on service account B enables this?

    1. roles/iam.serviceAccountAdmin, which manages the account's policy and lifecycle
    2. roles/iam.serviceAccountUser, which lets a principal attach the account to resources
    3. roles/iam.serviceAccountTokenCreator, which allows minting tokens as that account Correct answer.
    4. roles/iam.workloadIdentityUser, which binds external identities to the account

    Explanation

    Impersonation requires the Service Account Token Creator role on the target account, because that role permits generating access tokens and signing as the account. Service Account User authorizes attaching an account to a resource but not minting credentials for it.

  171. Question 171Configuring access and security

    Contractors should be able to use a project only during a two-week migration window. Which IAM mechanism expresses this most directly?

    1. An organization policy constraint enforced during the window
    2. A basic role granted at the organization level and revoked by hand
    3. An IAM conditional role binding with a date and time condition Correct answer.
    4. A custom role whose permissions are removed at the end of the window

    Explanation

    IAM Conditions attach attributes such as request time to a role binding, so the grant stops applying automatically when the condition no longer evaluates to true. This removes the need for manual revocation and keeps the permissions scoped to the intended period.

  172. Question 172Configuring access and security

    You want Cloud KMS to automatically produce new versions of a symmetric encryption key every 90 days. What should you configure?

    1. Add an IAM condition that expires the key after 90 days
    2. Set the key's rotation period to 90 days Correct answer.
    3. Create a new key ring every 90 days
    4. Use envelope encryption with a new data key for each object

    Explanation

    A Cloud KMS key has a rotation period; when it elapses, a new key version is created automatically and becomes the primary version for new encryptions. Older versions are retained so data encrypted with them can still be decrypted.

  173. Question 173Configuring access and security

    An application must fetch database passwords at runtime with versioning and an audit trail of access. Which service is designed for this?

    1. A Cloud Storage bucket with bucket-level IAM
    2. Secret Manager Correct answer.
    3. Cloud KMS
    4. Environment variables in the deployment YAML

    Explanation

    Secret Manager stores sensitive values, keeps multiple versions, and records access through audit logs and IAM. Workloads retrieve the current version at runtime instead of embedding the value in configuration.

  174. Question 174Configuring access and security

    What does a VPC Service Controls perimeter primarily provide?

    1. Automatic rotation of service account keys
    2. Network firewall rules applied between subnets in a VPC
    3. Encryption of data at rest in Cloud Storage buckets using CMEK
    4. A security boundary around selected projects and services Correct answer.

    Explanation

    A service perimeter defines a set of projects and services, and API requests that would move data across the boundary are blocked even when the caller holds valid IAM permissions. This mitigates data exfiltration rather than network-level attacks.

  175. Question 175Configuring access and security

    An organization must stop anyone from creating service account keys, regardless of the IAM roles they hold. What should be used?

    1. An organization policy that disables key creation Correct answer.
    2. A deny policy applied to each service account
    3. VPC Service Controls deployed in dry-run mode
    4. Removing the Editor role from every user in the organization

    Explanation

    Organization policy constraints such as constraints/iam.disableServiceAccountKeyCreation are evaluated independently of IAM and block the action even for project owners. A constraint set at the organization node applies to every current and future project beneath it.

  176. Question 176Configuring access and security

    You need Admin Activity audit logs retained for seven years in Cloud Storage. What should you configure?

    1. An exclusion filter on the default log bucket
    2. Changing the retention period of the _Default log bucket to seven years
    3. A log-based metric built on the Admin Activity log entries
    4. A log sink to a Cloud Storage bucket with a retention policy Correct answer.

    Explanation

    A log sink routes copies of matching log entries to a destination such as Cloud Storage, and the destination bucket's retention policy or bucket lock then governs how long the copies are kept. The _Default bucket is meant for recent operational access, not multi-year archives.

  177. Question 177Configuring access and security

    A workload running on AWS needs to call Google Cloud APIs without long-lived service account keys. Which feature fits?

    1. Downloading a service account key and shipping it with the workload
    2. Cloud Identity groups
    3. OS Login
    4. Workload Identity Federation Correct answer.

    Explanation

    Workload Identity Federation exchanges credentials from an external identity provider, such as AWS or OIDC, for short-lived Google Cloud access tokens. Nothing long-lived is stored with the external workload.

  178. Question 178Configuring access and security

    A team of 40 engineers needs identical access to a project. What is the recommended way to manage their bindings?

    1. Grant basic roles at the organization level so everyone inherits them
    2. Grant the roles to each engineer individually
    3. Grant the roles to a Google Group containing the engineers Correct answer.
    4. Create a custom role and assign it to a shared user account

    Explanation

    Binding roles to a Google Group means membership changes are handled in one place while the IAM policy stays stable and auditable. Individual bindings multiply and drift, and shared accounts destroy per-person accountability.

  179. Question 179Configuring access and security

    Which service protects an external HTTP(S) load balancer against SQL injection and cross-site scripting attacks?

    1. Identity-Aware Proxy
    2. Cloud NAT
    3. VPC firewall rules
    4. Cloud Armor Correct answer.

    Explanation

    Cloud Armor security policies attach to backend services behind an external HTTP(S) load balancer and evaluate layer 7 rules such as the preconfigured OWASP signatures. Firewall rules operate at the connection level and cannot inspect HTTP payloads.

  180. Question 180Configuring access and security

    You must determine which users read a specific object in a Cloud Storage bucket. What log configuration makes this possible?

    1. Admin Activity logs, which are enabled by default for every project
    2. Policy Denied logs
    3. Data Access logs, which are not enabled by default Correct answer.
    4. System Event logs

    Explanation

    Reads of object data are recorded only in Data Access audit logs, which are disabled by default because of their volume. They must be explicitly enabled for the service before object-read events appear.

  181. Question 181Configuring access and security

    A user holds roles/storage.objectViewer at the project level, and an IAM deny policy on one bucket denies storage.objects.get to that user. What happens when the user reads an object in that bucket?

    1. Denied, because deny rules take precedence over allow rules Correct answer.
    2. Allowed, because deny policies apply only to service accounts
    3. Denied only when the user also holds a basic role on the bucket
    4. Allowed, because an allow policy at the project level always wins

    Explanation

    IAM deny policies are evaluated before allow policies, and an applicable deny rule overrides any allow binding inherited from a project, folder, or organization. Deny policies can target users, groups, and service accounts alike.

  182. Question 182Configuring access and security

    Service account SA1 lives in project A. A workload running as SA1 must write objects into a bucket in project B. What is the minimal correct configuration?

