PCD Designing highly scalable, available, and reliable cloud-native applications Practice Test 6 — 15 Questions

Question 1

A company runs a real-time inventory service on Cloud Run that receives bursts of traffic from an upstream system. The service writes to Cloud SQL for PostgreSQL. During traffic spikes, the service experiences connection errors because the database reaches its maximum connection limit. The company wants to improve reliability without changing the application code. Which approach should they take?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Enable the Cloud SQL Auth Proxy sidecar with connection pooling, or use a connection pooler such as PgBouncer, and configure Cloud Run to connect through it.. Cloud Run scales horizontally by creating many instances, each of which may open a database connection. A connection pooler such as PgBouncer multiplexes those connections onto a limited set of backend connections, preventing the Cloud SQL instance from hitting its max_connections limit. This approach requires no application code changes because the application still connects to a PostgreSQL endpoint, just through the pooler.

Answer

Configure the Cloud Run service to use a serverless VPC access connector and connect directly to the Cloud SQL private IP address.

Answer

Deploy the Cloud SQL Auth Proxy as a sidecar in each Cloud Run instance and configure the application to connect through it.

Answer

Increase the maximum number of connections on the Cloud SQL instance by raising the max_connections database flag.

Question 2

A startup deploys a stateless containerized web front end on Cloud Run. They want the service to remain available even if an entire zone in the region becomes unavailable, and they want to avoid managing servers. Which Cloud Run configuration should they use?

Accepted Answer

Deploy the Cloud Run service to multiple regions behind a global external Application Load Balancer with a serverless network endpoint group backend.. Cloud Run is regional and automatically spreads instances across zones, so within a region it already tolerates a zone failure. To survive a full regional outage, the service must be deployed in multiple regions and fronted by a global external Application Load Balancer using serverless network endpoint group backends. That combination keeps the architecture serverless while providing the desired availability.

Answer

Deploy the service to a single zone by pinning the Cloud Run service to a specific zone using a resource label.

Answer

Deploy the Cloud Run service with CPU always allocated and set the maximum instance count to a high value.

Answer

Deploy the Cloud Run service in a region and set the minimum instance count to zero to allow multi-zone placement.

Question 3

A large media company is designing a video transcoding pipeline on Google Cloud. The pipeline processes video files uploaded to a Cloud Storage bucket. Each video file must be transcoded into multiple formats, and the process can take several minutes per file. The company expects thousands of files to be uploaded per hour, with occasional bursts of tens of thousands. They want to ensure the pipeline is highly scalable, reliable, and can recover from failures without losing work. The transcoding logic is packaged as a container. Which architecture should they use?

Accepted Answer

Use a Cloud Storage trigger to publish a message to Pub/Sub, and have a GKE Autopilot cluster with a Horizontal Pod Autoscaler process the messages.. The architecture should decouple file upload from processing to handle bursts and ensure reliability. Pub/Sub provides a durable queue that can buffer messages during spikes and allow retries on failure. GKE Autopilot offers a managed Kubernetes environment where pods can run the transcoding container and scale based on workload. The Horizontal Pod Autoscaler can scale pods using Pub/Sub queue depth as a custom metric, ensuring efficient resource use. This combination is scalable, reliable, and suitable for long-running tasks.

Answer

Use a Cloud Storage trigger to invoke a Cloud Function that runs the transcoding container with a 60-minute timeout.

Answer

Use a Cloud Storage trigger to publish a message to Pub/Sub for each uploaded file, and have a Cloud Run service process the messages.

Answer

Use a Cloud Function triggered by Cloud Storage object finalize events to start a Dataflow job for each file.

Question 4

A financial services company runs a critical transaction processing system on Google Kubernetes Engine (GKE) using a regional cluster. The application stores data in a Cloud SQL for PostgreSQL instance with a single zone. The compliance team requires an RPO of 5 minutes and an RTO of 15 minutes in case of a regional failure. The application must remain available if a single zone fails. What should the team do to meet these requirements?

