PCDE Practice Test 45 — 15 Questions

Question 1

A financial technology firm is designing a Cloud Spanner schema for a payments ledger. The ledger table will contain hundreds of billions of rows and must support high-throughput inserts of new transactions and efficient lookups by account ID for the most recent transactions. The team wants to avoid hotspots and minimize storage cost for older records. Which schema design should they choose?

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

Accepted Answer

Use a composite primary key of account ID followed by a reversed or hashed transaction timestamp. A composite primary key that starts with account ID spreads inserts across many key ranges, avoiding the hotspot caused by a sequential key, while ordering the timestamp portion in reverse or hashed form keeps recent transactions for each account physically close for fast lookups. This balances high insert throughput with efficient per-account recent-transaction queries and avoids secondary index overhead.

Answer

Use a monotonically increasing transaction ID as the primary key and rely on Spanner load balancing

Answer

Use a random UUID as the primary key and create a secondary index on account ID and timestamp

Answer

Use a composite primary key of transaction timestamp followed by account ID with a descending timestamp

Question 2

A financial services firm runs a Cloud SQL for PostgreSQL instance that must be highly available across two zones. They need automatic failover with minimal data loss and want to test failover without impacting production. Which configuration and action should they use?

Accepted Answer

Enable high availability and use the `gcloud sql instances failover` command to trigger a failover.. Enabling high availability on a Cloud SQL instance creates a synchronous standby in another zone, enabling automatic failover with no data loss. The `gcloud sql instances failover` command is the supported method to manually trigger a failover for testing. Read replicas use asynchronous replication and manual promotion, and restart does not exercise the failover path.

Answer

Create a read replica in another zone and promote it manually during an outage.

Answer

Use a Cloud SQL replica with automatic failover enabled via the `--failover` flag.

Answer

Enable high availability and use the `gcloud sql instances restart` command to test failover.

Question 3

A financial services company is designing a highly available database solution on Google Cloud for a trading application that requires strong consistency and low latency across multiple regions. They are evaluating Cloud Spanner and Cloud SQL. Which two statements are true regarding the capabilities of Cloud Spanner compared to Cloud SQL? (Choose two.)

Accepted Answer

Cloud Spanner offers multi-region replication with strong consistency, while Cloud SQL supports cross-region read replicas but not multi-region strong consistency for writes.. Cloud Spanner is designed for horizontal scalability and multi-region strong consistency, using automatic sharding and synchronous replication. Cloud SQL is a regional, vertically scalable relational database that supports cross-region read replicas but not multi-region strong consistency for writes. These fundamental architectural differences make Cloud Spanner the choice for globally distributed, strongly consistent, high-scale transactional workloads.

Answer

Cloud Spanner requires manual indexing for all queries, whereas Cloud SQL automatically creates indexes for primary keys and foreign keys.

Answer

Cloud Spanner supports standard SQL with ACID transactions, while Cloud SQL only supports eventual consistency for transactions.

Answer

Cloud Spanner is limited to a single region, while Cloud SQL can be deployed across multiple regions with strong consistency.

Question 4

A team wants to set up alerting on error budget burn rate for a service with an SLO of 99.9%. They want to detect when the error budget is being consumed at a rate that would exhaust it in less than 24 hours, using a 1-hour assessment window. What is the appropriate burn rate threshold for a fast burn alert?

Accepted Answer

100x burn rate. 100x burn rate. To exhaust a 30-day error budget in less than 24 hours, the burn rate must exceed 30x (30 days / 1 day = 30). A burn rate of 100x is a standard threshold for fast burn alerts that guarantees exhaustion in less than 24 hours. The 14x threshold would exhaust the budget in approximately 2.14 days, which does not meet the requirement.

Answer

5x burn rate

Answer

14x burn rate

Answer

2x burn rate

Question 5

A team wants to implement chaos engineering on Google Kubernetes Engine (GKE) to test resilience against pod failures. Which tool is designed for injecting faults into GKE clusters?

Accepted Answer

Chaos Mesh on GKE. Chaos Mesh is a CNCF-graduated chaos engineering platform that runs natively on Kubernetes, including GKE, and provides a Chaos Mesh Operator plus CRDs (PodChaos, NetworkChaos, StressChaos, IOChaos) to inject pod-kill, container-kill, network partition, latency, and resource stress faults. It is purpose-built for injecting faults into Kubernetes workloads, which is exactly what the team needs to test resilience against pod failures on GKE.

