Professional Data Engineer · Sample questions
10 real questions from the CertifyDeck Professional Data Engineer bank, spread across the domains of the official Google Cloud Certified Professional Data Engineer skills outline, with the correct answer and a full explanation of why the other options are wrong. They are original questions written from the public syllabus, not exam dumps, and they are the same style you will find in the practice tests and mock exams.
Correct answer: C. Create a VPC Service Controls perimeter around the approved projects that restricts the BigQuery and Cloud Storage APIs.
VPC Service Controls perimeters protect the resources of the services you specify and prevent data from being copied to unauthorized resources outside the perimeter with operations such as gcloud storage cp or bq mk, as a layer independent of IAM. Removing roles would also stop legitimate work and still not limit where authorized users copy data. The resource locations constraint controls the regions where resources are created, not which projects data can be copied to, and Private Google Access only provides a private path to Google APIs.
Correct answer: A. Create a regional key in europe-west1, grant the Dataflow and Compute Engine service accounts the CryptoKey Encrypter/Decrypter role, and pass the key with the dataflow_kms_key option.
Correct answer: A. Enforce a domain-restricted sharing organization policy, such as iam.managed.allowedPolicyMembers, on the organization node.
Correct answer: C. Add shared_views to raw_sales as an authorized dataset, and grant the analysts Data Viewer on the shared_views dataset only.
Correct answer: B. Move the preference columns to their own column family whose garbage collection keeps one version.
Correct answer: C. Create an object table over the bucket with a Cloud resource connection, and run inference on it.
Object tables are read-only tables over unstructured objects in Cloud Storage that you can analyze with BigQuery ML inference or remote functions and then join with structured data, and their access delegation means users are granted the object table instead of the underlying objects. BigLake tables are for structured formats such as Parquet, Avro or CSV, so they cannot represent JPEG images. Loading every image with Dataflow copies the data, and granting bucket access breaks the stated access requirement.
Correct answer: B. Give the reservation enough baseline slots so that it does not rely on idle slots.
Baseline slots remain constantly available to their reservation, and Google recommends that production, time-bound workloads use a reservation that is self-reliant in slot capacity instead of relying on idle slots, which can be reclaimed at any time. Leaving ignore_idle_slots at false only keeps the current borrowing behaviour that is causing the unpredictability. Batch priority makes queries more likely to wait in the queue, and on-demand capacity is lower in the capacity prioritization order than edition baselines.
Correct answer: D. Cover the steady baseline with a slot commitment and add autoscaling up to a maximum for the peaks.
Capacity commitments can save money over the pay-as-you-go rate for steady-state workloads and cover baseline slots, while autoscaling adds slots between the baseline and the maximum only when demand requires them. A commitment sized for the peaks pays for capacity that sits idle most of the day. A zero-baseline autoscaling reservation bills the steady workload at pay-as-you-go rates, because commitments don't apply to autoscaled slots, and assigning the ELT projects to None moves them to on-demand billing as well.
Correct answer: C. Use Connected Sheets to connect Google Sheets to the table, and schedule the data refresh.
Connected Sheets lets users analyze BigQuery data inside Google Sheets with charts, formulas and pivot tables without knowing SQL, and it runs the BigQuery queries on request or on a defined schedule instead of relying on manual exports. Nightly CSV exports create the copies the requirement rules out. Querying in BigQuery Studio or through an ODBC tool both require the analysts to write SQL.
Correct answer: C. Share the spreadsheet with the representatives without granting them any access in BigQuery.
Direct access to BigQuery datasets is controlled in BigQuery, so to give users Sheets-only access you share the spreadsheet without granting BigQuery access; those users can analyze the sheet but cannot refresh the data manually or schedule refreshes. Granting BigQuery roles, even only the ability to run jobs, gives the representatives a path to query BigQuery. A VPC Service Controls egress rule is what allows query results to reach spreadsheets from a protected perimeter, not a way to limit what sheet users can do.
Answer each question before opening the explanation, and write down why you discarded the other options. On Professional Data Engineer, most mistakes come from choosing an option that is technically possible but does not meet every requirement in the scenario.
These 10 questions are a taste of the format, not a diagnostic. To know which domains you fail, practice by domain for free and take a timed mock exam that mirrors the official weights and duration.
Start for free, no sign-up or card. Every question includes an explanation like the ones above, and the mock exam reproduces the official Professional Data Engineer format.
Dataflow supports only regional Cloud KMS keys in the same region as the job, both the Dataflow service account and the Compute Engine service account need the CryptoKey Encrypter/Decrypter role, and the key is passed with the dataflow_kms_key pipeline option; the key then protects persistent disks, shuffle state and the temporary buckets. Multi-regional and global keys are not supported for Dataflow pipelines, the launching user is not the identity that encrypts the state, and keys on the sink and bucket leave the job's own state on default encryption.
Domain-restricted sharing is an organization policy that allows only principals from allowed domains or organizations to be granted IAM roles, and a policy set on the organization node is inherited by every current and future project. A VPC Service Controls perimeter restricts API access and data movement but does not stop an external account from being added to an allow policy. An alert only detects the grant after it happens, and a custom role per project does not cover new projects.
An authorized dataset lets all views in one dataset access the data in a second dataset, so you don't need to authorize each view, and users can query the views without access to the source dataset. Authorizing views one by one works but fails the requirement to avoid per-view configuration. Granting Data Viewer on raw_sales gives the direct access that is forbidden, and nightly copies duplicate the raw data into a dataset analysts can read.
Garbage collection policies are set at the column family level, so the documentation recommends putting columns with different retention needs in different column families to limit storage costs; a family that keeps one version stops paying for unneeded old cells. A policy cannot be set on individual columns, which rules out the tempting per-column option. Creating a table per month goes against the advice not to create tables for new time periods, and adding a cluster replicates the data, which adds storage instead of removing it.