PCD Designing highly scalable, available, and reliable cloud-native applications Practice Test 13 — 15 Questions

Question 1

A startup is deploying a new web application on Compute Engine. They expect variable traffic and want to automatically adjust the number of VM instances based on demand while keeping costs low. They also want to ensure that the application remains available if a single VM fails. What should they do?

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

Accepted Answer

Create a managed instance group with autoscaling enabled and configure a health check to automatically replace unhealthy instances.. A managed instance group with autoscaling and health checks automatically adjusts capacity based on demand and replaces unhealthy instances, ensuring both scalability and high availability. This is the most cost-effective and resilient solution for variable traffic.

Answer

Create a managed instance group without autoscaling and set the size to a fixed number of instances that can handle peak load.

Answer

Create an unmanaged instance group and manually add or remove instances based on traffic patterns.

Answer

Create a single Compute Engine instance with a static external IP address and use a startup script to install the application.

Question 2

A development team deploys a containerized web application to Cloud Run. Users in Europe and Asia report high latency, although the service is healthy. The team wants the lowest possible latency for a global audience without managing clusters or rewriting the application. What should they do?

Accepted Answer

Deploy the Cloud Run service to multiple regions and place a global external Application Load Balancer in front of it.. Placing a global external Application Load Balancer in front of Cloud Run services deployed in multiple regions combines anycast edge termination with regional execution. Users connect to the nearest Google edge and are routed to the closest healthy backend, cutting round-trip latency. This needs no cluster management and no application rewrite.

Answer

Deploy the Cloud Run service with the --region flag set to a single region close to the majority of users.

Answer

Enable Cloud CDN on the Cloud Run service and configure a cache TTL for all responses.

Answer

Deploy the container to a GKE Autopilot cluster with a multi-zone node pool and a global external load balancer.

Question 3

A company runs a stateful web application on a regional managed instance group (MIG) of Compute Engine VMs. The application stores session data on each VM's local SSD and writes user uploads to a zonal persistent disk. The company wants to make the application highly available across zones within a region and survive a single-zone failure without data loss. Which two changes should they make? (Choose two.)

Accepted Answer

Move session state from local SSD to a Memorystore for Redis instance with a Standard Tier replica in a second zone.. To survive a single-zone failure without data loss, the application must not store state on zonal resources. Moving session state to a replicated Memorystore for Redis instance removes VM-local state, and replacing the zonal persistent disk with a regional persistent disk replicates uploads across zones. Together, these changes allow the regional MIG to recreate instances in healthy zones while preserving session and upload data.

Answer

Configure the MIG to use preemptible VMs to reduce cost and allow rapid replacement of failed instances.

Answer

Enable autoscaling on the MIG based on CPU utilization and set a minimum instance count of two.

Answer

Change the MIG from regional to zonal and increase the number of instances in that zone to handle the full load.

Question 4

A SaaS provider serves customers from a single GKE cluster in us-east1 backed by a single Cloud SQL for PostgreSQL instance. The service level objective requires the application to remain available during a regional outage. Which change best meets this objective?

Accepted Answer

Run the application in a second GKE cluster in another region, replicate the database with a cross-region replica, and direct traffic with a global load balancer with health-checked failover.. Meeting a regional-outage objective requires resources in at least two regions, not just multiple zones. A second regional GKE cluster provides compute that survives the loss of us-east1, a cross-region database replica provides a promotable copy of the data, and a global load balancer with health checks automatically routes users to the surviving region. Only a multi-region design satisfies the objective.

Answer

Enable high availability on the Cloud SQL instance and add a read replica in another zone.

Answer

Increase the Cloud SQL instance size and add more nodes to the existing GKE cluster to absorb the extra load during an outage.

Answer

Deploy the GKE cluster as a regional cluster and configure the Cloud SQL instance for high availability.

Question 5

Your team runs an order-processing service on Cloud Run (fully managed) in europe-west1. Traffic is spiky, and during peak shopping events the service occasionally returns HTTP 503 errors while new instances are starting. You want to reduce cold-start latency for this Java service without changing the container's behavior. What should you do?

Accepted Answer

Configure a minimum number of instances on the Cloud Run service so warm instances are always available.. Cold-start latency on a fully managed Cloud Run service is best mitigated by keeping a baseline of warm instances, which the minimum instance count setting provides. That way, burst traffic is absorbed by already-initialized containers rather than triggering new cold starts. Scaling limits, timeouts, and caching do not address Java runtime initialization delays for dynamic order requests.

Answer

Enable Cloud CDN in front of the Cloud Run service to cache responses during peak events.

Answer

Increase the Cloud Run maximum number of instances so more containers can be created during spikes.

Answer

Raise the Cloud Run request timeout so slow-starting instances have more time to respond.

Question 6

A financial services company runs a payment API on Cloud Run that calls a third-party provider. The provider enforces a strict rate limit and returns HTTP 429 when exceeded. During peak hours, the API retries immediately on 429 responses, causing a retry storm and cascading failures. The team wants to protect the provider and keep the API responsive. What should they do?

