Deploying and Implementing a Cloud Solution practice questions

Answer: A) Storage Transfer Service

Storage Transfer Service is Google Cloud's managed, schedulable, agentless service for online transfers from other clouds such as Amazon S3 into Cloud Storage. Transfer Appliance is for offline transfers, rsync scripts are self-managed, and BigQuery Data Transfer Service loads data into BigQuery.

• ACorrect. Storage Transfer Service is a managed service for moving data from Amazon S3, Azure Blob Storage, and other sources into Cloud Storage. Transfers from S3 need no agents, can be scheduled, and remove the need for manual processes and custom scripts.
• BIncorrect. Transfer Appliance is a physical device that is shipped to and from your site for offline transfers, typically when network bandwidth is limited. The company has ample bandwidth and wants to avoid shipping hardware.
• CIncorrect. gcloud storage rsync can copy data, but the team would have to script, schedule, monitor, and retry the jobs on infrastructure it runs. That is the self-managed approach the company wants to avoid.
• DIncorrect. BigQuery Data Transfer Service loads data into BigQuery datasets on a schedule. It is not designed for bulk object migration into Cloud Storage buckets.

Answer: A) CREATE MATERIALIZED VIEW ds.mv OPTIONS (enable_refresh = true, refresh_interval_minutes = 60) AS SELECT ...

BigQuery materialized views refresh automatically when enable_refresh is true. The refresh_interval_minutes option sets the frequency cap, which defaults to 30 minutes.

• ACorrect. enable_refresh = true keeps automatic refresh on. refresh_interval_minutes = 60 caps the refresh frequency at once per hour, up from the default of 30 minutes. The allowed range is 1 minute to 7 days.
• BIncorrect. Setting enable_refresh to false turns off automatic refresh, so the view would only be refreshed manually. That doesn't meet the requirement.
• CIncorrect. A logical view stores no precomputed results and re-runs its query each time it's referenced. Refresh options apply only to materialized views.
• DIncorrect. Without refresh_interval_minutes, automatic refresh uses the default 30-minute cap, so it can run more often than once an hour. Adding a scheduled query doesn't change the view's refresh frequency cap.

Answer: A) The instance edition: Developer, Basic, or Enterprise

You choose a Cloud Data Fusion edition (Developer, Basic, or Enterprise) when you create the instance. The edition sets the instance's hourly price and its features, such as RBAC in Enterprise.

• ACorrect. The edition is chosen when the instance is created (gcloud --edition=developer|basic|enterprise). It sets the hourly instance price and the feature set. For example, Enterprise is meant for large, mission-critical pipelines and includes RBAC.
• BIncorrect. You do pick a region for each instance, but the region only sets where the instance runs. It does not decide the feature tier or whether RBAC is available.
• CIncorrect. A private instance controls how the instance connects to your VPC network. It does not set the edition's price or features.
• DIncorrect. Compute profiles set the Dataproc clusters that run pipelines, which affects Dataproc costs. They are not the instance-level setting that decides the instance's hourly price or RBAC availability.

Answer: A) The region, the machine series, the amounts of vCPUs and memory, and a 1-year or 3-year term

A resource-based CUD commits to quantities of resources, such as vCPUs and memory, for one machine series in one region, for a 1-year or 3-year term.

• ACorrect. Resource-based commitments are bought per region. They apply to a specific machine series in that region, commit to minimum amounts of resources such as vCPUs and memory, and last 1 or 3 years.
• BIncorrect. Commitments are bought per region, not per zone, and aren't tied to specific VM instances. The discount applies to eligible usage of that machine series anywhere in the region.
• CIncorrect. Resource-based commitments can't be global; they're bought per region. The term is 1 or 3 years, not a monthly renewal, even though billing happens monthly.
• DIncorrect. A dollar spend commitment describes a spend-based CUD. A resource-based CUD commits to quantities of resources such as vCPUs and memory.

Answer: A) Create a Serverless VPC Access connector in the VPC and configure the function to use it

A serverless function needs VPC egress to reach private IP resources. A Serverless VPC Access connector provides that, and Direct VPC egress is the newer, recommended alternative. Cloud NAT, authorized networks, and Private Google Access don't.

• ACorrect. A Serverless VPC Access connector lets serverless workloads send traffic into a VPC network, so the function can reach resources with internal IP addresses, such as the Cloud SQL private IP. Direct VPC egress is a newer alternative that Google now recommends, but it isn't one of the options here.
• BIncorrect. Cloud NAT translates outbound traffic from resources inside the VPC to reach the internet. It gives the serverless function no path into the VPC's internal addresses.
• CIncorrect. Authorized networks control access to an instance's public IP address. This instance has only a private IP, so the setting gives the function no path to it.
• DIncorrect. Private Google Access lets VMs without external IPs reach Google APIs and services. It doesn't connect a serverless function to internal IP addresses in the VPC.

