Google PCA Ensure solution and operations reliability — 30 Questions

Question 1

A company runs a critical application on Compute Engine instances in a managed instance group (MIG) with autoscaling. During a traffic spike, some instances become unhealthy but are not automatically replaced. What is the most likely cause?

Accepted Answer

The autohealing health check is misconfigured.. The most likely cause is that the autohealing health check is misconfigured. In a managed instance group, autohealing relies on a health check to detect unhealthy instances and trigger replacement. If the health check is misconfigured (e.g., wrong port, path, or protocol), the MIG will not recognize instances as unhealthy and will not automatically replace them, even during a traffic spike.

Answer

The MIG is regional and one zone failed.

Answer

The instance template has a startup script error.

Answer

The HTTP load balancer's health check is failing.

Question 2

A company is designing a disaster recovery plan for a Cloud SQL for PostgreSQL instance. They want to failover to a different region with minimal data loss and recovery time under 10 minutes. The database is 500 GB and experiences 2,000 write transactions per second. Which solution should they use?

Accepted Answer

Configure a cross-region replica instance using Cloud SQL's cross-region replication feature.. Cloud SQL for PostgreSQL offers a managed cross-region replication feature that creates a replica instance in a different region, using synchronous or asynchronous replication to keep data nearly in sync. This solution meets the RPO (minimal data loss) and RTO (under 10 minutes) requirements because the replica is continuously updated and can be promoted to primary in minutes, without needing to restore from a backup or export.

Answer

Export the database daily using gsutil and import in the other region using pg_restore.

Answer

Create a cross-region read replica and promote it to primary during failover.

Answer

Automated backups with point-in-time recovery to a new instance in the other region.

Question 3

A company uses Cloud Spanner for a global financial application. They experience increased latency and transaction aborts during peak hours. Which measure should they take first to improve reliability?

Accepted Answer

Optimize transactions to reduce lock contention.. Option C is correct because transaction aborts and latency in Cloud Spanner are most commonly caused by lock contention during peak hours. By optimizing transactions—such as reducing their scope, using read-only transactions where possible, and avoiding hot-spot writes—you directly address the root cause of contention without incurring additional cost or schema changes. This aligns with Google's best practices for Spanner reliability.

Answer

Increase the number of nodes in the Spanner instance.

Answer

Reduce the number of indexes on frequently updated columns.

Answer

Use interleaved tables to co-locate related data.

Question 4

A company deploys a microservices application on Google Kubernetes Engine (GKE). Pods in one deployment are frequently OOMKilled. The team sets memory requests and limits, but pods still crash. What is the most likely remaining cause?

Accepted Answer

The application has a memory leak that eventually exceeds the limit.. Option D is correct because OOMKilled errors occur when a container exceeds its memory limit. Setting memory requests and limits prevents unbounded usage, but if the application has a memory leak, it will continue to consume memory until it hits the configured limit, causing the kernel's Out-Of-Memory (OOM) killer to terminate the pod. The fact that pods still crash after setting limits indicates the application itself is the root cause, not resource configuration.

Answer

CPU requests are too low, causing throttling and eventual crash.

Answer

The node pool is too small, causing memory pressure on the node.

Answer

Memory limits are set higher than the node's allocatable memory.

Question 5

An organization uses Cloud Functions (2nd gen) for event-driven processing. They notice that some functions fail with 'memory limit exceeded' errors during peak load. The function processes messages from Pub/Sub and writes to Firestore. What should they do to improve reliability without sacrificing throughput?

Accepted Answer

Increase the memory allocated to the Cloud Function.. The 'memory limit exceeded' error indicates that the function's allocated memory is insufficient for the workload during peak load. Increasing the memory allocation (Option B) directly resolves this by providing more RAM for processing larger messages or concurrent operations, without altering the invocation pattern or throughput. Cloud Functions (2nd gen) allow memory to be set up to 32 GiB, and this change does not reduce the number of events processed per second.

Answer

Increase the maximum number of concurrent function instances.

Answer

Enable Pub/Sub batching to reduce the number of function invocations.

Answer

Split the function into multiple smaller functions, each handling a subset of the data.

Question 6

A company deploys a stateful workload using StatefulSets on GKE. They want to ensure that if a pod is evicted, its persistent volume claim (PVC) is reattached to the replacement pod in the same zone. Which configuration achieves this?