Accepted Answer

Configure Cloud SQL high availability with a standby instance in a different zone, and enable GKE regional cluster with nodes spread across multiple zones.. Cloud SQL high availability with a standby in a different zone provides automatic failover and synchronous replication, ensuring minimal data loss (RPO) and quick recovery (RTO). A GKE regional cluster spreads nodes across multiple zones, so the application continues to run if one zone fails. Together, these configurations meet the compliance requirements for zone-level failure without application changes.

Answer

Set up Cloud SQL read replicas in multiple regions and configure the application to use them for failover.

Answer

Use a multi-region GKE cluster with a global load balancer and store data in Cloud Spanner.

Answer

Enable Cloud SQL point-in-time recovery and perform regular backups to a multi-region bucket.

Question 5

A healthcare company is building a patient portal on Google Cloud. The portal must be highly available and able to handle sudden increases in traffic. The company wants to use a managed database that scales automatically and provides strong consistency. Which database should they choose?

Accepted Answer

Cloud Spanner with a multi-region instance configuration.. Cloud Spanner multi-region instances provide strong consistency, automatic scaling, and high availability across regions. This meets all requirements: managed service, automatic scaling, strong consistency, and resilience to regional failures. Other options either lack strong consistency, automatic scaling, or multi-region availability.

Answer

Cloud Bigtable with a multi-region cluster.

Answer

Cloud SQL for MySQL with high availability enabled and a read replica for scaling reads.

Answer

Cloud Spanner with a regional instance configuration.

Question 6

A media company streams video-on-demand content to a global audience using Cloud Storage buckets. Viewers in Asia report slow startup times and buffering, while viewers in North America have no issues. The buckets are in the us-central1 region. The company wants to improve global read performance and availability without changing the application's storage API calls. Which TWO actions should they take? (Choose two.)

Accepted Answer

Enable Cloud CDN on an external Application Load Balancer with a backend bucket pointing to the Cloud Storage bucket.. Multi-region buckets replicate objects across regions so reads are served from a nearby location, improving latency and availability. Cloud CDN caches content at edge points of presence, which is especially effective for video segments that many viewers request. Both approaches preserve the existing Cloud Storage API, so the application does not need modification.

Answer

Enable uniform bucket-level access and set IAM conditions to allow reads only from specific regions.

Answer

Create a Cloud Storage transfer job to copy objects to buckets in each continent and update the application to select the nearest bucket.

Answer

Set the bucket's default storage class to Nearline and enable Object Lifecycle Management to move older videos to Coldline.

Question 7

A logistics company runs an order-processing pipeline on Google Cloud. Orders arrive in a Pub/Sub topic and are processed by a Cloud Run service that calls a third-party shipping API. The team observes duplicate shipments during traffic spikes and wants to design for at-least-once delivery without creating duplicate side effects in the shipping system. Which TWO design changes should they implement? (Choose two.)

Accepted Answer

Make the Cloud Run handler idempotent by recording each processed order ID in Firestore and skipping orders already marked complete.. At-least-once delivery means duplicates are expected, so the application must make its side effects safe to repeat. Tracking processed order IDs in Firestore prevents a second shipment when the same message is redelivered, and a longer acknowledgment deadline reduces redeliveries caused by slow third-party calls. Together they close the duplicate-shipment gap while preserving the delivery guarantee.

Answer

Increase the Cloud Run maximum instance count so messages are processed faster and duplicates cannot occur.

Answer

Enable exactly-once delivery on the Pub/Sub subscription and rely on it to prevent all duplicate processing.

Answer

Switch the subscription to pull delivery and have each Cloud Run instance pull messages in a tight loop.

Question 8

A retail company runs a web application on Compute Engine instances behind an HTTP(S) Load Balancer. The application experiences sudden spikes in traffic. The operations team wants to ensure that the application remains available and responsive during these spikes without manual intervention. They also want to minimize cost during low-traffic periods. What should they do?