Answer

GKE Node Auto-Repair

Answer

Traffic Director fault injection

Answer

Cloud Armor

Question 6

A retail company is deploying a new order-processing application on Google Cloud. The application requires a relational database that supports automatic, multi-region replication with strong consistency and 99.999% availability. The database will be accessed from users in North America, Europe, and Asia. The team wants minimal operational overhead and the ability to scale horizontally for writes. Which Google Cloud database should they choose?

Accepted Answer

Cloud Spanner. Cloud Spanner is the only Google Cloud database that combines relational semantics, horizontal write scalability, strong consistency, and multi-region replication with a 99.999% availability SLA. The other options either lack global strong consistency, horizontal write scaling, or relational capabilities, making them unsuitable for this scenario.

Answer

Cloud Bigtable

Answer

Cloud SQL for PostgreSQL with cross-region replicas

Answer

Firestore in Datastore mode

Question 7

A team wants to visualize real-time CPU utilization across a fleet of Compute Engine instances in a single dashboard. Which chart type in Cloud Monitoring is most appropriate?

Accepted Answer

Line chart. A line chart is the correct choice because it plots metric values against a continuous time axis, making trends and spikes in CPU utilization across a fleet immediately visible. Cloud Monitoring's time-series model is natively rendered as line charts, and they support multiple series (one per instance) overlaid on the same axes. This is the standard visualization for real-time utilization metrics.

Answer

Stacked bar chart

Answer

Scatter plot

Answer

Heatmap

Question 8

Which of the following best describes 'toil' in SRE?

Accepted Answer

Manual, repetitive work that provides no enduring value and scales linearly with service growth. In SRE, toil is defined as manual, repetitive, tactical work that is automatable, provides no enduring value, and scales linearly with service growth. Google's SRE book explicitly caps toil at 50% of an SRE's time to preserve engineering capacity.

Answer

Work that is creative and requires deep domain knowledge

Answer

Automating infrastructure provisioning

Answer

Incident response and on-call duties

Question 9

A DevOps engineer is optimizing a GKE workload that is CPU-bound. They want to ensure proper resource allocation to improve performance. Which three actions should they take? (Choose THREE).

Accepted Answer

Set Horizontal Pod Autoscaler based on CPU utilization. Option B is correct because a Horizontal Pod Autoscaler targeting CPU utilization directly addresses a CPU-bound workload by adding or removing pod replicas as average CPU usage crosses the target threshold, improving throughput under load. Option D is correct because enabling the Cluster Autoscaler on the node pool ensures that when HPA-created pods cannot be scheduled due to insufficient node capacity, new nodes are provisioned automatically, keeping the CPU-bound workload supplied with compute. Option E is correct because running the Vertical Pod Autoscaler in Auto mode right-sizes each pod's CPU (and memory) requests and limits based on observed usage, which improves scheduling accuracy and performance for CPU-bound containers. Option A is not relevant because min instances is a Cloud Run serverless setting, not a GKE resource-allocation control. Option C is not appropriate because preemptible (Spot) nodes can be reclaimed at any time, which risks disrupting a performance-sensitive CPU-bound workload rather than improving its resource allocation.

Answer

Set min instances on Cloud Run

Answer

Use preemptible nodes

Question 10

You want to visualize the 99th percentile latency of a service on a dashboard. The metric is a distribution metric. Which reducer should you use in the chart configuration?

Accepted Answer

99TH_PERCENTILE. For a distribution metric, the chart reducer determines how the underlying distribution is aggregated into a single value per time interval. To visualize the 99th percentile latency, you must select the 99TH_PERCENTILE reducer, which computes the 99th percentile of the distribution rather than an average or sum. This is the correct reducer because latency is a distribution metric and percentiles are the standard way to surface tail latency.

Answer

MEAN

Answer

SUM

Answer

COUNT

Question 11

A Cloud SQL instance is using InnoDB and has a large buffer pool. The query performance is slower after a failover. What is the most likely cause?

