GCP-ADP Practice Test 20 — 15 Questions

Question 1

You are using BigQuery and need to query a large dataset that is partitioned by day. Which clause should you always include for efficiency?

Accepted Answer

WHERE clause on the partition column. BigQuery partitioned tables store data in separate segments by partition column (e.g., DATE or TIMESTAMP). Including a WHERE clause on the partition column allows BigQuery to prune partitions and scan only the relevant data, dramatically reducing bytes processed and cost. Without it, the query scans the entire table, which is inefficient and expensive for large datasets.

Answer

ORDER BY

Answer

LIMIT

Answer

GROUP BY

Question 2

A media company ingests clickstream events into Pub/Sub and processes them with Dataflow before loading into BigQuery. They need to reduce the number of duplicate records that reach BigQuery while keeping the pipeline streaming and low-latency. Which two approaches should they use? (Choose two.)

Accepted Answer

Use the BigQuery Storage Write API with exactly-once semantics in the Dataflow sink.. Combining the BigQuery Storage Write API with exactly-once semantics at the sink and stateful deduplication in Dataflow keyed on a stable event or message ID addresses duplicates both at write time and during processing. Together they handle retries, restarts, and at-least-once delivery while keeping the pipeline streaming and low-latency.

Answer

Set the Pub/Sub subscription to exactly-once delivery and rely on it alone to prevent duplicates.

Answer

Increase the number of Dataflow workers to reduce the chance of bundle retries.

Answer

Enable BigQuery streaming inserts and schedule a periodic MERGE job to remove duplicates after the fact.

Question 3

You are deploying a Cloud Composer 2 environment and need to ensure that the DAG files are automatically synchronized to all workers. Where should you store these files?

Accepted Answer

In the 'dags/' folder of the environment's bucket.. In Cloud Composer 2, the environment's Cloud Storage bucket contains a 'dags/' folder that is automatically synchronized to all Airflow worker nodes (and the scheduler). Uploading DAG files to gs://<environment-bucket>/dags/ is the supported, automatic way to distribute DAGs, and Composer's built-in sync process pushes changes to every component.

Answer

Inside a Docker volume configured in the environment's YAML manifest.

Answer

In the local /home/airflow/gcs/dags directory of each worker node.

Answer

In a custom persistent disk attached to the Airflow scheduler.

Question 4

In BigQuery, what does the 'Data Preview' tab show?

Accepted Answer

A sample of the table data.. The Data Preview tab in BigQuery shows a sample of the table data. It allows you to quickly inspect the first few rows of a table without running a query, which is useful for understanding the data structure and content.

Answer

The edit history of the table.

Answer

The billing usage for that table.

Answer

The full schema definition.

Question 5

A Cloud Composer DAG is failing because the Airflow database is locking up during high-concurrency periods. What is the most effective way to address this?

Accepted Answer

Scale up the Cloud Composer environment size to increase database resources.. Cloud Composer's Airflow metadata database (backed by Cloud SQL) is the bottleneck during high-concurrency DAG runs because every task instance, DAG run, and log entry writes to it. Scaling up the Composer environment size increases the Cloud SQL tier (CPU, memory, IOPS), which directly relieves database lock contention and connection saturation. This is the recommended remediation when the scheduler/DB cannot keep up with task state updates.

Answer

Delete all past DAG run logs.

Answer

Increase the number of DAGs in the repository.

Answer

Set the 'parallelism' parameter to a very high value.

Question 6

You are troubleshooting a stuck task in Cloud Composer. You want to see the stdout/stderr logs for that specific task attempt. Where can you find them?

Accepted Answer

In the 'logs/' folder in the environment's GCS bucket.. In Cloud Composer, task logs (stdout/stderr) for each task attempt are stored in the environment's GCS bucket under the logs/ folder, organized by DAG and task. This is the canonical location for retrieving raw task output, especially for tasks that have completed or whose pods have been deleted.

Answer

In the Airflow metadata database directly.

Answer

In the Cloud Composer environment console logs.

Answer

Via the 'kubectl' logs command on the worker pod.

Question 7

You are running a Dataflow job that joins two large datasets. Which join strategy should you avoid to prevent OOM errors?

Accepted Answer

Side Input Join with a large dataset. A Side Input Join involves loading one dataset as a side input into memory, which is then available to each worker. If the side input dataset is large, it can cause out-of-memory (OOM) errors on workers because the entire dataset must fit in memory. Therefore, using a side input join with a large dataset is the strategy to avoid. Other options like CoGroupByKey are designed for large-scale joins.

Answer

Flatten

Answer

CoGroupByKey

Answer

ParDo

Question 8

You are designing a Cloud Workflows pipeline that calls a series of Cloud Functions. The output of the first function must be passed as an argument to the second function, and the output of the second to the third. You need to define the workflow in YAML. Which Workflows syntax correctly assigns the result of a function call to a variable and then uses that variable in a subsequent call?

Accepted Answer

- call: firstFunction
  result: firstResult
- call: secondFunction
  args:
    input: ${firstResult}. In Cloud Workflows, the 'result' keyword captures a call's return value into a variable. Subsequent calls can reference that variable within their 'args' block using ${}. This allows chaining outputs to inputs. The correct syntax uses 'result' to store the output and then 'args' with a named parameter to pass it along.

Answer

- call: firstFunction
  output: firstResult
- call: secondFunction
  args:
    input: ${firstResult}

Answer

- call: firstFunction
  result: firstResult
- call: secondFunction
  args: ${firstResult}

Answer

- call: firstFunction
  result: firstResult
- call: secondFunction
  input: ${firstResult}

Question 9

You have a BigQuery table that is partitioned by day on a DATE column. You need to ensure that data older than 90 days is automatically deleted to reduce storage costs, while keeping the table structure and recent data intact. What should you do?

