PCDE Practice Exam 57 — 20 Questions

Question 1

Refer to the exhibit. A developer deployed these Firestore security rules. What is a security concern with this configuration?

Accepted Answer

All authenticated users can read all documents. The Firestore security rules shown grant read access to all authenticated users without any document-level or collection-level restrictions. The rule `match /databases/{database}/documents { allow read: if request.auth != null; }` applies to all documents in the database, meaning any authenticated user can read every document, including those they should not have access to. This violates the principle of least privilege and can lead to unauthorized data exposure.

Answer

Only the owner can write to documents

Answer

The rules do not explicitly allow delete operations

Answer

The wildcard path matches all databases

Question 2

A company notices that their Cloud SQL for PostgreSQL instance, as shown in the exhibit, frequently runs out of storage, causing downtime. They have set up automated backups with point-in-time recovery. What is the most likely cause of the storage issue?

Accepted Answer

Transaction logs for point-in-time recovery are consuming disk space.. Cloud SQL for PostgreSQL uses transaction logs (WAL files) to enable point-in-time recovery (PITR). These logs accumulate on the disk until they are automatically removed, but if the rate of log generation exceeds the cleanup rate or if the backup retention period is long, the logs can fill the disk, causing storage exhaustion and downtime.

Answer

The activation policy is set to ALWAYS, causing continuous writes.

Answer

The instance tier (db-custom-4-15360) is too low for the workload.

Answer

The data disk type (PD_SSD) is not suitable for PostgreSQL.

Question 3

A news website uses Cloud SQL for MySQL for content management. They experience slow reads during breaking news events. They have a single primary instance in us-east1. They need to improve read scalability globally. They also want to ensure data is backed up in another region. What should they do?

Accepted Answer

Add cross-region read replicas in multiple regions and use replica read for queries.. Adding cross-region read replicas in multiple regions allows the website to offload read queries to replicas located closer to global users, reducing latency during traffic spikes. Cloud SQL for MySQL supports cross-region replicas, which also provide a backup copy of data in another region for disaster recovery, meeting both scalability and backup requirements without changing the database engine.

Answer

Use Cloud Spanner with multi-region configuration.

Answer

Enable automatic failover to a standby instance in another region.

Answer

Use Bigtable for content storage.

Question 4

Refer to the exhibit. The application requires low-latency reads for users in Europe. The current cluster is in us-central1. What should they do?

Accepted Answer

Add a new cluster in a European region (e.g., europe-west1).. Adding a new cluster in a European region (e.g., europe-west1) is correct because it places data physically closer to users, reducing network latency for read operations. In a multi-region deployment, the application can read from the nearest cluster, achieving low-latency reads without changing the existing cluster's configuration. This approach leverages geographic proximity to minimize round-trip time (RTT) for European users.

Answer

Increase the number of nodes in the existing cluster.

Answer

Use a multi-cluster instance with existing cluster.

Answer

Change the storage type to SSD in the existing cluster.

Question 5

You are optimizing a Cloud SQL for MySQL instance for an OLTP application. You observe frequent buffer pool contention and high disk reads per second. The instance has 16 vCPUs and 120 GB memory. What is the most effective initial tuning action?

Accepted Answer

Increase innodb_buffer_pool_size to 80% of available memory.. Buffer pool contention and high disk reads per second are classic symptoms of an undersized InnoDB buffer pool. For a dedicated Cloud SQL for MySQL OLTP instance with 120 GB memory, increasing innodb_buffer_pool_size to 80% (96 GB) maximizes cached data and index pages in memory, reducing physical disk I/O and contention. This directly addresses the root cause—insufficient memory for the working set—without introducing side effects like connection overhead or query cache invalidation.

Answer

Increase the max_connections parameter to handle more concurrent sessions.

Answer

Reduce max_heap_table_size to avoid in-memory temp tables spilling to disk.

Answer

Enable query cache to cache SELECT results.

Question 6

A healthcare company uses Cloud SQL for MySQL for patient records. They need to export data for a compliance audit. They must ensure the export includes all changes within a specific time window (e.g., last 24 hours). They have binary logging enabled. What is the best method to obtain a consistent snapshot of the data as of the audit time?