Accepted Answer

Use a managed instance group with autoscaling based on HTTP load balancing utilization, and set a minimum of 2 instances across multiple zones.. Autoscaling based on HTTP load balancing utilization directly responds to the traffic handled by the load balancer, making it ideal for web applications with variable load. A minimum of 2 instances across multiple zones ensures high availability and baseline capacity. This setup automatically scales out during spikes and scales in during low traffic, balancing performance and cost.

Answer

Configure the managed instance group to use autoscaling based on CPU utilization, and set a minimum number of instances to 1 and a maximum to 10.

Answer

Deploy the application on a single large Compute Engine instance with a preemptible VM to reduce cost, and use a network load balancer.

Answer

Configure a managed instance group with autoscaling based on Cloud Monitoring custom metrics that track the number of active users, and set a minimum of 1 instance.

Question 9

A fintech company is designing a payment processing system on Google Cloud. The system uses Pub/Sub to ingest transactions, Cloud Run services to process them, and Firestore in Native mode to store account balances. During a regional outage, the team must continue processing with an acceptable recovery point objective. Which TWO design changes should they implement? (Choose two.)

Accepted Answer

Create a second Pub/Sub topic in another region and configure the publishers to write to both topics, then have regional Cloud Run consumers subscribe to their local topic.. To keep processing payments through a regional outage, both the messaging and the data layers must have a live presence in a second region. Publishing to topics in two regions with local consumers maintains ingestion, while a second Native mode Firestore database with dual writes and reconciliation preserves balances. Together these reduce the recovery point objective to the replication or dual-write interval.

Answer

Set the Cloud Run services to use a minimum instance count of 10 in each region to keep warm capacity ready.

Answer

Configure Firestore in Datastore mode with a multi-region location and rely on its automatic replication for balance writes.

Answer

Increase the Pub/Sub message retention duration to 31 days so messages are preserved during the outage.

Question 10

A company is designing a new application that must store user profiles and support millions of reads per second with single-digit millisecond latency at global scale. The data model is key-value with no complex joins, and the application must remain available if a single region fails. Which Google Cloud service should they use?

Accepted Answer

Firestore in Native mode with a multi-region location.. Firestore in Native mode is a fully managed, serverless NoSQL document database that automatically replicates data across multiple regions when configured with a multi-region location. It provides strong consistency, low-latency reads and writes, and scales seamlessly to millions of users. Its document model fits user profiles well, and the multi-region configuration satisfies the requirement to survive a single-region failure without manual intervention.

Answer

Cloud SQL for PostgreSQL with a read replica in a second region.

Answer

Cloud Bigtable with a multi-cluster routing policy and replication across regions.

Answer

Cloud Spanner with a regional configuration and automatic failover.

Question 11

A retail company uses Pub/Sub to ingest orders from thousands of stores. The orders are processed by a set of subscriber applications running on Compute Engine instances in a managed instance group. The team notices that during peak hours, some messages are processed multiple times, causing duplicate orders. They want to ensure exactly-once processing without modifying the subscriber code significantly. What should they do?

Accepted Answer

Configure the subscription to use exactly-once delivery and ensure the subscriber acknowledges messages using the ack ID.. Pub/Sub exactly-once delivery prevents duplicate message delivery after a successful acknowledgment. By configuring the subscription for exactly-once delivery and ensuring the subscriber acknowledges using the ack ID, the system avoids processing the same message multiple times. This meets the requirement without significant code changes, as the subscriber only needs to handle the ack ID properly.

Answer

Enable message ordering on the Pub/Sub topic and use the same ordering key for all orders.

Answer

Create a separate subscription for each store and use a dead-letter topic to handle duplicates.

Answer

Use Pub/Sub exactly-once delivery and set the ack deadline to a high value.

Question 12

A healthcare company runs a patient portal on a regional Google Kubernetes Engine (GKE) cluster in europe-west1. The portal stores session data in a Memorystore for Redis instance and static assets in a Cloud Storage bucket. The compliance team requires the portal to keep serving reads if the europe-west1 zone hosting the Redis primary fails, with no manual intervention. Writes and sessions must remain consistent. Which change should the team make?

