• 60 questions

Google Data Engineer Certification Practice Test

Scenario questions for the Professional Data Engineer exam, each answer linked to the Google Cloud documentation page behind it.

A team wants to turn on Pub/Sub exactly-once delivery. Which subscription type supports it?

Pick an answer, then press Check answer. Then you see Correct or Not quite, with the reason. Keys: A to D pick, Enter checks.

Print covers the next 50 questions. Share this set copies a link to the same questions in the same order.

Exam mode: 50 questions, 120 minutes.

More options

Unofficial practice questions, not affiliated with Google LLC. Bank checked October 9, 2026.

Next step Next: AWS Cloud PractitionerAnother IT certification? Pick it

All 60 questions

Designing data processing systems 13

  • A company wants every new resource in its organization to be created only in European locations. Which control sets that limit? Organization Policy documentation: Restrict resource locations
  • A data center must move a very large archive to Cloud Storage, and its network link is too slow to upload it in time. Which offering ships the data physically? Transfer Appliance documentation: Transfer Appliance overview
  • A marketing team wants data from a supported SaaS platform loaded into BigQuery on a schedule, without writing code. Which service does this? BigQuery documentation: Introduction to the BigQuery Data Transfer Service
  • A payments ledger spans several regions and needs relational transactions that behave as if they ran one after another. Which database documents that guarantee? Spanner documentation: TrueTime and external consistency
  • A pipeline must tokenize 16-digit account numbers so that each token keeps the length and the character set of the original. Which Sensitive Data Protection method does that? Sensitive Data Protection documentation: Pseudonymization
  • A regulator requires that the company, not Google, control the key encryption keys that protect its BigQuery tables. What should the data engineer set up? BigQuery documentation: Customer-managed Cloud KMS keys
Show all 60 questions (54 more)

Designing data processing systems, continued

  • A team is moving a PostgreSQL database to Google Cloud and wants a continuous migration with low downtime, without running replication servers itself. Which service fits? Database Migration Service documentation: Database Migration Service overview
  • A team needs change data capture: each change made in a source database is read and passed on as it happens. Which service is documented as serverless CDC and replication? Datastream documentation: Datastream overview
  • An analyst holds BigQuery Data Viewer on a dataset but gets a permission error when starting a query in the project. Which role is missing? BigQuery documentation: BigQuery IAM roles and permissions
  • Analysts keep their BigQuery transformations as SQL in a version-controlled repository and want a query run after each build to check that the data is correct. Which service offers this? Dataform documentation: Dataform overview
  • Analysts may query an employee table, but only the payroll team may read its salary column, which carries a policy tag. Which role lets a payroll analyst read that column? BigQuery documentation: Introduction to column-level access control
  • Regional managers share one dashboard on a single sales table. Each manager must see only the rows of their own region. Which BigQuery feature fits? BigQuery documentation: Introduction to BigQuery row-level security
  • Security wants to lower the risk that BigQuery and Cloud Storage data is copied out to a project outside the company. Which control draws a boundary around those services? VPC Service Controls documentation: Overview of VPC Service Controls

Ingesting and processing the data 15

  • A CI/CD process packages a Dataflow pipeline for reuse as a Flex Template. In what form is the pipeline packaged? Dataflow documentation: Dataflow templates
  • A dashboard needs a one-minute running average that is recalculated every thirty seconds. Which window type overlaps in this way? Dataflow documentation: Streaming pipelines
  • A few malformed messages are redelivered again and again because the subscriber can never acknowledge them. Where can Pub/Sub forward them? Pub/Sub documentation: Dead-letter topics
  • A short process calls a Cloud Run service, then a BigQuery job, then an external HTTP API, in a fixed order. Which fully managed service runs steps like these in the order you define? Workflows documentation: Workflows overview
  • A streaming Dataflow job must be stopped for good, but the data it has already buffered should still be processed. Which way of stopping does that? Dataflow documentation: Stop a running Dataflow pipeline
  • A streaming pipeline must count events in back-to-back 30-second intervals that never overlap. Which window does the Dataflow documentation name for this? Dataflow documentation: Streaming pipelines
  • A streaming writer uses the BigQuery Storage Write API and needs each record to be readable as soon as it is written. Which stream type gives that? BigQuery documentation: Introduction to the Storage Write API (gRPC)
  • A subscriber is an HTTPS service that should receive each Pub/Sub message as a request sent by the Pub/Sub server, with no polling. Which subscription type is that? Pub/Sub documentation: Choose a subscription type
  • A team wants to turn on Pub/Sub exactly-once delivery. Which subscription type supports it? Pub/Sub documentation: Exactly-once delivery
  • Clickstream events should be grouped per burst of user activity, with a new group starting after a quiet gap. Which window type is this? Dataflow documentation: Streaming pipelines
  • In a Dataflow streaming job, an element arrives with a timestamp inside a window after the watermark has passed the end of that window. How is the element classed? Dataflow documentation: Streaming pipelines
  • Messages on a Pub/Sub topic need to land in a BigQuery table, and the team wants no pipeline in between. What should it create? Pub/Sub documentation: Choose a subscription type
  • Objects in Amazon S3 buckets must be moved into Cloud Storage without custom scripts. Which service is made for this? Storage Transfer Service documentation: Storage Transfer Service overview
  • Pipeline worker VMs have internal IP addresses only, yet must call Google APIs such as Cloud Storage. Which subnet feature allows that? Virtual Private Cloud documentation: Private Google Access
  • Updates for the same customer must reach subscribers in the order they were published. What does the publisher attach to the related messages? Pub/Sub documentation: Order messages