    1. Move SA1 into project B, since service accounts can only access resources in their own project
    2. Grant roles/storage.objectCreator to SA1 at the organization level so it covers every bucket
    3. Grant roles/storage.objectCreator to SA1 on the bucket in project B Correct answer.
    4. Grant roles/iam.serviceAccountUser to SA1 in project A

    Explanation

    A service account is an identity that can be granted roles on resources in any project, so the minimal fix is a binding on the target bucket. Organization-wide grants would exceed the requirement and serviceAccountUser grants no storage permissions.

  183. Question 183Configuring access and security

    A VM's service account has roles/storage.objectViewer, but the instance was created with access scopes that do not cover Cloud Storage. What is the result when the application on the instance tries to read an object?

    1. It fails only if OS Login has been disabled on the instance
    2. It fails, because access scopes also restrict the instance's API calls Correct answer.
    3. It succeeds, because IAM grants always take precedence over access scopes
    4. It succeeds as long as the bucket is in the same project

    Explanation

    For credentials issued to a Compute Engine instance, a request must be permitted by both the IAM policy and the instance's access scopes. Scopes can only narrow what the service account is allowed to do, never widen it.

  184. Question 184Configuring access and security

    A compliance team must view IAM policies and audit logs across every project but must not be able to change resources. Which grant meets this with least privilege?

    1. roles/iam.securityReviewer combined with roles/editor granted at the folder level
    2. roles/owner granted at the organization level
    3. roles/editor granted on every project in the organization
    4. roles/iam.securityReviewer and roles/logging.viewer at the org level Correct answer.

    Explanation

    Security Reviewer grants get and list access to IAM policies and related resource metadata without write permissions, and Logging Viewer grants read access to logs. Granting both at the organization node covers all current and future projects read-only.

  185. Question 185Configuring access and security

    You want to evaluate a new VPC Service Controls perimeter without interrupting production traffic yet. What should you do?

    1. Apply a deny policy to the projects inside the perimeter
    2. Add a VPC firewall rule that blocks outbound traffic on port 443
    3. Deploy the perimeter in enforced mode and review the audit logs afterwards
    4. Deploy the perimeter in dry-run mode Correct answer.

    Explanation

    In dry-run mode a perimeter logs the requests that would have been violated without blocking them, so you can tune rules before enforcing. Switching to enforced mode later begins denying those requests.

  186. Question 186Configuring access and security

    You create a Compute Engine instance with the default service account and do not modify any IAM bindings. What role does that service account have in the project by default?

    1. roles/editor Correct answer.
    2. No role; only access scopes apply.
    3. roles/viewer
    4. roles/owner

    Explanation

    The Compute Engine default service account is granted the basic Editor role on the project when it is created. This role is broad and should be replaced with a least-privilege predefined or custom role.

  187. Question 187Configuring access and security

    Which statement best describes the basic IAM roles such as Owner, Editor, and Viewer?

    1. They can be granted only to service accounts.
    2. They are broad legacy roles that span many services. Correct answer.
    3. They apply only to one named resource.
    4. They are custom roles created by the organization.

    Explanation

    Basic roles are legacy project-wide roles that bundle permissions across many Google Cloud services. For least privilege, predefined or custom roles are preferred instead.

  188. Question 188Configuring access and security

    You want to grant a role to every user in the Cloud Identity domain example.com. Which IAM member identifier should you use?

    1. group:example.com
    2. serviceAccount:example.com
    3. user:example.com
    4. domain:example.com Correct answer.

    Explanation

    The domain: prefix grants a role to all identities in a Cloud Identity or Google Workspace domain. Individual users and groups use the user: and group: prefixes.

  189. Question 189Configuring access and security

    You need to determine why a specific user can access a Cloud Storage bucket. Which tool should you use?

    1. Policy Troubleshooter Correct answer.
    2. IAM Recommender
    3. Cloud Audit Logs
    4. VPC Flow Logs

    Explanation

    Policy Troubleshooter evaluates the principal's IAM policies, including inherited bindings and deny policies, and explains whether access is granted. Audit logs show past activity rather than the policy evaluation result.

  190. Question 190Configuring access and security

    A developer must create a Compute Engine VM that runs as an existing service account. The developer already has roles/compute.instanceAdmin.v1. What additional IAM role is required on the service account?

    1. roles/iam.roleAdmin
    2. roles/iam.serviceAccountUser Correct answer.
    3. roles/iam.serviceAccountKeyAdmin
    4. roles/iam.securityAdmin

    Explanation

    Attaching a service account to a resource requires the iam.serviceAccounts.actAs permission, which is included in roles/iam.serviceAccountUser. Compute instance administration alone does not grant the right to act as a service account.

  191. Question 191Configuring access and security

    An application needs short-lived OAuth 2.0 access tokens for a service account without downloading a key. Which role should be granted to the calling identity on the service account?

    1. roles/iam.serviceAccountUser
    2. roles/iam.serviceAccountKeyAdmin
    3. roles/iam.workloadIdentityUser
    4. roles/iam.serviceAccountTokenCreator Correct answer.

    Explanation

    The Service Account Token Creator role allows an identity to mint short-lived credentials for a service account. Service Account User allows attaching the account to resources rather than generating tokens.

  192. Question 192Configuring access and security

    Pods in a GKE cluster must call Google Cloud APIs as a Google service account without storing service account keys. Which feature should you enable?

    1. Cloud Identity-Aware Proxy
    2. GKE Workload Identity Correct answer.
    3. Compute Engine OS Login
    4. VPC Service Controls

    Explanation

    GKE Workload Identity maps a Kubernetes service account to a Google service account so pods receive short-lived credentials. It removes the need for exported service account keys in the cluster.

  193. Question 193Configuring access and security

    An organization wants to prevent IAM policies from granting access to identities outside its Cloud Identity domain. Which organization policy constraint should it enforce?

    1. constraints/storage.uniformBucketLevelAccess
    2. constraints/compute.vmExternalIpAccess
    3. constraints/iam.allowedPolicyMemberDomains Correct answer.
    4. constraints/iam.disableServiceAccountKeyCreation

    Explanation

    The allowedPolicyMemberDomains constraint restricts the domains that can appear as IAM members in policies. It blocks bindings to external users or domains while still allowing permitted internal identities.

  194. Question 194Configuring access and security

    A bucket currently uses fine-grained ACLs. You want IAM policies to be the only way access is granted to objects in the bucket. What should you do?

    1. Enable uniform bucket-level access Correct answer.
    2. Enable Object Lifecycle Management
    3. Set the bucket's default storage class
    4. Enable versioning on the bucket

    Explanation

    Uniform bucket-level access disables object ACLs and grants access only through IAM policies on the bucket and project. Lifecycle settings, storage class, and versioning do not change the authorization model.

  195. Question 195Configuring access and security

    A user needs to query tables in one BigQuery dataset but must not see any other dataset in the project. What is the most appropriate grant?