Answer: A) Standard environment with automatic scaling

The App Engine standard environment with automatic scaling provides Google-managed language runtimes and can scale to zero. The flexible environment always keeps at least one instance running, and manual scaling uses a fixed instance count.

• AThe standard environment runs code in Google-provided runtimes for supported language versions. With automatic scaling, it scales quickly with traffic and can scale down to zero instances when idle.
• BWith manual scaling, the team sets a fixed number of instances that keep running regardless of load, so the app doesn't scale automatically with traffic or down to zero.
• CFlexible environment instances run in Docker containers on Compute Engine VMs and use the Compute Engine autoscaler. Each active version needs at least one instance, so the app can't scale to zero.
• DA custom runtime makes the team responsible for defining the runtime in a Dockerfile, which they want to avoid. It also runs on the flexible environment, which keeps at least one instance running.

Answer: A) A runtime environment variable in the function's deployment configuration, read by the code when it runs

Use runtime environment variables in the deployment configuration to give a Cloud Run function settings that differ by environment without code changes. A change to a variable is applied by deploying a new revision. Secrets belong in Secret Manager.

• ACorrect. Runtime environment variables are part of the function's deployment configuration and the code reads them from its environment. Each environment can set its own value with the same code. Changing the value still means deploying a new revision, but the source code stays the same. Sensitive values like API keys should come from Secret Manager instead.
• BIncorrect. Labels are key-value metadata for organizing resources and breaking down costs. Their values are not passed into the function's runtime environment. Label values also have character limits that don't allow full URLs.
• CIncorrect. Build-time environment variables only exist while the container image is being built. They are not available to the function at runtime, which is when this function needs the value.
• DIncorrect. A Scheduler payload only reaches invocations that Scheduler triggers. It is not a configuration mechanism, and invocations from other sources would not get the value.

Answer: A) Enable PSC on the instance with the consumer project in its allowed projects list, then create a PSC endpoint in that VPC targeting the instance's service attachment

With Cloud SQL PSC, you enable PSC on the instance, list the allowed consumer projects, and create an endpoint (internal IP plus forwarding rule) in the consumer VPC that targets the instance's service attachment. VPC peering and an allocated IP range are not needed.

• ACorrect. A PSC-enabled Cloud SQL instance publishes a service attachment (pscServiceAttachmentLink). You can only create endpoints from projects in the instance's allowed consumer projects list. The endpoint is an internal IP address plus a forwarding rule in the consumer VPC that points to the service attachment. No peering is needed.
• BIncorrect. This is private services access, which gives the instance a private IP through VPC Network Peering with Google's service producer network. That is the peering-based approach the team wants to avoid, not PSC.
• CIncorrect. A Serverless VPC Access connector lets serverless services such as Cloud Run reach resources in a VPC. It is not attached to Cloud SQL instances and does not create PSC connectivity for VMs.
• DIncorrect. This sends traffic to the instance's public IP. It doesn't give private connectivity, and it doesn't use Private Service Connect.

Answer: A) Infrastructure Manager, which deploys Terraform configurations

Infrastructure Manager is Google Cloud's managed infrastructure-as-code service, built on Terraform, with deployments, revisions, and previews. Deployment Manager reached end of support on March 31, 2026, and Google recommends migrating to Infrastructure Manager.

• ACorrect. Infrastructure Manager is a managed service that deploys and manages Google Cloud resources from Terraform configurations. It records each update as a revision, and its previews run a Terraform plan so you can see changes before you apply them.
• BIncorrect. Deployment Manager is deprecated. Support ended on March 31, 2026, it gets no new features, and Google tells customers to migrate to Infrastructure Manager. A new project should not adopt it.
• CIncorrect. Cloud Build is a CI/CD service that runs build steps. It does not track infrastructure deployments, revisions, or change previews by itself.
• DIncorrect. Cloud Asset Inventory lets you view, search, and export metadata and history for existing resources. It does not deploy or update infrastructure from configuration files.

Answer: A) Set the machine type, and enable cluster autoscaler on the pool with --min-nodes and --max-nodes values

Cluster autoscaler, enabled per node pool with minimum and maximum node counts, automatically changes the number of nodes based on demand. The HPA scales Pods rather than nodes, auto-repair replaces unhealthy nodes, and a fixed node count needs manual resizing.

• ACorrect. Cluster autoscaler resizes a node pool based on workload demand. It adds nodes up to the maximum when Pods can't be scheduled and removes underused nodes down to the minimum. The machine type is set when the pool is created.
• BIncorrect. The Horizontal Pod Autoscaler changes the number of Pod replicas in a workload. It does not add or remove nodes, so Pods it creates can stay unscheduled when the nodes are full.
• CIncorrect. Node auto-repair recreates unhealthy nodes to keep the pool healthy. It does not change the number of nodes based on demand.
• DIncorrect. A fixed node count with manual resizing does not scale automatically, so it fails the requirement.