Accepted Answer

Use a StatefulSet with a volumeClaimTemplate referencing a persistent disk in the same zone.. StatefulSets are designed for stateful workloads and guarantee stable network identities and persistent storage. When a pod is evicted, the StatefulSet controller ensures the replacement pod uses the same PVC, which is bound to a GCE Persistent Disk in the same zone as the original pod, provided the volumeClaimTemplate specifies a disk in that zone. This maintains data locality and avoids cross-zone reattachment.

Answer

Use a Deployment with a PVC that has allowedTopologies restricting to the desired zone.

Answer

Use a Deployment with a persistent volume that is manually attached after pod creation.

Answer

Use a StatefulSet with a persistent disk that has access mode ReadOnlyMany.

Question 7

A company monitors their application with Cloud Monitoring. They set up an alerting policy to notify the on-call team when the 99th percentile latency exceeds 500 ms for 5 minutes. However, they receive false positive alerts due to short bursts. How should they refine the policy?

Accepted Answer

Increase the evaluation window to 10 minutes.. Option D is correct because increasing the evaluation window to 10 minutes smooths out short bursts of high latency, ensuring the alert triggers only when the 99th percentile latency exceeds 500 ms for a sustained period. Cloud Monitoring evaluates metrics over the specified window, so a longer window reduces false positives from transient spikes while still detecting genuine degradation.

Answer

Set up alerting on each data point individually.

Answer

Decrease the threshold to 400 ms.

Answer

Change the metric to average latency instead of 99th percentile.

Question 8

A company runs a web application on Compute Engine behind an HTTP load balancer. They want to improve reliability by implementing failover across two regions. Which TWO actions should they take?

Accepted Answer

Deploy a global external HTTP load balancer with backends in both regions.. A global external HTTP load balancer is required for cross-region failover because it uses a single anycast IP address and routes traffic to the closest healthy backend. By deploying backends in both regions, the load balancer automatically fails over to the secondary region if the primary region's backends become unhealthy, improving reliability without DNS propagation delays.

Answer

Configure DNS-based failover using Cloud DNS with health checks.

Answer

Use an internal load balancer to route traffic between regions.

Answer

Use a regional external HTTP load balancer with a multi-region backend.

Question 9

A company uses Cloud CDN to accelerate content delivery. They notice that some users receive stale content even after purging the cache. Which THREE factors could cause this?

Accepted Answer

The purge request did not complete successfully.. Option B is correct because a purge request that does not complete successfully will leave cached content intact, causing users to receive stale data. Cloud CDN processes purge requests asynchronously, and if the request fails (e.g., due to network issues or invalid paths), the cache is not invalidated. This directly explains why stale content persists despite an attempted purge.

Answer

The content is compressed with gzip.

Answer

The CDN is configured with signed URLs.

Question 10

A company deploys a critical application on Google Kubernetes Engine (GKE) and wants to ensure high availability during cluster upgrades. Which TWO practices should they follow?

Accepted Answer

Use multiple node pools across different zones within the cluster.. Option B is correct because deploying multiple node pools across different zones ensures that if one zone fails or is taken down for maintenance, the application can continue serving from the other zones. This aligns with GKE's best practice for high availability by distributing workloads across failure domains. Option C is correct because PodDisruptionBudgets (PDBs) define the minimum number of pods that must remain available during voluntary disruptions like cluster upgrades, preventing the upgrade from taking down too many replicas at once.

Answer

Use a single-zone node pool with multiple replicas.

Answer

Enable cluster autoscaling to add nodes during upgrades.

Answer

Enable regional clusters for multi-zone control plane.

Question 11

A company runs a multi-tier application on Google Cloud: a frontend on App Engine Standard, a backend on Cloud Run, and a Cloud SQL database. The application experiences intermittent 500 errors when users submit forms. The errors correlate with high CPU usage on the Cloud SQL instance (db-n1-standard-2, 7.5 GB memory). The Cloud Run service has a concurrency setting of 80 and a maximum of 10 instances. The App Engine service uses automatic scaling. The team has verified that the application code is not the issue. They suspect the database is hitting connection limits. Current max_connections on Cloud SQL is 250. The Cloud Run service uses a connection pool of 10 connections per instance. The App Engine service uses a connection pool of 5 connections per instance. They also have a few batch jobs that run occasionally, using up to 10 connections. The team wants to resolve the errors with minimal cost and complexity. Which course of action should they take?