Accepted Answer

Set a partition expiration on the table to 90 days.. Setting partition expiration on a partitioned table automatically removes partitions older than the specified retention period. This is the simplest and most cost-effective way to delete old data while preserving the table and recent partitions. It requires no manual intervention and integrates with BigQuery's storage management.

Answer

Enable table expiration on the table for 90 days.

Answer

Create a scheduled query that deletes rows older than 90 days daily.

Answer

Use a lifecycle rule on the dataset to delete tables older than 90 days.

Question 10

Which TWO of the following are valid ways to monitor Dataflow performance?

Accepted Answer

Dataflow Monitoring UI. Option C (Dataflow Monitoring UI) is correct because the Dataflow console provides a built-in, real-time monitoring interface that displays job graphs, step-level throughput, autoscaling, worker CPU/memory utilization, and system/worker error logs for a running pipeline. Option D (Cloud Logging) is correct because Dataflow writes pipeline logs (e.g., worker logs, job messages, and errors) to Cloud Logging, where you can filter, search, and create log-based metrics and alerts to monitor performance and failures. Option A (BigQuery logs) is not a Dataflow monitoring mechanism; BigQuery audit logs track BigQuery activity, not Dataflow pipeline execution. Option B (Cloud DNS logs) is unrelated, as DNS logging records name-resolution queries, not Dataflow job metrics. Option E (Manual packet sniffing) is not a supported or practical way to monitor Dataflow performance and would not surface job-level metrics.

Answer

BigQuery logs

Answer

Cloud DNS logs

Answer

Manual packet sniffing

Question 11

A Cloud Composer 2 task runs a Python script that performs a long-running data transformation and regularly exceeds the default worker pod timeout, causing tasks to be killed and marked as failed. The transformation cannot be shortened and must complete within a single task. What should you do to allow the task to run to completion?

Accepted Answer

Increase the execution_timeout and dagrun_timeout parameters on the task and DAG, and verify the Celery worker visibility timeout is sufficient.. Long-running tasks in Cloud Composer 2 are bounded by Airflow execution timeouts and by the Celery worker visibility timeout used by the broker. Increasing execution_timeout and dagrun_timeout, and ensuring the worker visibility timeout is long enough, allows the transformation to finish in one attempt. Adding workers, retrying, or switching operator type does not extend the per-task runtime limit.

Answer

Convert the task to use the KubernetesPodOperator with a larger machine type.

Answer

Set retries=5 and retry_delay=timedelta(minutes=10) on the task.

Answer

Change the task's queue to default and increase the environment's worker count.

Question 12

You need to move an asset from one Dataplex Zone to another. What is the correct procedure?

Accepted Answer

Delete the asset from the first zone and recreate it in the second. In Dataplex, assets are immutable with respect to their zone assignment. To move an asset from one zone to another, you must delete it from the source zone and recreate it in the target zone. There is no direct move operation. This procedure ensures that metadata and permissions are correctly re-established in the new zone.

Answer

Update the IAM policy to include the new zone

Answer

Use the 'move' command in the CLI

Answer

Update the asset metadata to point to the new zone

Question 13

You need to delete all files in a GCS bucket older than 90 days. What is the most efficient approach?

Accepted Answer

Cloud Lifecycle Management. Cloud Lifecycle Management is the most efficient approach because it lets you define a lifecycle rule on the bucket that automatically deletes objects older than 90 days, with no code, no compute, and no manual intervention. It is a native GCS feature designed exactly for age-based object deletion, applying at scale across all objects in the bucket.

Answer

Manual deletion in the console

Answer

Use Dataflow to filter files

Answer

Write a cron job with gsutil

Question 14

You are optimizing a Dataflow job. Which THREE of the following actions can help improve job performance and reduce costs?

Accepted Answer

Enable Dataflow Autoscaling. Option A (Enable Dataflow Autoscaling) is correct because it dynamically adjusts the number of worker instances based on the pipeline's workload, scaling up during peak demand and down during idle periods, which both improves throughput and reduces costs by avoiding over-provisioning. Option B (Use Dataflow Shuffle Service) is correct because it offloads shuffle operations from worker VMs to a dedicated, Google-managed service, reducing worker CPU/memory pressure, eliminating the need for persistent disk-based shuffle storage, and improving performance for pipelines with heavy group-by or join operations. Option D (Use Flexible Resource Scheduling, FlexRS) is correct because it uses a combination of preemptible VMs and advanced scheduling to run batch jobs at a significantly lower cost, at the tradeoff of a longer, non-deterministic completion time, which is ideal for non-time-critical batch workloads. Option C (Disable pipeline monitoring) is not correct because disabling monitoring does not improve performance or reduce costs; it removes visibility into job health and can hinder troubleshooting. Option E (Increase the machine type for all workers) is not correct because uniformly increasing worker machine size can raise costs without guaranteeing better performance, and it is less effective than autoscaling or shuffle offloading for optimizing resource usage.

Answer

Disable pipeline monitoring

Answer

Increase the machine type for all workers

Question 15

Which property in a Workflows YAML definition is required to define the entry point of the workflow?

Accepted Answer

main. In Google Cloud Workflows YAML definitions, the 'main' property defines the entry point of the workflow — the first step that executes when the workflow runs. It is a required top-level field that names the starting step, and all other steps are reached via transitions from it.

Answer

entry_point

Answer

init

Answer

start