Accepted Answer

Implement exponential backoff with jitter on 429 responses and add a circuit breaker that opens after repeated failures.. Backoff with jitter desynchronizes retries so the provider is not hammered simultaneously, and a circuit breaker sheds load when failures persist, allowing recovery. This directly addresses the retry storm and cascading failure. Simply adding capacity or shortening timeouts does not reduce pressure on the rate-limited dependency.

Answer

Increase the Cloud Run maximum instance count so more requests can be handled concurrently and the provider limit is less likely to be hit.

Answer

Configure a Cloud Run timeout shorter than the provider's response time so failed calls are terminated before they can be retried.

Answer

Move the third-party calls to a Cloud Tasks queue that delivers each request exactly once with no retry configuration.

Question 7

A startup runs a stateless REST API on a managed instance group of Compute Engine VMs behind an external Application Load Balancer. The API is deployed in a single region with three zones, but during a regional outage the service was completely unavailable. The startup wants the API to remain available even if an entire region fails, with minimal changes to the application code. What should they do?

Accepted Answer

Deploy the managed instance group in a second region and add a global external Application Load Balancer with backend services in both regions.. Regional resilience requires capacity in more than one region plus a traffic distribution mechanism that can react to a regional failure. A global external Application Load Balancer fronts backend services in multiple regions behind one anycast IP and automatically steers users to healthy backends. Because the API is stateless, deploying the same managed instance group template in a second region completes the design with no application changes.

Answer

Add more zones to the existing regional managed instance group and increase the minimum number of instances per zone.

Answer

Enable autoscaling on the existing managed instance group and raise the maximum instance count to absorb the outage.

Answer

Create a Cloud DNS failover routing policy with health checks that points to a static IP in a second region.

Question 8

A startup is building a new REST API on Cloud Run. The API must handle unpredictable traffic spikes and scale automatically. The startup wants to minimize costs when there is no traffic. What should they do?

Accepted Answer

Configure the Cloud Run service to allow scaling to zero by not setting a minimum instance count.. Cloud Run automatically scales to zero when there is no traffic, so you only pay for what you use. This is ideal for unpredictable traffic and minimizing costs during idle periods. Allowing the service to scale to zero by not setting a minimum instance count achieves this, while still enabling automatic scaling during spikes.

Answer

Use a Cloud Run service with a minimum of 0 instances and set a maximum of 1 instance.

Answer

Deploy the API on a Compute Engine managed instance group with autoscaling enabled.

Answer

Set the minimum number of instances to 1 and the maximum to 1000.

Question 9

A development team is deploying a containerized microservices application on Google Kubernetes Engine (GKE). They need to ensure that the application remains available even if a single node fails, and that pods are automatically rescheduled onto healthy nodes. They also want to minimize manual operational overhead. Which configuration should they use?

Accepted Answer

Create a regional GKE cluster with multiple node pools across zones and rely on the Kubernetes scheduler and ReplicaSet controller to reschedule pods.. A regional GKE cluster distributes nodes across multiple zones, so a single node or even an entire zone failure does not take down the application. The ReplicaSet controller ensures the desired number of pod replicas exists, and the Kubernetes scheduler places replacement pods on healthy nodes. This built-in automation minimizes manual intervention and directly addresses the availability and operational overhead requirements.

Answer

Create a zonal GKE cluster and manually create standby nodes in a second zone that can be added to the cluster during a failure.

Answer

Create a regional GKE cluster and configure each pod to use a node selector that pins it to a specific node for predictable performance.

Answer

Create a zonal GKE cluster with a single node pool and configure pod anti-affinity to spread pods across nodes.

Question 10

A financial analytics platform ingests market data into Cloud Bigtable and serves point reads to a trading dashboard. The application currently uses a single-cluster instance in us-central1. The business requires that the dashboard remain readable even if the entire us-central1 region becomes unavailable, with a recovery point objective of zero. Which Bigtable configuration should the team implement?

Accepted Answer

Configure a multi-cluster routing instance with clusters in us-east1 and us-west1, and use an application profile with cluster group routing.. Bigtable achieves cross-region availability by placing clusters in multiple regions within one multi-cluster instance; replication is synchronous, so the recovery point objective is zero. The application profile's routing policy determines whether clients actually use the other clusters. Cluster group routing sends requests to any cluster in the group, enabling automatic failover when a region is lost, whereas single-cluster routing would keep clients pinned to the failed region.

Answer

Export nightly snapshots of the Bigtable tables to Cloud Storage in us-east1 and restore them if us-central1 fails.

Answer

Enable a read replica in us-east1 on the existing single-cluster instance and point the dashboard at the replica endpoint.

Answer

Create a multi-cluster routing instance with clusters in us-east1 and us-west1, and configure the application profile with single-cluster routing.