    1. roles/bigquery.dataViewer at the project level
    2. roles/bigquery.admin at the project level
    3. roles/bigquery.dataViewer on the dataset Correct answer.
    4. roles/bigquery.jobUser on the dataset

    Explanation

    Granting data viewer on the dataset limits table read access to that dataset. Project-level grants would affect all datasets, and jobUser only allows running jobs rather than reading table data.

  196. Question 196Configuring access and security

    A security team wants to list every resource in an organization that a specific principal can access with a given role. Which tool should they use?

    1. Policy Troubleshooter
    2. Access Transparency
    3. IAM Recommender
    4. Policy Analyzer Correct answer.

    Explanation

    Policy Analyzer evaluates IAM policies across a scope and reports which resources a principal can access. Policy Troubleshooter explains a single access decision rather than enumerating resources.

  197. Question 197Configuring access and security

    You want to grant a group roles/storage.objectViewer on a bucket only for objects whose names start with reports/. What should you use?

    1. A service account access scope
    2. An IAM condition with a resource name prefix Correct answer.
    3. An organization policy constraint
    4. A bucket ACL

    Explanation

    IAM conditions can restrict a role binding based on attributes such as the resource name. A condition using a prefix on resource.name limits the grant to matching objects.

  198. Question 198Configuring access and security

    Service account A can impersonate service account B, and B can impersonate service account C. A workload running as A requests credentials for C directly. What is required for this request to succeed?

    1. No grant is needed if all accounts are in the same project.
    2. A only needs Token Creator on B.
    3. A must have Token Creator on C directly. Correct answer.
    4. B's Token Creator role on C is sufficient.

    Explanation

    Service account impersonation is not transitive: permission to impersonate B does not imply permission to impersonate C. The caller must be granted the token-creating permission on the target service account, C.

  199. Question 199Configuring access and security

    A regulated customer wants a record of when Google Support accessed their project data and wants to approve such access before it happens. Which pair of services should they use?

    1. Cloud Audit Logs and VPC Service Controls
    2. Security Command Center and IAM Recommender
    3. Access Transparency and Access Approval Correct answer.
    4. Access Transparency and Cloud KMS

    Explanation

    Access Transparency provides logs of Google personnel access to customer data. Access Approval requires explicit customer approval before that access occurs.

  200. Question 200Configuring access and security

    An organization policy constraint is enforced at the organization node. A folder administrator tries to set a less restrictive value on the same constraint at the folder. What happens?

    1. The folder cannot loosen the organization value. Correct answer.
    2. The constraint is automatically disabled for the folder.
    3. The folder value applies only to newly created resources.
    4. The folder value overrides the organization value.

    Explanation

    Organization policy inheritance is hierarchical: a child resource can tighten a constraint but cannot relax a value enforced by an ancestor. The organization-level restriction therefore continues to apply.

GCP ACE flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Resource hierarchy order
The resource hierarchy is Organization > folders > projects > resources. IAM allow policies are inherited downward, and a child policy cannot remove a parent grant.
Billing account link
A project can be linked to at most one billing account, while one billing account can pay for multiple projects; closing the billing account disables linked projects' billable services.
gcloud default project
Running gcloud config set project PROJECT_ID changes the default project used by subsequent gcloud commands; it does not grant any IAM permissions.
Quota versus limit
Quotas are project- or region-scoped allocations enforced by APIs. Allocation quotas cap resource counts, rate quotas cap request rates, and increases are requested through the console.
Cloud Identity
Cloud Identity provides identity, group, and endpoint management without Google Workspace collaboration apps; Google Groups from Cloud Identity can be granted IAM roles.
Labels versus tags
Labels are key-value metadata used for cost allocation, filtering, and automation. Tags are key-value metadata used in IAM conditions and firewall targeting; neither grants access by itself.
VPC scope
A VPC network is global, while its subnets are regional. Resources in different regions of the same VPC can communicate using internal IP addresses.
Firewall rule priority
VPC firewall rules are stateful and evaluated by priority; a lower number means higher priority. There are implied deny ingress and allow egress rules.
Storage class minimums
Cloud Storage classes are Standard, Nearline with a 30-day minimum, Coldline with 90 days, and Archive with 365 days. Deleting objects early incurs early deletion charges.
Cloud SQL HA
Cloud SQL regional high availability uses a standby instance in a second zone with synchronous replication. Failover is automatic; read replicas are separate and serve read scaling.
CUD versus SUD
Committed use discounts require a 1- or 3-year commitment for steady usage. Sustained use discounts apply automatically to qualifying Compute Engine usage within a month.
BigQuery partitioning
Partitioning divides a BigQuery table by date, timestamp, or integer range. Clustering sorts data within partitions, reducing bytes scanned by filtered queries.
App Engine deploy
gcloud app deploy app.yaml creates a new App Engine version and routes traffic according to promotion settings such as --promote or --no-promote.
Cloud Run deploy
gcloud run deploy SERVICE --image IMAGE creates or updates a Cloud Run revision; --allow-unauthenticated permits public invocation of the service.
GKE Autopilot
GKE Autopilot is a mode where Google manages nodes, scaling, and infrastructure; users deploy pods and are billed for requested CPU and memory.
Artifact Registry
Artifact Registry stores container images and language packages, integrates with IAM, and supports vulnerability scanning; it is the successor to Container Registry.
Cloud Build steps
Cloud Build runs each step in cloudbuild.yaml as a container. Substitutions provide variables, and triggers start builds from repository events.
Terraform plan
terraform plan previews changes by comparing configuration with state and real resources; terraform apply then creates, updates, or deletes resources to match.
Pub/Sub subscriptions
Publishers send messages to a Pub/Sub topic, and each subscription receives a copy. Default delivery is at-least-once, governed by the acknowledgment deadline.
Uptime check
A Cloud Monitoring uptime check probes an endpoint from multiple locations. An alerting policy can fire when the check fails for a configured duration.
Log sink
A Cloud Logging sink routes log entries to Cloud Storage, BigQuery, Pub/Sub, or another project. Inclusion and exclusion filters determine which entries are routed.
Audit log defaults
Admin Activity and System Event audit logs are always enabled. Data Access audit logs are disabled by default and must be explicitly enabled for most services.
SLO error budget
An SLO specifies a target and a time window. The error budget equals 1 minus the SLO target, representing allowed unreliability.
Error Reporting
Error Reporting aggregates and groups application exceptions found in logs. It can alert on new error groups but is not a general metrics store.
Billing export
Billing export to BigQuery writes detailed cost and usage data into datasets. Queries over these tables support cost analysis and chargeback reporting.
Basic IAM roles
Owner, Editor, and Viewer are broad legacy roles. Owner includes billing and IAM control; predefined and custom roles provide narrower permissions.
Service account keys
Service account keys are long-lived credentials. Attached service accounts, Workload Identity Federation, or short-lived tokens are preferred over downloaded key files.
Workload Identity Federation
Workload Identity Federation lets external workloads exchange identity-provider credentials for short-lived Google Cloud credentials, avoiding service account key files.
IAM conditions
IAM conditions are CEL expressions attached to role bindings. The grant applies only when the expression evaluates true for attributes such as resource name or time.
VPC Service Controls
VPC Service Controls define a service perimeter that restricts access to supported Google Cloud services and helps reduce data exfiltration; they are not network firewalls.
Cloud KMS
Cloud KMS manages key rings and cryptographic keys. Customer-managed encryption keys let supported services encrypt data with keys the customer controls.
Organization Policy
Organization Policy constraints are set at organization, folder, or project level and restrict resource configurations, such as allowed locations or public IP addresses.