Accepted Answer

Use Cloud SQL clone to create a new instance from a point-in-time, then export from clone.. Cloud SQL's clone feature can create a new instance from a specific point-in-time using binary logs, providing a consistent snapshot of the database as of the audit time. This ensures all changes within the last 24 hours are captured without impacting the production instance, and the export can then be performed from the clone.

Answer

Use Database Migration Service's continuous export.

Answer

Use mysqldump to export at the audit time.

Answer

Use Cloud SQL's export feature with a specific backup-id.

Question 7

Your Cloud Spanner instance has high latency for point reads. The workload is evenly distributed across all nodes. Which metric should you examine first to identify the bottleneck?

Accepted Answer

CPU utilization per node. High latency for point reads in Cloud Spanner, even with evenly distributed workload, often points to CPU saturation on individual nodes. Spanner uses a shared-nothing architecture where each node handles a portion of the data and queries; if CPU utilization per node is high, it indicates that the node is overloaded, causing queuing and increased latency. This metric directly reflects processing capacity and is the first place to look for a bottleneck in point-read performance.

Answer

Storage utilization

Answer

Rows returned per second

Answer

Number of committed nodes

Question 8

A Cloud SQL for PostgreSQL instance is experiencing high CPU usage during peak hours. Query Insights shows that a complex reporting query is causing full table scans on a large table. The query filters on a column used in JOINs. Which optimization should be applied first?

Accepted Answer

Add a composite index on the filtered column and join columns.. Adding a composite index on the filtered column and the join columns directly addresses the root cause of the full table scans. Query Insights indicates the query filters on a column used in JOINs; a composite index covering both the filter and join columns allows PostgreSQL to perform an Index Scan instead of a sequential scan, reducing CPU usage without requiring additional infrastructure or data duplication.

Answer

Increase the instance size.

Answer

Create a read replica for reporting.

Answer

Use query rewriting with materialized views.

Question 9

A company has a GKE cluster with cluster autoscaler enabled. They notice that after a batch job completes, the cluster takes a long time to scale down, leaving idle nodes running and incurring costs. Which configuration change would reduce the scale-down delay?

Accepted Answer

Reduce the scale-down delay from 10 minutes to 2 minutes.. The cluster autoscaler has a default scale-down delay (e.g., 10 minutes for unneeded nodes). Reducing this delay causes nodes to be removed sooner after becoming idle.

Answer

Enable node auto-provisioning.

Answer

Increase the min node count to match the peak.

Answer

Decrease the max node count.

Question 10

A company uses Pub/Sub to ingest events from multiple services. They notice a backlog of unacknowledged messages and want to increase the throughput of their subscriber. The subscriber is a single process running on Compute Engine. What is the most effective way to increase throughput?

Accepted Answer

Add more subscriber instances running in parallel.. Running multiple subscriber instances (parallel pull consumers) increases the overall throughput by allowing more messages to be pulled and processed concurrently.

Answer

Decrease the flow control max outstanding messages.

Answer

Increase the acknowledgement deadline to 10 minutes.

Answer

Enable ordering keys on the subscription.

Question 11

A BigQuery table stores daily sales data. The team commonly queries data for a specific date range. Which schema optimization will reduce query cost and improve performance?

Accepted Answer

Partition the table by date column. Partitioning the table by the date column allows BigQuery to prune entire partitions when querying a specific date range, drastically reducing the amount of data scanned. Since BigQuery charges by the bytes processed, this directly lowers query cost and improves performance by reading only the relevant partitions.

Answer

Create a view over the table

Answer

Create a materialized view with a filter on date

Answer

Cluster the table by date column

Question 12

Which TWO are valid methods to define an SLI for a request-driven service? (Choose 2.)

Accepted Answer

Proportion of requests completed within a latency threshold. Option A is correct because an SLI for a request-driven service is commonly expressed as the proportion of requests whose latency falls within a defined threshold (e.g., requests served faster than 300 ms), which directly measures user-perceived responsiveness. Option E is correct because the proportion of requests returning a successful HTTP status (e.g., non-5xx responses) is a standard availability/error-rate SLI for request-driven services. Options B and D describe resource-level metrics (CPU utilization and instance count) that reflect infrastructure capacity rather than the service's request outcomes, so they are not valid SLIs. Option C, code commits per week, is a development activity metric and has no bearing on service-level performance or reliability.