Accepted Answer

Increase the max_connections parameter on Cloud SQL to 500.. The intermittent 500 errors are caused by the Cloud SQL instance hitting its max_connections limit of 250. With Cloud Run using 10 connections per instance and up to 10 instances (100 connections), App Engine using 5 connections per instance (unknown instance count but likely significant), and batch jobs using up to 10 connections, the total can easily exceed 250. Increasing max_connections to 500 directly addresses the connection limit without changing instance size or scaling behavior, which is the simplest and most cost-effective fix.

Answer

Increase the maximum number of Cloud Run instances to 20 to handle more requests.

Answer

Upgrade the Cloud SQL instance to db-n1-standard-4 (15 GB memory) to handle more connections.

Answer

Reduce the concurrency setting on Cloud Run from 80 to 40.

Question 12

A company runs a web application on Google Kubernetes Engine (GKE) with Cluster Autoscaler enabled. During a traffic spike, the application becomes slow and some requests timeout. The cluster has sufficient CPU and memory headroom. What is the most likely cause and solution?

Accepted Answer

Configure Horizontal Pod Autoscaler (HPA) based on CPU utilization or custom metrics.. The correct answer is D because the cluster has sufficient CPU and memory headroom, indicating that the issue is not about cluster capacity but about pod-level scaling. The Horizontal Pod Autoscaler (HPA) automatically scales the number of pod replicas based on observed CPU utilization or custom metrics, which directly addresses the application slowdown and timeouts during traffic spikes by distributing the load across more pods.

Answer

Increase the node pool's machine type to a larger size.

Answer

Enable Cluster Autoscaler to add more nodes.

Answer

Deploy the application in a regional cluster for higher availability.

Question 13

An organization is migrating a legacy monolithic application to Google Cloud. The application currently runs on a single server with an on-premises database. The application is stateful and requires low-latency access to the database. The migration must minimize downtime and ensure high availability. Which architecture should the company adopt?

Accepted Answer

Deploy on Compute Engine with a regional persistent disk and use Cloud SQL for PostgreSQL with regional high availability.. Option B is correct because it combines Compute Engine with a regional persistent disk for synchronous replication across zones, ensuring high availability with minimal downtime during a zonal failure. Cloud SQL for PostgreSQL with regional high availability provides a managed, low-latency database with automatic failover, meeting the stateful application's need for low-latency access and high availability without the complexity of container orchestration.

Answer

Deploy on GKE with StatefulSets and use Cloud Spanner for global consistency.

Answer

Deploy on App Engine Standard Environment and use Cloud Firestore in Datastore mode.

Answer

Deploy on Cloud Run and use Cloud SQL with read replicas.

Question 14

A company uses Cloud SQL for MySQL to host its production database. The database experiences high read traffic. The team wants to improve read performance without modifying the application. What should they do?

Accepted Answer

Add read replicas and configure the application to use them for read queries.. Option C is correct because adding read replicas offloads read queries from the primary Cloud SQL instance, distributing the read load across multiple replicas. This improves read performance without any application code changes, as the application can be configured to direct read queries to the replica endpoints. Cloud SQL for MySQL replicas use asynchronous replication, ensuring near-real-time data consistency for read-heavy workloads.

Answer

Increase the number of CPUs on the primary Cloud SQL instance.

Answer

Use Cloud SQL Proxy with connection pooling.

Answer

Enable automatic storage increase to allow more data.

Question 15

A company is running a critical application on Compute Engine. The application writes logs to a local persistent disk. The operations team wants to ensure logs are not lost if the VM fails. What should they do?

Accepted Answer

Configure the application to write logs to Cloud Logging using the Logging agent.. Option D is correct because Cloud Logging with the Logging agent provides a centralized, durable, and managed log storage solution. The agent streams logs from the VM to Cloud Logging in near real-time, ensuring logs are preserved even if the VM or its local persistent disk fails. This decouples log storage from the VM's lifecycle, meeting the operations team's requirement for log durability.

Answer

Use a regional persistent disk to replicate data across zones.

Answer

Schedule persistent disk snapshots every 5 minutes.

Answer

Create a script to copy logs to a Cloud Storage bucket every minute.

Question 16

Which TWO options are best practices for ensuring high availability of an application running on Google Kubernetes Engine (GKE)?