Accepted Answer

Buffer pool warm-up time. After a failover in Cloud SQL, the new primary instance starts with a cold buffer pool. InnoDB relies on the buffer pool to cache data and index pages in memory for fast queries. Since the buffer pool is empty after the failover, queries must read from disk until the cache warms up, causing significantly slower performance.

Answer

Read replica lag

Answer

Binary log not enabled

Answer

Data corruption

Question 12

A financial analytics firm uses Cloud Bigtable to store time-series trade data. The row key is designed as reversed timestamp (e.g., #<reverse_timestamp>) to enable quick retrieval of the latest trades. They now need to query trades by account ID and time range, and also by symbol and time range. The current row key does not support these queries efficiently. What should they do to optimize for these access patterns?

Accepted Answer

Create a new table with row key as account_id#reverse_timestamp and another table with symbol#reverse_timestamp, and write data to both tables using an application-level dual-write.. To support multiple access patterns in Bigtable, you must design row keys that match the query patterns and often duplicate data into separate tables. Using account_id#reverse_timestamp and symbol#reverse_timestamp as row keys in two tables allows efficient range scans for each pattern. Dual-writes keep the tables in sync. Other options rely on non-existent features or inefficient full scans.

Answer

Use Cloud Bigtable's built-in secondary index feature to create indexes on account_id and symbol.

Answer

Add a column family for account_id and another for symbol, and use filters to query by those columns.

Answer

Change the row key to account_id#symbol#reverse_timestamp to support both queries in a single table.

Question 13

A media analytics team stores 300 TB of clickstream events in Bigtable and needs to serve two access patterns: (1) fetch the latest events for a specific user in reverse chronological order, and (2) scan all events for a given day across all users for batch aggregation. The current row key is a random UUID, which makes both patterns slow. They want a design that keeps writes evenly distributed while supporting both patterns. Which row key design should they use?

Accepted Answer

Use a row key of salted hash bucket followed by reversed timestamp, and add a separate table keyed by date plus user ID for the daily scan.. Because Bigtable has no secondary indexes, a single table cannot efficiently serve both a per-user reverse-chronological lookup and a cross-user daily scan. Salting the row key with a hash bucket keeps writes spread across tablets, and a denormalized second table keyed by date plus user ID provides the batch aggregation path without full scans.

Answer

Use a row key of reversed timestamp followed by user ID, and rely on a column family per user for the daily scan.

Answer

Use a row key of date followed by user ID and timestamp, and query the per-user latest events with a filter on the user ID column.

Answer

Use a row key of user ID followed by reversed timestamp, and rely on Bigtable's built-in secondary indexes for the daily scan.

Question 14

A company wants to use Cloud Monitoring dashboards to display real-time metrics for their application, but they also need to version control the dashboard configurations. Which approach should they use?

Accepted Answer

Define dashboards as code using the Cloud Monitoring API and store the configuration in a Git repository.. Defining dashboards as code via the Cloud Monitoring API (or Terraform's google_monitoring_dashboard resource) lets the JSON configuration live in a Git repository, giving full version control, peer review, and reproducible deployments. This is the only option that treats dashboard configuration as a first-class, auditable artifact rather than a manual export. It aligns with infrastructure-as-code principles required for change tracking and rollback.

Answer

Manually create dashboards in the Cloud Monitoring console and export them periodically.

Answer

Use Grafana with the Cloud Monitoring data source and export dashboard JSON.

Answer

Use Cloud Monitoring's built-in 'Clone' feature to backup dashboards.

Question 15

A media streaming service uses Bigtable to store user viewing history. The row key is currently a hash of the user ID followed by a timestamp. They need to retrieve the most recent 100 viewing events for a given user efficiently. What change should they make to the row key design?

Accepted Answer

Use the user ID hash as a prefix, followed by an inverted timestamp.. To efficiently retrieve the most recent events for a user, the row key should start with a user identifier to group events, followed by an inverted timestamp to sort events in descending time order. Hashing the user ID ensures even distribution across tablets. This design allows a single range scan to fetch the latest events without scanning unrelated data.

Answer

Reverse the timestamp and prepend it to the user ID hash.

Answer

Store the timestamp as a separate column qualifier instead of in the row key.

Answer

Use a sequential user ID without hashing, followed by the timestamp.