Answer

CPU utilization of the server

Answer

Number of code commits per week

Answer

Number of running instances

Question 13

Which TWO statements about Cloud Spanner schema changes are correct? (Choose 2)

Accepted Answer

Schema changes are applied asynchronously and may take minutes to hours to complete.. Option A is correct because Cloud Spanner schema changes are applied asynchronously in the background, and depending on the size of the database and the nature of the change, they can take from minutes up to hours to complete. Option C is correct because ALTER TABLE statements in Cloud Spanner are non-blocking: existing reads and writes continue to work while the schema change is being applied, so the database remains fully in use. Option B is incorrect because Cloud Spanner never requires an instance restart to apply schema changes. Option D is incorrect because Cloud Spanner supports both adding and dropping columns via ALTER TABLE. Option E is incorrect because creating a secondary index is an online, non-blocking operation that does not take the table offline.

Answer

Schema changes require the instance to be restarted.

Answer

You can only add columns, not drop them.

Answer

Creating a secondary index requires taking the table offline.

Question 14

A Cloud Pub/Sub subscription is used to ingest real-time events. The subscriber's processing rate is slower than the publish rate, causing messages to back up. The team needs to increase throughput without losing messages. Which three actions should they take? (Choose three.)

Accepted Answer

Increase the number of subscriber clients (parallel pull consumers). Option A is correct because Pub/Sub scales throughput horizontally: adding more subscriber clients (parallel pull consumers) increases the aggregate number of concurrent StreamingPull streams and message deliveries, so the backlog drains faster without dropping messages. Option D is correct because raising the max outstanding messages (flow control) per subscriber client allows more messages to be in flight concurrently, increasing per-client throughput as long as the client can process them. Option E is correct because an asynchronous puller (e.g., the high-level asynchronous client libraries) fetches and processes messages concurrently rather than blocking on each message, which substantially raises processing throughput. Option B is not appropriate because ordering keys enforce per-key ordered delivery, which can serialize processing and reduce throughput rather than increase it. Option C is not required because switching to push does not by itself raise throughput; push and pull are both viable delivery modes, and the bottleneck is subscriber processing capacity.

Answer

Set ordering keys on the subscription

Answer

Use a push subscription instead of pull

Question 15

A company is migrating a PostgreSQL database to Cloud SQL using DMS continuous migration. After the full dump, the CDC phase is replicating changes. To prepare for cutover, which TWO actions should the engineer take? (Choose 2)

Accepted Answer

Verify that the DMS migration job lag is 0 seconds.. Option A is correct because during the CDC phase of a Cloud SQL DMS continuous migration, the migration job lag (replication delay) must be verified as 0 seconds to confirm that all changes from the source have been applied to the destination before cutover, ensuring no data loss. Option E is correct because quiescing all write operations to the source database stops new transactions from being generated, allowing the CDC replication to fully catch up and the destination to become consistent with the source at the moment of cutover. Option B is not needed because binary logging is a MySQL concept; PostgreSQL uses logical replication (e.g., logical decoding via the pgoutput plugin), and DMS for PostgreSQL does not require enabling binary logging. Option C is not required at cutover because the full dump has already been completed and CDC is handling ongoing changes; taking another full backup is unnecessary and could introduce inconsistency. Option D is incorrect because deleting the DMS migration job would stop replication entirely, which is the opposite of what is needed to prepare for a controlled cutover.

Answer

Enable binary logging on the source.

Answer

Take a full backup of the source.

Answer

Delete the DMS migration job to stop replication.

Question 16

A site reliability engineer is leading a blameless postmortem for an incident. Which THREE practices should be included? (Choose 3.)

Accepted Answer

List contributing factors. Option A (List contributing factors) is correct because blameless postmortems focus on the systemic and environmental conditions—such as configuration drift, missing alerts, or unclear runbooks—that combined to cause the incident, rather than a single fault. Option C (Assign action items with owners and due dates) is correct because a postmortem is only valuable if it produces concrete, tracked remediation work with accountability, ensuring findings translate into reliability improvements. Option D (Use the 5 Whys technique to find root cause) is correct because iteratively asking 'why' drives the analysis past surface symptoms to underlying process, tooling, or design weaknesses that can actually be fixed. Option B (Determine the incident commander for future incidents) does not belong because incident command roles are established during incident response planning, not as a postmortem output. Option E (Identify the person who caused the incident) does not belong because it directly contradicts the blameless principle, which seeks to remove fear and encourage honest disclosure rather than assign individual blame.