GCP ACE study guide

Setting Up a Cloud Solution Environment

The resource hierarchy

Everything in Google Cloud hangs off a four-level tree: organization (tied to a Cloud Identity or Workspace domain) → folders → projects → resources. IAM policies inherit downward, so a role granted on a folder applies to every project beneath it. Billing accounts sit outside the hierarchy: a project links to exactly one billing account, and unlinking it disables most paid services. A project is the fundamental unit of API enablement, quota, and cost attribution - questions asking how to isolate environments usually want separate projects, not just separate VPCs.

Bringing a project online

Typical flow: create or select the project, attach the billing account, enable the required APIs (gcloud services enable), then grant least-privilege IAM. Traps: APIs are per project, so a service enabled in one project does nothing for another; a role granted on project A does not carry to project B; roles/owner is enormous while roles/editor still cannot read some data or manage IAM. New projects may auto-grant the creator Owner - do not depend on it in production.

Interacting with the platform

Cloud Shell, the gcloud CLI, the Console, client libraries, and the REST API all reach the same control plane. Know gcloud config set project, gcloud auth list, and named configurations for switching contexts. Cloud Shell is ephemeral, free to use, and ships with the SDK preinstalled.

Cost and quota guardrails

Budgets and budget alerts notify; they do not cap spend - a very common distractor. Quotas are per project and often per region, and may be rate or allocation quotas. Labels drive cost allocation and reporting; network tags and service accounts drive firewall targeting.

Rule of thumb: environment isolation → separate projects inside a folder; shared guardrails → IAM and organization policies set at the organization or folder level.

Planning and Configuring a Cloud Solution

Match workload to service

Compute: Compute Engine for full VM control, GKE for container orchestration, Cloud Run for stateless containers with no cluster to operate, App Engine for app-code-only deployment, Cloud Functions for event-driven glue. Data: Cloud SQL for regional relational OLTP up to a few TB, Spanner for globally consistent horizontally scalable relational, Firestore for document/mobile, Bigtable for high-throughput wide-column time series, BigQuery for analytics. A frequent trap is choosing Cloud SQL for a globally distributed write-heavy workload, or Bigtable for transactional queries.

Regions, zones, and high availability

Resources are regional or zonal. Real HA needs at least two zones: a multi-zone managed instance group, a regional persistent disk, or Cloud SQL with a failover replica. A single-zone MIG with autoscaling is scalable but not highly available - read the question for the word "availability."

Network design

A VPC network is global; subnets are regional. Firewall rules are stateful, attach to the network, and target instances by network tag or service account; baseline behavior is default-deny ingress and implied-allow egress. Cloud NAT provides outbound-only internet for instances with no external IP; Private Google Access reaches Google APIs privately; Shared VPC centralizes networking under a host project so many projects share subnets. Pick the load balancer deliberately: global external HTTP(S) for multi-region web, regional internal for private tiers, network LB for non-HTTP protocols.

Cost levers

Sustained-use and committed-use discounts, Spot VMs for interruptible batch work, storage class selection, and network egress dominate savings questions.

Rule of thumb: identify the hard constraint first - global consistency, latency, availability, or operational burden - and it will usually select the product for you.

Deploying and Implementing a Cloud Solution

Deployment mechanisms

Deploy with the Console, gcloud, infrastructure-as-code (Terraform or Deployment Manager), Cloud Build pipelines, or Marketplace for third-party solutions. Cloud Build runs containerized steps from cloudbuild.yaml: build, test, push to Artifact Registry, deploy. In exams the mechanism rarely matters - the resource configuration does.

Compute deployments

GKE: choose zonal versus regional (regional gives control-plane HA), then apply manifests. Cloud Run: deploy an image, set concurrency and scaling limits, and decide whether to allow unauthenticated invocations or require IAM. App Engine: app.yaml defines runtime and scaling. Cloud Functions: runtime, entry point, trigger, memory, timeout. Compute Engine fleets: instance template plus MIG for repeatability, startup scripts for bootstrapping.

Storage and data deployment

Bucket names are globally unique, and the name and location are immutable after creation - you must create a new bucket and copy to change either. Set storage class, versioning, lifecycle rules, retention policy, and uniform bucket-level access at creation. Enabling uniform bucket-level access disables object ACLs. For Cloud SQL, choose tier, region, HA, backups, and private IP via VPC peering.

Traps worth memorizing

  • A deploy failing with 403 is usually a missing API enablement or IAM grant, not bad code.
  • Lifecycle rules act on object age, class, or version count; they do not move data between buckets.
  • IaC questions test the concept (reproducible, version-controlled environments), not syntax.

Rule of thumb: every manual step you perform should be encoded in a template or pipeline so the same environment can be rebuilt identically.

Ensuring Successful Operation of a Cloud Solution

Observability stack

Cloud Monitoring handles metrics, dashboards, uptime checks, and alerting policies. Cloud Logging stores entries in log buckets. Error Reporting aggregates exceptions, Cloud Trace shows request latency, and Profiler shows CPU/heap for supported runtimes. Critically, default VM metrics cover only CPU, network, and disk I/O - memory, disk-space, and process metrics require the Ops Agent. That is a classic exam trap.

The logs pipeline

Entries flow from resources into log buckets (_Default, _Required) and can be routed by sinks. Aggregated sinks export to another project, BigQuery, Cloud Storage, or Pub/Sub; the destination must grant permission to the sink's writer identity. Exclusions drop entries before storage, while a sink with no filter exports everything - do not confuse the two. Log-based metrics turn log entries into counters or distributions you can alert on.

Audit logs

Admin Activity audit logs are always enabled and cannot be turned off. Data Access audit logs are disabled by default and, at volume, cost real money. Policy Denied logs record organization policy and VPC Service Controls denials.

Alerting and reliability

Alerting policies trigger on metric thresholds, absence of data, or log-based metrics, and route to notification channels (email, SMS, Pub/Sub, webhooks). Uptime checks probe HTTP/TCP from multiple regions. SLOs express reliability targets and can back burn-rate alerts.