Storing the data 12

  • A Bigtable table stores sensor readings, and its row keys begin with the reading's timestamp. Writes are piling onto one node. What does the schema guidance recommend? Bigtable documentation: Schema design best practices
  • A bucket holds objects whose access pattern nobody can predict. Which feature has Cloud Storage move objects between storage classes by itself? Cloud Storage documentation: Storage classes
  • A line-of-business application uses a SQL Server database, and the team wants a fully managed relational service for that engine. Which service supports it? Cloud SQL documentation: Cloud SQL overview
  • A Spanner table uses a timestamp as the first part of its primary key and has a write hotspot. Which key does the schema guidance recommend? Spanner documentation: Schema design best practices
  • An application writes and reads huge volumes of rows by a single key and needs low latency at high throughput. Which service does its documentation call ideal for this? Bigtable documentation: Bigtable overview
  • Analysts must query Parquet files in Cloud Storage through BigQuery, but must not be given access to the bucket itself. Which table type separates the two kinds of access? BigQuery documentation: Introduction to BigLake tables
  • Backups are kept only for disaster recovery and are almost never read. Which Cloud Storage class offers the lowest storage cost for them? Cloud Storage documentation: Storage classes
  • Before writing data quality rules, a team wants statistics for its BigQuery tables, such as common values and how the data is distributed. What does Dataplex offer for this? Dataplex documentation: About data profiling
  • How many clustering columns can a BigQuery clustered table specify at most? BigQuery documentation: Introduction to clustered tables
  • Most queries on a very large events table filter on the event date. Which table design lets BigQuery skip the data for dates that the filter rules out? BigQuery documentation: Introduction to partitioned tables
  • Orders and their line items are always queried together, and the join is slow. Which BigQuery modelling practice does the documentation recommend to denormalize them? BigQuery documentation: Use nested and repeated fields
  • Two Object Lifecycle Management rules match the same object at once: one deletes it and one changes its storage class. Which action is taken? Cloud Storage documentation: Object Lifecycle Management

Preparing and using data for analysis 9

  • A partner team should query a filtered subset of a dataset without being granted access to the source dataset. What should the data administrator create? BigQuery documentation: Authorized views
  • A query is much slower than expected, and the engineer wants to see which step is taking the time. What does BigQuery provide for this? BigQuery documentation: Get query performance insights
  • A retrieval-augmented generation system keeps text embeddings in BigQuery and must find, with SQL, the rows closest in meaning to a query. Which function does that? BigQuery documentation: Introduction to embeddings and vector search
  • A subscriber finds a listing in BigQuery sharing (formerly Analytics Hub) and subscribes to it. What is created in the subscriber's project? BigQuery documentation: Introduction to BigQuery sharing
  • An analyst who knows SQL wants to train a model on data that is already in BigQuery. Which statement creates the model in BigQuery ML? BigQuery documentation: Introduction to ML in BigQuery
  • Images and PDF files sit in a Cloud Storage bucket, and the team wants to analyze this unstructured data from BigQuery. Which table type is made for it? BigQuery documentation: Introduction to object tables
  • Interactive reports on BigQuery need faster responses. Which service accelerates queries by caching the most frequently used data in memory? BigQuery documentation: Introduction to BI Engine
  • Queries name a base table, yet BigQuery reroutes them to a precomputed object whenever that object can resolve part of the query. Which object has this smart tuning? BigQuery documentation: Introduction to materialized views
  • Support staff may query a customer table, but should get an obscured value in the email column instead of an access error. Which BigQuery feature does this? BigQuery documentation: Introduction to data masking