Question 11

A media streaming company runs a read-heavy API on Cloud Run (fully managed) serving a global audience. The service reads from Cloud Storage and occasionally writes user playback history to Cloud Firestore. Users in Asia report high latency, while users in North America experience fast responses. The team wants to reduce latency for all users without changing the application code. What should they do?

Accepted Answer

Deploy the Cloud Run service to multiple regions and use a global external HTTP(S) load balancer to route users to the nearest region.. Deploying the Cloud Run service in multiple regions and using a global external HTTP(S) load balancer ensures that users are routed to the nearest regional deployment, minimizing network latency. This approach requires no application code changes and leverages Google Cloud's global load balancing and Cloud Run's regional scalability. It also improves availability because traffic can fail over to another region if one becomes unhealthy.

Answer

Increase the maximum number of instances for the Cloud Run service and set concurrency to 1.

Answer

Enable Cloud CDN with a global external HTTP(S) load balancer in front of the Cloud Run service and cache the API responses.

Answer

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

Question 12

A startup is developing a new web application that will be deployed on Google Cloud. They expect the application to have unpredictable traffic patterns, with periods of no traffic followed by sudden spikes. The development team wants to minimize operational overhead and only pay for resources when the application is handling requests. They also need the application to scale automatically to handle the spikes. Which Google Cloud service should they use for the application's frontend?

Accepted Answer

Cloud Run. Cloud Run is a fully managed serverless compute platform that automatically scales instances based on incoming requests, including scaling to zero when there is no traffic. This means the startup only pays for the resources consumed during request handling, aligning with cost efficiency. It also removes the need to manage servers or clusters, minimizing operational overhead. Thus, Cloud Run is the best fit for unpredictable traffic patterns.

Answer

Google Kubernetes Engine (GKE) Autopilot cluster

Answer

App Engine standard environment

Answer

Compute Engine managed instance group with autoscaling

Question 13

A financial services company runs a Python service on Cloud Run that writes transaction records to Cloud Spanner. During a regional outage in us-central1, the service in us-east1 continues to serve reads but writes fail with 'Session not found' errors after about 10 minutes. The Spanner instance is configured as a regional instance in us-central1. What should the company do to make writes survive a regional outage with minimal application changes?

Accepted Answer

Recreate the Spanner instance as a multi-region instance with a configuration that includes us-central1 and us-east1, and update the application to use the new instance ID.. Cloud Spanner regional instances keep all replicas and the leader in one region, so a regional outage stops writes. A multi-region instance replicates across regions and automatically elects a new leader in a surviving region, providing write availability with no application-level failover logic. Changing the instance configuration and updating the connection string is the only application change required.

Answer

Add a read replica in us-east1 and configure the application to write to the replica when the primary is unavailable.

Answer

Configure the Cloud Run service to retry failed writes using exponential backoff and a longer session timeout.

Answer

Increase the Spanner instance's node count and enable automatic scaling to handle the failover traffic.

Question 14

A global logistics company runs a microservices application on Google Kubernetes Engine (GKE) across multiple regions. They use a multi-region Cloud Spanner instance for their database. They want to ensure that the application remains available even if an entire region fails, and they want to minimize data replication lag. Which design should they implement?

Accepted Answer

Deploy GKE clusters in multiple regions and configure the application to connect to the nearest Cloud Spanner replica using a multi-region instance configuration.. Cloud Spanner multi-region instances provide automatic replication and failover across regions with strong consistency. Deploying GKE clusters in multiple regions and connecting to the nearest Spanner replica ensures low latency and high availability. This design allows the application to survive a region failure without manual intervention.

Answer

Deploy GKE clusters in multiple regions and use a Cloud SQL for PostgreSQL instance with cross-region replicas, manually promoting the replica during a regional failure.

Answer

Deploy GKE clusters in multiple regions and use a Cloud Spanner multi-region instance, but configure the application to write to a single region to avoid conflicts.

Answer

Use a single GKE cluster in one region and a Cloud Spanner regional instance with a read replica in another region.

Question 15

A team is designing a new microservices application on Cloud Run and wants to maximize reliability and availability without over-engineering. Which TWO practices should they adopt? (Choose two.)

Accepted Answer

Configure each service with a minimum instance count greater than zero and a health-check-compatible startup so cold starts do not cause request failures during traffic ramp-up.. Reliability on Cloud Run comes from reducing cold-start exposure and handling transient failures gracefully. Warm minimum instances smooth the ramp-up so requests are not rejected while containers start, and client-side retries with backoff and jitter combined with idempotent operations let the system recover from brief failures without duplicating effects. Together these practices raise availability with modest cost and complexity.

Answer

Set the container concurrency to 1 for every service so each instance handles exactly one request at a time.

Answer

Disable Cloud Run's built-in health checks and rely on client-side timeouts to detect unhealthy instances.

Answer

Pin every Cloud Run service to a single region to reduce cross-region latency and operational complexity.