Troubleshooting cues

403 = IAM or disabled API; 404 = wrong project or resource; quota errors = request an increase or redesign; SSH failures usually mean a missing firewall rule, no external IP, or IAP not configured.

Rule of thumb: for "why were we not alerted?" check three things in order - does the metric exist, does the policy condition match it, and is the notification channel wired to a human?

Configuring Access and Security

The IAM model

An allow policy binds members (user, group, service account, domain) to roles on a resource. Policies inherit down the hierarchy and are additive - there is no implicit deny unless a deny policy or organization policy applies. Role types: primitive (Owner/Editor/Viewer - broad, usually wrong in production), predefined (service-scoped, usually the right answer), and custom (curated permission lists). Grant to Google groups rather than individuals so membership changes do not require policy edits.

Service accounts

A service account is both an identity and a resource. Attach one to a VM, GKE workload, or Cloud Run service so code receives credentials automatically. To let a user create a resource as a service account, they need iam.serviceAccounts.actAs, contained in roles/iam.serviceAccountUser. Prefer short-lived credentials - impersonation or Workload Identity Federation - because exported JSON keys are long-lived and easily leaked. Organization policy can block key creation entirely.

Encryption and secrets

Cloud KMS holds key rings and keys (regional or global) with rotation schedules; CMEK uses your key for disks, storage buckets, BigQuery, and databases. Secret Manager stores API keys and passwords with versions and IAM control. Customer-supplied encryption keys are not stored by Google.

Guardrails and edge protection

Organization policies are inherited constraints - disable service account key creation, restrict domain sharing, enforce uniform bucket-level access - that block actions regardless of IAM. VPC Service Controls build a perimeter around APIs to reduce exfiltration risk. Cloud Armor protects external HTTP(S) load balancers; Identity-Aware Proxy gates application access by identity instead of network position.

Traps

  • Roles are additive; you cannot subtract access with IAM alone.
  • Data Access audit logs are off by default.
  • roles/editor still cannot read some data or change IAM.

Rule of thumb: least privilege means the narrowest predefined role, at the lowest resource in the hierarchy that still works, granted to a group.

GCP ACE cheat sheet

Quick reference

Setting up a cloud solution environment

  • Resource hierarchy order is Organization > folders > projects > resources; IAM allow policies are inherited downward, and a child policy cannot revoke a parent grant.

  • A project belongs to exactly one organization or to no organization; folders can nest and can contain projects or other folders.

  • A project can be linked to at most one billing account, while one billing account can pay for multiple projects; budgets and alerts attach to billing accounts or projects.

  • Running gcloud config set project PROJECT_ID changes the default project for subsequent gcloud commands; it does not grant IAM permissions.

  • Quotas are enforced per project and per region: allocation quotas cap resource counts, rate quotas cap API call rates, and increases are requested in the console.

  • Cloud Identity provides identity, group, and endpoint management without Google Workspace collaboration apps; Google Groups can be granted IAM roles.

  • Labels are key-value pairs used for cost allocation, filtering, and automation; tags are key-value pairs used in IAM conditions and firewall targeting.

  • Enabling an API is per project; the API must be enabled in the project where the dependent resources are created.

  • gcloud configurations are named profiles; gcloud config configurations activate NAME switches the default project, account, and region settings.

Quick reference

Planning and configuring a cloud solution

  • A VPC network is global while its subnets are regional; resources in different regions of the same VPC communicate over internal IP addresses.

  • VPC firewall rules are stateful and evaluated by priority; a lower number means higher priority, with implied deny ingress and allow egress rules.

  • Region choice is driven by latency, data residency, and service availability; zones within a region are independent failure domains.

  • Cloud Storage classes are Standard, Nearline with a 30-day minimum, Coldline with 90 days, and Archive with 365 days; early deletion incurs prorated charges.

  • Cloud SQL regional high availability uses a standby in a second zone with synchronous replication and automatic failover; read replicas serve read scaling.

  • Committed use discounts require a 1- or 3-year commitment for steady usage; sustained use discounts apply automatically to qualifying monthly Compute Engine usage.

  • BigQuery partitioning divides tables by date, timestamp, or integer range; clustering sorts data within partitions to reduce bytes scanned by filtered queries.

  • Load balancer selection: global external Application LB for HTTP(S) across regions; regional external proxy Network LB for TCP or SSL; passthrough Network LB preserves client IP.

  • A Cloud Storage bucket location is immutable after creation; regional locations reduce latency, multi-region locations increase availability, and dual-region uses two specific regions.

Quick reference

Deploying and implementing a cloud solution

  • gcloud app deploy app.yaml creates a new App Engine version and routes traffic according to promotion settings such as --promote or --no-promote.

  • gcloud run deploy SERVICE --image IMAGE creates or updates a Cloud Run revision; --allow-unauthenticated permits public invocation of the service.

  • GKE Autopilot manages nodes, scaling, and infrastructure; users deploy pods and billing is based on requested CPU and memory.

  • Artifact Registry stores container images and language packages, integrates with IAM, and supports vulnerability scanning; it succeeds Container Registry.

  • Cloud Build runs each step in cloudbuild.yaml as a container; substitutions provide variables and triggers start builds from repository events.

  • terraform plan previews changes by comparing configuration with state and real resources; terraform apply creates, updates, or deletes resources to match configuration.

  • Pub/Sub publishers send messages to a topic, and each subscription receives a copy; default delivery is at-least-once and governed by the acknowledgment deadline.

  • Deployment Manager uses YAML, Jinja, or Python templates; a preview creates resources in a dry-run mode before an update is applied.

  • kubectl apply -f manifest.yaml declaratively creates or updates GKE resources, while kubectl get pods lists pods in the current namespace.

Quick reference

Ensuring successful operation of a cloud solution

  • Cloud Monitoring uptime checks probe endpoints from multiple locations; an alerting policy fires after a configured failure duration.

  • Log sinks route entries to Cloud Storage, BigQuery, Pub/Sub, or another project; inclusion and exclusion filters control which entries are routed.

  • Admin Activity and System Event audit logs are always enabled; Data Access audit logs are disabled by default and must be explicitly enabled for most services.

  • An SLO specifies a target and a rolling or calendar window; the error budget equals 1 minus the SLO target.

  • Error Reporting aggregates and groups application exceptions from logs; it can alert on new error groups but is not a general metrics store.

  • Billing export to BigQuery writes detailed cost and usage data into datasets; queries over these tables support cost analysis and chargeback.

  • Cloud Monitoring metrics are time series; dashboards and alerting policies use metric filters, aggregation, and alignment.

  • gcloud logging read queries entries with filters, and gcloud logging sinks create creates a sink at project, folder, or organization level.

  • The Ops Agent collects metrics and logs from Compute Engine VMs and replaces the legacy monitoring and logging agents.