Maintaining and automating data workloads 11

  • A BigQuery dataset is replicated to a second region for disaster recovery. Which statement about the secondary replica is true? BigQuery documentation: Cross-region dataset replication
  • A Cloud SQL for MySQL instance is configured for high availability. When is a transaction reported as committed? Cloud SQL for MySQL documentation: About high availability
  • A Dataproc cluster needs more compute for a few hours but no more HDFS storage. Why do secondary workers fit? Dataproc documentation: Secondary workers
  • A faulty job overwrote rows in a BigQuery table three days ago. With default settings, how can the engineer read the table as it was before? BigQuery documentation: Data retention with time travel and fail-safe
  • A nightly report does not need to start at once. Which query job type has lower priority than the default and may wait in the queue when compute resources are fully used? BigQuery documentation: Run a query
  • A nightly Spark job should run on Dataproc on a cluster that exists only while the job runs. What does a workflow template with a managed cluster do? Dataproc documentation: Overview of workflow templates
  • A team schedules its pipelines in Cloud Composer. In what kind of file is each Airflow DAG defined? Cloud Composer documentation: Write Airflow DAGs
  • An administrator wants to list the queries that ran in a project this week and what each one used. Which view holds near real-time metadata about the project's jobs? BigQuery documentation: JOBS view
  • On a streaming Dataflow job, the data freshness metric keeps rising. What does that metric measure? Dataflow documentation: Dataflow job metrics
  • Under capacity-based pricing, test queries are slowing production workloads. How can the administrator stop the two competing for slots? BigQuery documentation: Introduction to workload management
  • With on-demand pricing, an engineer wants a single query to fail instead of running when it would read more than a set amount of data. Which setting does that? BigQuery documentation: Estimate and control costs

How this list was compiled: Own questions, written to the five sections and weightings of the Professional Data Engineer certification exam guide that Google publishes at https://services.google.com/fh/files/misc/professional_data_engineer_exam_guide_english.pdf (read 2026-10-09), linked from https://cloud.google.com/learn/certification/data-engineer (read 2026-10-09), which gives the exam length and format. Facts were read on 2026-10-09 in Google Cloud's public documentation at docs.cloud.google.com. Each question names the page it was written from and links to it. Licence: Own work. Every question, wrong choice and explanation was written for this site from the sections of the exam guide and the public Google Cloud documentation pages named in each question. Nothing is copied or reworded from a Google exam, Google's sample questions, a practice product or a question dump.

How it works

This unofficial Google data engineer certification practice test has 60 of our own questions across the five sections of the Professional Data Engineer exam guide. It is not affiliated with Google LLC, and no question comes from the real exam.

Read the full guide

What the Professional Data Engineer exam asks

Google's exam guide, read on October 9, 2026, gives designing data processing systems about 22% of the exam, ingesting and processing the data 25%, storing the data 20%, preparing and using data for analysis 15%, and maintaining and automating data workloads 18%.

Names that have changed

The exam guide says Dataproc, Cloud Composer and Dataplex, and so do these questions. Google's documentation pages for them are now titled Managed Service for Apache Spark, Managed Service for Apache Airflow and Knowledge Catalog. The real exam also has multiple-select questions; every question here has one right answer.

  • Draining a Dataflow job finishes the data it has buffered. Canceling a job stops processing that data.
  • Pub/Sub exactly-once delivery works with pull subscriptions only.
  • A Bigtable row key that starts with a timestamp pushes sequential writes onto a single node.

Questions and answers

How many questions are on the Google Professional Data Engineer exam?

40 to 50 multiple choice and multiple select questions in 2 hours, by Google's certification page read on October 9, 2026. Exam mode here draws 50 questions on a 120-minute clock.

What is the passing score for the Google data engineer certification?

Google's certification page lists the length, fee, languages and format of the exam and gives no passing score. Exam mode here shows your percent right in each section and no pass or fail.

How much does the Google data engineer certification exam cost?

The registration fee is $200 plus tax where applicable, and the certification is valid for 2 years, by Google's certification page read on October 9, 2026. Check that page before you register.

Is this Google data engineer practice test free?

Yes. All 60 questions and their explanations are free, with no account.

More tests

Show 7 more

Browse Safety and Trades