Answer

Determine the incident commander for future incidents

Answer

Identify the person who caused the incident

Question 17

An SRE team wants to implement a blameless postmortem culture after incidents. Which TWO practices are essential for a blameless postmortem?

Accepted Answer

Conducting a root cause analysis using the 5 Whys technique. Option C is correct because the 5 Whys technique is a core blameless postmortem tool that iteratively asks 'why' to trace an incident back to systemic or process causes rather than stopping at human error, which aligns with the goal of learning from failure without blaming individuals. Option E is correct because a postmortem is only actionable when it produces concrete action items with assigned owners and due dates, ensuring the identified systemic improvements are actually implemented and tracked to completion. Options A, B, and D do not belong: escalating to executives is an incident-management communication step, not a defining practice of blameless postmortems; monetary fines directly contradict the blameless principle by punishing people; and identifying the individual responsible is precisely the blame-oriented behavior a blameless culture seeks to eliminate in favor of systemic analysis.

Answer

Escalating the incident to the executive team

Answer

Assigning monetary fines to the team responsible

Answer

Identifying the individual responsible for the incident

Question 18

A company uses Cloud Spanner for a global application. They need to capture real-time changes to certain tables for downstream processing. Which two services or features can be used together to achieve this? (Choose TWO.)

Accepted Answer

Spanner change streams. Spanner change streams (A) are the native Cloud Spanner feature that records data changes (inserts, updates, deletes) to specified tables and columns in real time, providing the change data capture source required by this scenario. Cloud Dataflow (E) is the correct companion service because it offers a built-in Spanner change streams connector (via the SpannerIO API) that reads the change stream records and processes or routes them for downstream consumption. Together, change streams plus Dataflow form the standard Google-recommended pattern for streaming Spanner changes to downstream systems. Cloud Pub/Sub Lite (B) is a messaging service, not a change capture mechanism, and cannot read Spanner table changes on its own. Cloud SQL for PostgreSQL (C) is an unrelated managed database and has no role in capturing Cloud Spanner changes. Cloud Functions (D) is an event-driven compute service that lacks a native Spanner change stream source, so it cannot directly consume change stream data without an intermediary such as Dataflow.

Answer

Cloud Pub/Sub Lite

Answer

Cloud SQL for PostgreSQL

Answer

Cloud Functions

Question 19

An SRE team is defining SLIs for a data pipeline that ingests events from a pub/sub topic and writes to BigQuery. Which two metrics are good SLIs for pipeline freshness? (Choose TWO.)

Accepted Answer

Age of the oldest unprocessed message in the subscription. Option B is correct because the age of the oldest unprocessed message in the subscription directly measures pipeline freshness — it quantifies how far behind the pipeline is in consuming the backlog, which is exactly what a freshness SLI should capture. Option E is correct because the end-to-end latency between event ingestion and availability in BigQuery measures the time it takes for an event to become queryable, which is the canonical definition of freshness for a data pipeline. Option A is not a freshness metric; CPU utilization of Dataflow workers is a resource/saturation signal, not a measure of data staleness. Option C reflects throughput (volume of events published per second), not how fresh the data is. Option D is a capacity/volume metric (total rows in BigQuery) and says nothing about how current or delayed the data is.

Answer

CPU utilization of Dataflow workers

Answer

Number of events published per second

Answer

Total number of rows in BigQuery

Question 20

An SRE team wants to conduct chaos engineering on a GKE cluster to test resilience. Which TWO tools or services can be used? (Choose 2.)

Accepted Answer

Traffic Director fault injection. Traffic Director fault injection (C) is correct because Traffic Director supports fault injection policies that can inject delays, aborts, and other failures into traffic routed to services, which is a valid way to test the resilience of workloads on a GKE cluster. Chaos Mesh (D) is correct because it is a Kubernetes-native chaos engineering platform that can be installed on GKE to run controlled experiments such as pod failures, network partitions, and stress tests. Cloud Scheduler (A) is only a cron-based job scheduler and does not provide chaos experiment capabilities. Cloud Build (B) is a CI/CD build service, not a chaos engineering tool. Cloud Run for Anthos (E) is a serverless application platform and does not itself provide fault injection or chaos testing features.

Answer

Cloud Scheduler

Answer

Cloud Build

Answer

Cloud Run for Anthos