Accepted Answer

Replace Memorystore for Redis with a Memorystore for Redis Cluster and configure the application to use the cluster endpoint with automatic failover.. The portal needs session reads and writes to continue without manual action when a zone fails. Memorystore for Redis Cluster spreads shards across zones and promotes a new primary automatically, and the application uses one cluster endpoint. This provides the automatic failover and consistency the compliance team requires, unlike a single-zone or self-managed design.

Answer

Configure Memorystore for Redis Standard tier with a replica in another zone and let the service promote the replica automatically on primary failure.

Answer

Move session data into the GKE cluster as a StatefulSet with a PersistentVolume and use a headless Service for discovery.

Answer

Keep Memorystore for Redis in Basic tier and add a read replica in a second zone, updating the application to read from the replica during an outage.

Question 13

A company is designing a new microservices application on Google Cloud. They need to ensure that the application can handle a regional failure with minimal data loss and automatic failover. The application uses Cloud SQL for PostgreSQL as its database. They want a solution that provides a recovery point objective (RPO) of less than 1 minute and a recovery time objective (RTO) of less than 5 minutes. What should they do?

Accepted Answer

Use Cloud SQL for PostgreSQL with a cross-region replica, and configure the application to connect to the primary instance's IP. In the event of a regional failure, use a Cloud Function triggered by Cloud Monitoring alert to promote the replica and update the connection string in Secret Manager.. A cross-region replica provides asynchronous replication to another region, enabling a low RPO. Automating the promotion and connection string update using Cloud Monitoring and Cloud Functions ensures a low RTO without manual intervention. This design provides automatic failover for the database, meeting the RPO and RTO requirements. The application must be designed to fetch the connection string from Secret Manager dynamically.

Answer

Configure a Cloud SQL for PostgreSQL instance with high availability (HA) and enable point-in-time recovery (PITR).

Answer

Deploy Cloud SQL for PostgreSQL with a cross-region replica, and use a global external HTTP(S) load balancer with a failover backend service that automatically redirects traffic to the replica upon regional failure.

Answer

Use a Cloud SQL for PostgreSQL instance with a cross-region replica, and implement a manual failover process using a script that promotes the replica.

Question 14

A logistics company uses a Cloud Run service to process shipment tracking updates. The service writes to Cloud Firestore and publishes events to Pub/Sub. During a sudden surge in traffic, some requests fail with HTTP 429 errors, and the service scales out slowly. The team wants to handle the surge gracefully without dropping requests and without manual intervention. What should they do?

Accepted Answer

Increase the maximum number of instances for the Cloud Run service and set a higher concurrency value.. Increasing the maximum number of instances allows Cloud Run to scale out further to handle the traffic surge. Raising the concurrency value lets each instance process more requests simultaneously, reducing the total number of instances required and improving scaling efficiency. Together, these settings enable the service to handle the surge gracefully without dropping requests, and they are simple configuration changes.

Answer

Configure a Cloud Run serverless VPC connector and route traffic through a NAT gateway with higher throughput.

Answer

Deploy the Cloud Run service to multiple regions and use a global load balancer to distribute traffic.

Answer

Enable Cloud Run's CPU always allocated and set the minimum instance count to a high value.

Question 15

A company is designing a cloud-native application on Google Cloud that requires low-latency access to a global user base. The application serves static content and dynamic APIs. Which strategy best minimizes latency while maintaining high availability?

Accepted Answer

Use Cloud CDN to cache static content and deploy the API across multiple regions with global load balancing.. It combines Cloud CDN for caching static content at edge locations worldwide, reducing latency for static assets, and deploys the dynamic API across multiple regions with global load balancing (using Google Cloud's global external HTTP(S) load balancer) to route users to the nearest healthy backend, minimizing latency for dynamic requests while ensuring high availability through regional redundancy.

Answer

Deploy the application in a single region and use Cloud Interconnect for global access.

Answer

Use Cloud Armor to protect the application and rely on Google's backbone for low latency.

Answer

Store all content in Cloud Storage and serve directly from there.