Quick reference

Configuring access and security

  • An IAM allow policy binds members to roles on a resource; allow policies are additive and inherited down the hierarchy, while deny policies can override allows.

  • Basic roles Owner, Editor, and Viewer are broad and legacy; predefined roles are service-specific, and custom roles use an explicit permission list.

  • A service account is an identity for workloads; attaching a service account to a resource supplies credentials through the metadata server.

  • Service account keys are long-lived secrets; attached service accounts, Workload Identity Federation, and short-lived tokens avoid downloaded key files.

  • IAM conditions are CEL expressions on role bindings that limit when a grant applies, based on resource, request, or time attributes.

  • VPC Service Controls define a service perimeter that restricts access to supported services and helps reduce data exfiltration; they are not network firewalls.

  • Cloud KMS manages key rings and keys; CMEK lets supported services use customer-managed keys, and envelope encryption uses a DEK wrapped by a KEK.

  • Organization Policy constraints apply at organization, folder, or project level and restrict configurations such as allowed locations or public IP addresses.

  • Cloud Audit Logs record Admin Activity, Data Access, System Event, and Policy Denied events; Data Access logs usually require explicit enablement.

GCP ACE key terms

Resource hierarchy

The nested order of organization, folders, projects, and individual resources in Google Cloud, used to organize assets and inherit IAM policies and organization policies from parent to child.

Related: Project, Folder, Organization node
Project

The basic container in Google Cloud that holds resources, links to a billing account, and defines boundaries for API enablement, quotas, and most IAM allow policies.

Related: Resource hierarchy, Billing account, Quota
Folder

A grouping container inside an organization or another folder that lets administrators apply IAM policies and organization constraints to many projects at once.

Related: Organization node, Project
Organization node

The top of the Google Cloud resource hierarchy, created from a Cloud Identity or Google Workspace domain, giving central control over every folder and project beneath it.

Related: Folder, Resource hierarchy
Billing account

A Google Cloud entity holding payment details that pays for one or more linked projects; budgets and programmatic spend exports are configured against it.

Related: Project
Quota

A limit on how much of a resource or API call rate you can consume in a project or region, protecting shared capacity; some quotas can be raised by request.

Related: Project
Cloud Shell

A browser-based shell with gcloud, kubectl, and other tools preinstalled, plus a small persistent disk, giving command-line access without any local setup.

Related: gcloud CLI
gcloud CLI

The primary command-line tool for creating and managing Google Cloud resources, with commands grouped by service and support for output formatting and scripting.

Related: Cloud Shell
VPC network

A global virtual network that provides routing, firewall rules, and connectivity for regional subnets and the resources placed in them, whether on Google Cloud or connected on-premises.

Related: Subnet, Firewall rule
Subnet

A regional range of IP addresses within a VPC network; each subnet belongs to exactly one region, and VMs placed in it receive internal IP addresses automatically.

Related: VPC network
Firewall rule

A stateful VPC-level rule that permits or denies traffic to or from instances based on direction, source or destination, protocol, port, and target tags or service accounts.

Related: VPC network, Service account
Shared VPC

A configuration where one host project owns a VPC network and its subnets that other service projects attach to, centralizing network administration while letting teams manage their own resources.

Related: VPC network, IAM role
Cloud NAT

A managed service that lets instances with only private IP addresses reach the internet for outbound traffic, without allowing unsolicited inbound connections from outside.

Related: VPC network, Subnet
Cloud Load Balancing

A family of managed, software-defined load balancers, global or regional and covering HTTP(S), TCP, and UDP, that distribute traffic to backends with health checking and autoscaling integration.

Related: Managed instance group
Cloud Storage bucket

A globally uniquely named container for objects in Cloud Storage; its name, location, storage class, and IAM policies determine how data is stored and who can reach it.

Related: Storage class, Object lifecycle management
Storage class

A category applied to a bucket or object, such as Standard, Nearline, Coldline, or Archive, that trades storage price against retrieval cost and minimum storage duration.

Related: Object lifecycle management, Cloud Storage bucket
Object lifecycle management

A set of rules on a bucket that automatically changes an object's storage class or deletes it once conditions such as age, creation date, or version count are met.

Related: Storage class, Cloud Storage bucket
Committed use discount

A billing discount obtained by committing to a set amount of vCPU and memory usage for one or three years in a region, in exchange for a significantly lower price.

Related: Project, Quota
Instance template

A reusable definition of VM properties such as machine type, image, disks, network, metadata, and startup script, used to create managed instance groups and individual VMs consistently.

Related: Managed instance group
Managed instance group

A group of identical VMs created from an instance template that supports autoscaling, autohealing, rolling updates, and distribution across zones in a region.

Related: Instance template, Cloud Load Balancing
Cloud Build

A serverless CI/CD service that runs build steps in containers according to a configuration file, producing artifacts such as container images or deployable packages.

Related: Artifact Registry
Artifact Registry

A managed repository service for storing and scanning container images and language packages, with fine-grained IAM and regional repositories; it supersedes Container Registry.

Related: Cloud Build, Cloud Run
Cloud Run

A fully managed serverless platform that runs stateless containers that respond to HTTP requests or events, scaling down to zero instances when idle.

Related: Artifact Registry, Pub/Sub topic
Google Kubernetes Engine

A managed Kubernetes service where Google operates the control plane; you choose cluster mode, node pools, and versions while GKE handles upgrades and node repairs.

Related: Artifact Registry, Managed instance group
Pub/Sub topic

A named channel in Pub/Sub to which publishers send messages; subscriptions attached to the topic deliver copies to subscribers such as Cloud Run services or push endpoints.

Related: Cloud Run
Cloud Monitoring

A managed observability service that collects metrics, logs-based metrics, and uptime results, and powers dashboards, alerting policies, and notification channels.

Related: Alerting policy, Uptime check
Alerting policy

A rule in Cloud Monitoring that opens an incident and notifies configured channels when a metric or condition crosses a threshold for a defined duration.

Related: Cloud Monitoring, Uptime check
Log router sink

A configuration that routes matching log entries from Cloud Logging to destinations such as Cloud Storage, BigQuery, or Pub/Sub, using inclusion and exclusion filters.

Related: Admin Activity audit log, Cloud Monitoring
Uptime check

A Cloud Monitoring probe that regularly requests a public or private URL or TCP port from multiple locations to verify that a service is reachable and responding.

Related: Cloud Monitoring, Alerting policy
Admin Activity audit log

An always-on audit log recording administrative operations such as resource creation or IAM changes; it is retained for 400 days, cannot be disabled, and is visible to authorized viewers.

Related: Log router sink
Service level objective

A target level for a service level indicator, such as 99.9 percent of requests succeeding over 30 days; the error budget derives from how far actual performance sits from it.

Related: Cloud Monitoring, Alerting policy
IAM role

A named collection of permissions, either basic, predefined, or custom, that can be granted to a principal; individual permissions cannot be granted on their own.