Accepted Answer

Use pod anti-affinity to spread pods across multiple zones.. Option A is correct because pod anti-affinity ensures that pods from the same application are scheduled on different nodes across multiple zones, reducing the blast radius of a zonal failure. This is a key pattern for achieving high availability in GKE, as it prevents a single zone outage from taking down all replicas of your application.

Answer

Deploy all nodes in the same zone to simplify networking.

Answer

Prefer using preemptible VMs for cost savings.

Answer

Use a single zonal cluster to avoid cross-zone latency.

Question 17

Which THREE options are valid strategies for disaster recovery (DR) in Google Cloud?

Accepted Answer

Deploy a mirrored environment in another region and use Traffic Director to fail over.. Option B is correct because Traffic Director, based on the xDS API (Envoy), can manage traffic routing across regions. By deploying a mirrored environment in another region and configuring Traffic Director with failover policies, you can redirect traffic to the secondary region if the primary fails, enabling a robust active-passive or active-active DR strategy.

Answer

Store hourly snapshots of Compute Engine disks in the same region.

Answer

Enable Cloud CDN to cache static content from multiple origins.

Question 18

A company runs a batch processing workload on Compute Engine that processes financial transactions. The workload runs daily and must complete within a 4-hour window. The application reads input data from Cloud Storage, processes it, and writes output to another Cloud Storage bucket. The current implementation uses a single VM with a 500 GB persistent disk. Recently, the data volume has increased, and the job is now taking over 6 hours, exceeding the SLA. The team is tasked with redesigning the solution to be faster and more reliable. They want to minimize costs and operational overhead. The data is critical and must not be lost. Which approach should they take?

Accepted Answer

Use Cloud Dataflow with autoscaling to process the data in parallel.. Cloud Dataflow with autoscaling is the correct choice because it provides a fully managed, serverless service for parallel data processing that can automatically scale resources based on the volume of data. This directly addresses the need to complete the batch workload within the 4-hour SLA, as Dataflow can distribute the processing across many workers, significantly reducing execution time. It also ensures reliability and data durability through checkpointing and exactly-once processing semantics, meeting the critical data loss prevention requirement.

Answer

Use a managed instance group with a startup script that processes data, and use Cloud Pub/Sub to coordinate.

Answer

Increase the VM to a high-CPU machine type with a regional persistent disk for HA.

Answer

Deploy the processing logic in Cloud Functions and trigger from Cloud Storage events.

Question 19

A company has deployed a critical application on Google Kubernetes Engine (GKE) with a Regional cluster (us-central1). The application uses a Cloud SQL for PostgreSQL database with a cross-region replica for disaster recovery. The SRE team needs to ensure that the application can survive a regional outage with minimal data loss. Which TWO actions should the team take to improve the reliability of the solution?

Accepted Answer

Configure the application to automatically promote the Cloud SQL cross-region replica to a primary instance when the primary region is unavailable.. Option A is correct because promoting a Cloud SQL cross-region replica to a primary instance is the standard procedure for disaster recovery when the primary region fails. This operation is supported by Cloud SQL and can be automated using Cloud Functions or Cloud Run triggers, ensuring minimal manual intervention and reduced RTO. The cross-region replica maintains a near-synchronous copy of the data, so data loss is limited to the replication lag.

Answer

Configure Cloud SQL cross-region replication to be synchronous to ensure zero data loss during failover.

Answer

Deploy a secondary GKE cluster in the same region as the primary to provide a hot standby that can take over immediately.

Answer

Use a TCP/UDP load balancer to route traffic to both regions based on latency.

Question 20

You are investigating a Vertex AI Workbench instance (instance-2) that is showing UNHEALTHY status. Based on the exhibit, what is the most likely cause of the issue?

Accepted Answer

The container image gcr.io/my-project/my-image:latest does not exist, or the service account used by the Workbench instance does not have storage.objectViewer access to the container registry.. The UNHEALTHY status in Vertex AI Workbench typically occurs when the instance fails to start its container. The most likely cause is that the specified container image (gcr.io/my-project/my-image:latest) does not exist in the Container Registry, or the service account attached to the instance lacks the storage.objectViewer role on the registry bucket. Without this permission, the instance cannot pull the image, leading to a container creation failure and an UNHEALTHY state.

Answer

The container registry endpoint is blocked by a firewall rule that does not allow egress to gcr.io.

Answer

The instance's underlying Compute Engine resources are exhausted, causing the container creation to timeout.

Answer

The Workbench instance is using an outdated custom image that is not compatible with the latest runtime version.