Related: IAM policy binding, Custom role
IAM policy binding

An entry that grants a role to a member such as a user, group, service account, or domain on a resource; bindings merge into the allow policy and inherit downward.

Related: IAM role, Service account
Service account

A non-human identity used by workloads and VMs to authenticate to Google APIs, identified by an email address and authenticated with keys or attached credentials.

Related: IAM policy binding, IAM role
Custom role

A user-defined IAM role containing a chosen list of permissions, created at project or organization level so a workload receives exactly the access it needs.

Related: IAM role, Principle of least privilege
Principle of least privilege

The practice of granting identities only the permissions required for their task and nothing broader, limiting the impact of mistakes or compromised credentials.

Related: Custom role, IAM role
Organization policy constraint

A rule set at organization, folder, or project level that restricts how resources can be configured, such as blocking external IPs or limiting allowed locations, independent of IAM.

Related: Folder, Resource hierarchy
Cloud KMS

A managed service for creating and using cryptographic keys and key rings, supporting symmetric and asymmetric keys, rotation, and customer-managed encryption keys for other Google Cloud services.

Related: IAM role, Service account
VPC Service Controls

A security perimeter that restricts access to supported Google Cloud APIs from inside a defined boundary, helping prevent data exfiltration even if credentials are stolen.

Related: VPC network, Organization policy constraint

GCP ACE study plan

8 weeks · About 7 hours per week

Week 1

Setting up a cloud solution environment

  • Create a Google Cloud project under a folder in an organization and switch between them with gcloud config set project

  • Configure billing budgets and alerts on the new project and confirm the alert fires on a threshold breach

  • Assign predefined IAM roles to two test users and verify effective permissions with the IAM policy troubleshooter

  • Install and configure the Cloud SDK and Cloud Code, then authenticate with both user credentials and a service account

  • Request a quota increase for a compute region and identify which quotas are project-level versus region-level

Your checkpoint

A written one-page lab log showing the resource hierarchy you built (org, folder, two projects), the IAM bindings applied, and a screenshot-free text transcript of the gcloud commands that created them

Week 2

Setting up a cloud solution environment & Planning and configuring a cloud solution

  • Choose a compute option (Compute Engine, GKE, Cloud Run, App Engine) for three stated workloads and justify each choice in writing

  • Design a VPC with two subnets in different regions and document the CIDR ranges without overlap

  • Select storage classes and lifecycle rules for three data access patterns and estimate monthly cost for each

  • Size a Cloud SQL instance and choose between Cloud SQL, Spanner, and Firestore for two given schemas

  • Produce a cost estimate in the pricing calculator for a three-tier web app and export the result

Your checkpoint

A single architecture and cost brief covering all five decisions, with the pricing calculator export total within 15 percent of a hand calculation you show

Week 3

Planning and configuring a cloud solution & Deploying and implementing a cloud solution

  • Deploy a Compute Engine instance from a custom instance template using gcloud compute instances create-from-template

  • Create a managed instance group with an autoscaler tied to a CPU target and drive load until it scales past two instances

  • Deploy a containerized workload to Cloud Run and to a GKE Autopilot cluster, then expose both on HTTPS

  • Write a startup script and a metadata key that the instance reads at boot, and confirm the output in the serial console

  • Attach a persistent disk to an instance, format and mount it, and remount it after a stop and start

Your checkpoint

A repeatable deployment script checked into a repository that provisions the instance template, managed instance group, and Cloud Run service from an empty project with one command

Week 4

Planning and configuring a cloud solution & Deploying and implementing a cloud solution

  • Create Cloud Storage buckets with uniform bucket-level access, versioning, and a lifecycle rule that deletes objects after 30 days

  • Load a CSV into BigQuery with a defined schema and run three queries including one that joins two tables

  • Deploy a Cloud SQL for PostgreSQL instance, connect from Compute Engine over the private IP, and import a dump

  • Publish and consume messages on a Pub/Sub topic with a pull subscription and verify message retention

  • Configure a Cloud Storage bucket as a static website origin behind Cloud CDN

Your checkpoint

A data pipeline note demonstrating one object landing in Cloud Storage, being loaded to BigQuery, and being queried, with the query results and row counts pasted inline

Week 5

Deploying and implementing a cloud solution & Configuring access and security

  • Create a custom-mode VPC with two subnets, firewall rules scoped by service account, and a Cloud NAT gateway for egress

  • Deploy an external HTTP(S) load balancer with a managed certificate across two instance groups in different regions

  • Restrict a Cloud Storage bucket to a single service account using IAM conditions and prove that a second account is denied

  • Create a service account, grant it least-privilege roles, and attach it to an instance instead of using a downloaded key

  • Apply an organization policy that disables service account key creation and confirm a key creation attempt fails

Your checkpoint

A network and access diagram in text form paired with a test matrix showing each firewall rule and IAM binding you validated, including at least two expected-deny cases that actually denied

Week 6

Ensuring successful operation of a cloud solution & Configuring access and security

  • Create a Cloud Monitoring dashboard with CPU, latency, and error-rate charts for the load-balanced service from week 5

  • Define alerting policies with notification channels for uptime checks and error-rate thresholds

  • Write three Cloud Logging queries using filters and severity, then export a log sink to Cloud Storage

  • Perform a rolling update on the managed instance group with a surge setting and roll it back

  • Enable audit logs and KMS-encrypted disks, then verify the encryption key usage in Cloud Audit Logs

Your checkpoint

An operations runbook listing the alert policies, the log-based queries you saved, and the exact commands for the rolling update and rollback, each with the observed outcome recorded

Week 7

Configuring access and security & Ensuring successful operation of a cloud solution

  • Build a custom IAM role with only the permissions needed for a read-only analyst and test it against two denied actions

  • Configure Workload Identity Federation so an external identity can impersonate a service account without a key

  • Store an application secret in Secret Manager and retrieve it from a Cloud Run service via the runtime service account

  • Back up a Compute Engine disk with a snapshot schedule and restore it into a new disk

  • Diagnose a simulated failure where a service account lacks storage.objects.get and document the fix

Your checkpoint

A security hardening checklist completed against the week 5 environment, with each item marked pass or fail and the remediation command for every fail

Week 8

Setting up a cloud solution environment & Planning and configuring a cloud solution & Deploying and implementing a cloud solution & Ensuring successful operation of a cloud solution & Configuring access and security

  • Take three full-length timed practice exams and log every missed question by domain

  • Re-teach the two lowest-scoring domains by rebuilding the specific labs those questions tested

  • Rewrite your one-page cheat sheet for gcloud commands, IAM role names, and pricing decision rules from memory

  • Re-run the week 3 deployment script end to end from an empty project to prove recall under time pressure

Your checkpoint

A scored practice-exam log across all three attempts with per-domain percentages, plus a final cheat sheet no longer than two pages, showing every domain at or above 80 percent on the last attempt

Milestones

Week 2: You can create a project, wire billing alerts, apply IAM roles, and justify a compute and storage choice for a new workload without looking anything up

Week 4: One command or script stands up compute, storage, and a database from an empty project, and you have queried data that flowed through Pub/Sub into BigQuery

Week 6: Your deployed service has a dashboard, at least two firing alert policies, an exported log sink, and a successful rolling update with rollback

Week 8: Three timed practice exams completed with the final attempt at or above 80 percent in every domain and no domain below 70 percent on any attempt

GCP ACE exam day

Your checklist

Government-issued photo ID

Bring the original physical ID you registered with (passport, driver's license, or national ID). The name on it must match your registration name. Confirm which ID forms your provider accepts before test day; photocopies and digital copies are generally not accepted.

Appointment confirmation and login details

Have your confirmation email, confirmation/registration number, and the provider's login credentials accessible on the morning of the exam. Double-check the start time in your local time zone, not the provider's.

Completed system/pre-flight test (online proctored)

If testing remotely, run the provider's system check ahead of time on the exact machine and network you will use: webcam, microphone, browser, and bandwidth. Re-run it the day before in case of OS or browser updates.

Private, quiet room with a clear desk

Remove papers, books, pens, sticky notes, phones, smartwatches, and any second screen from the desk and the area around you. Walls should be reasonably bare of study material. Expect a room/desk scan at the start.

Clean computer state

Close every application except what the provider requires, quit messaging and email clients, silence notifications and alarms, and disconnect or disable additional monitors. Unplug unnecessary USB devices.

Reliable internet with a backup

Prefer wired ethernet or a strong Wi-Fi signal. Have a phone hotspot or a second connection ready as a fallback, and keep the provider's support number handy in case of a disconnect.

Power and battery

Plug the laptop into power and confirm the charger works. If on battery, make sure it comfortably exceeds the exam length.

Approved scratch/whiteboard method

Scratch paper and physical whiteboards are often restricted and some providers supply an on-screen whiteboard instead. Confirm the allowed method for your delivery format with the provider before the exam day rather than assuming.

Test-center logistics (if testing in person)

Confirm the address, parking or transit options, and arrival buffer. Expect a locker or secure area for personal belongings, since most items stay outside the testing room.

Uninterrupted block of time and personal readiness

Use the restroom and drink water before starting. Tell housemates or family not to enter the room, and put pets where they will not appear on camera or make noise.

Know the exam length and question count in advance

Look up the current duration and number of questions on the official certification page when you book, so you can plan your pacing. Do not rely on secondhand numbers.

Provider support contact and incident process

Save the proctor/support phone number or chat link. If something breaks, you want to reach a human immediately and get a case number rather than troubleshooting alone.

Plan the day

72 hours before

Stop adding new material. Take one timed practice set under exam conditions to calibrate pacing, then review only the explanations you got wrong. Re-read the official exam guide domain weightings so your last hours go to the heaviest domains (for this exam that usually means IAM/security, networking, and deploying/managing workloads). Confirm appointment time, time zone, and delivery format; validate your ID; run the provider's system test if testing remotely.

24 hours before

Do a short, light review of a one-page personal cheat sheet: gcloud/gsutil/kubectl command patterns, role-selection logic, when to choose Cloud Run vs GKE vs Compute Engine, and load-balancer/network choice rules. Do not start a new course or lab series. Prepare your room and desk, lay out your ID, charge devices, confirm internet and backup, set two alarms, and get a full night of sleep.

Morning of

Eat something light, avoid over-caffeinating, use the restroom, and clear the room again. Log in 15-30 minutes early to absorb any check-in delays. Keep your ID beside you, silence your phone and put it out of reach, and close background apps one more time. If at a test center, arrive with a comfortable buffer before your slot.

First 5 minutes of the test

Read the on-screen instructions and note the timer behavior. Write down (if an approved scratch method exists) your pacing target: divide total minutes by question count, then round slightly down to leave review time. Skim the flag/review controls so you do not waste time learning the interface mid-question.

During the test

Answer in passes: work steadily, flag anything that takes more than roughly one pacing unit, and move on. Read the full question, then eliminate clearly wrong options before choosing. Watch for qualifiers such as most cost-effective, least operational overhead, most secure, or highest availability, since they decide between two technically valid answers. Answer every question; there is no value in blanks. Reserve the final minutes to revisit flagged items and confirm nothing is unanswered.

After submitting

Follow whatever result-delivery method your provider uses and note any confirmation or case number. If you hit a technical failure at any point, report it to the proctor immediately rather than after the fact, and record the incident details in case you need to open a ticket.

Rules to check

  • Your ID must match the name on your registration exactly. If it does not, or if the ID is expired or a copy, you may be denied entry or your session may be terminated. Confirm the accepted ID list with the provider in advance.

  • No unauthorized materials in reach: notes, printed pages, phones, smartwatches, earbuds, second monitors, or any study aids. A violation can end the session and void your result, regardless of intent.

  • Stay inside the permitted camera and testing environment for the whole session. Leaving the view, letting another person enter, or talking to yourself aloud in a way the proctor flags can lead to termination. Confirm the provider's specific rules on movement, breaks, and water before you begin.

  • Scheduled breaks and whether the timer pauses during them vary by delivery format and provider. Ask before the exam starts, and assume unscheduled breaks are not compensated with extra time unless you are told otherwise.

  • Do not attempt to photograph, screen-record, copy, or verbally relay exam content. This breaches the exam agreement and is grounds for score invalidation and account sanctions.

  • Expect a pass/fail outcome rather than a numeric score unless the provider states otherwise, and confirm when and where results are released. Do not count on an instant score at the end of the session.

  • If the connection drops or the platform misbehaves, stop and contact the proctor or support line immediately, follow their instructions, and keep the case reference. Reconstructing what happened days later is much harder.

  • Retake eligibility, waiting periods, and fees are set by the provider and change over time. Check the current policy on your certification page rather than relying on what a colleague told you.

GCP ACE FAQ

How many GCP ACE practice questions are available?

The practice set has 200 GCP ACE questions. Each one includes the choices, the correct answer, and an explanation, and the set is free to read with no signup.

How many GCP ACE flashcards are available?

The flashcard set has 32 GCP ACE cards. Each card shows the term and its definition.

Which topics does the GCP ACE study material cover?

The GCP ACE study material has 5 sections, including Setting Up a Cloud Solution Environment, Planning and Configuring a Cloud Solution, and Deploying and Implementing a Cloud Solution.

What format facts does the catalog list for the GCP ACE exam?

The catalog lists a duration of 120 minutes and these item types: multiple choice and multiple select for the GCP ACE exam.