# Free COF-C03 practice questions

> 10 free COF-C03 practice questions with the correct answer and an explanation.

Page: https://devopscodelab.com/exams/snowflake-cof-c03/free-practice-questions/
Practice exams: https://devopscodelab.com/exams/snowflake-cof-c03/

### Question 1 of 10

A data science team trains a scikit-learn model in a Python stored procedure on a large dataset, and the procedure fails with out-of-memory errors on a standard Large warehouse. Which change is MOST appropriate?

- A. Enable the query acceleration service so the training step can borrow serverless compute
- B. Run the procedure on a Snowpark-optimized warehouse, which provides much more memory per node
- C. Increase MAX_CONCURRENCY_LEVEL so the procedure is allotted a greater share of the cluster
- D. Convert the warehouse to a multi-cluster warehouse so the procedure can spread across clusters

**Answer:** B

**Explanation:** Snowpark-optimized warehouses are designed for memory-intensive workloads such as model training in stored procedures and offer substantially more memory per node than standard warehouses. Multi-cluster warehouses add concurrency, not memory for a single procedure; query acceleration offloads portions of eligible scans, not Python training; and raising MAX_CONCURRENCY_LEVEL lets more queries share resources rather than giving one more memory.

### Question 2 of 10

A table stores intermediate ETL output that can easily be regenerated. The team wants to avoid Fail-safe storage costs but keep the table available to other sessions. Which table type should they use?

- A. An external table pointing to a stage
- B. A temporary table
- C. A transient table
- D. A permanent table with a 0-day retention period

**Answer:** C

**Explanation:** Transient tables persist across sessions and are visible to other users, but they have no Fail-safe period, which removes Fail-safe storage costs. Temporary tables exist only for the session that created them, permanent tables always have 7 days of Fail-safe even with zero Time Travel, and external tables only reference files in a stage rather than storing ETL output inside Snowflake.

### Question 3 of 10

An analyst runs the exact same SELECT statement twice within an hour. The underlying table did not change, and the second run returned instantly without the warehouse being resumed. Which cache served the second request?

- A. The metadata cache of micro-partition statistics
- B. The query result cache held in the cloud services layer
- C. The remote storage cache that keeps recently read micro-partitions in object storage
- D. The local disk cache of the virtual warehouse that ran the first query

**Answer:** B

**Explanation:** Persisted query results live in the cloud services layer, so an identical query against unchanged data can be answered without any running warehouse. The warehouse local disk cache requires the warehouse to be running, the metadata cache answers only certain metadata-based queries such as COUNT or MIN/MAX, and there is no separate remote storage cache of micro-partitions.

### Question 4 of 10

A financial company on Business Critical Edition must ensure that Snowflake cannot decrypt its data if the company revokes access to its own key. Which feature provides this control?

- A. Client-side encryption of files in an external stage with a master key
- B. Periodic rekeying of data encrypted with keys that are older than one year
- C. Tri-Secret Secure using a customer-managed key in the cloud provider's KMS
- D. Snowflake's hierarchical key model with automatic rotation of keys every 30 days

**Answer:** C

**Explanation:** Tri-Secret Secure combines a Snowflake-maintained key with a customer-managed key to create a composite master key; revoking the customer key makes the data unreadable. Periodic rekeying and automatic rotation strengthen Snowflake-managed encryption but leave Snowflake in control of the keys, and client-side stage encryption protects only staged files, not data in tables.

### Question 5 of 10

A company wants a secondary account in another region that can take over if the primary region has an outage, including databases, users, roles, and warehouses. Which feature fits BEST?

- A. A reader account in the secondary region with secure views over data
- B. A share containing those objects created for the secondary account
- C. Zero-copy clones of each database made in the secondary account
- D. A failover group replicating those objects to the secondary account

**Answer:** D

**Explanation:** Failover groups replicate databases and account objects such as users, roles, and warehouses to another account and allow promotion of the secondary for business continuity (Business Critical Edition or higher). Shares cannot contain users or roles, clones work only within one account, and reader accounts are for data consumers.

### Question 6 of 10

A security team does not want cloud storage credentials stored in stage definitions. Which object should be used to give external stages access to an S3 bucket?

- A. An API integration that calls the S3 REST endpoint
- B. A storage integration that uses an IAM role
- C. A secret holding an access key
- D. A network policy that allows the bucket's IP range

**Answer:** B

**Explanation:** A storage integration stores a generated identity (for AWS, an IAM role trust relationship) so stages can reach the bucket without embedding credentials, and the integration can be limited to specific locations. Network policies control inbound client access, API integrations are used for external functions and Git repositories, and storing an access key in a secret still relies on long-lived credentials.

### Question 7 of 10

A team receives Parquet files with a schema they do not yet know and wants Snowflake to derive the column definitions and create the table automatically. Which combination achieves this?

- A. CREATE EXTERNAL TABLE with AUTO_REFRESH set to TRUE
- B. COPY INTO with MATCH_BY_COLUMN_NAME on a table with no columns
- C. CREATE TABLE ... CLONE pointing at the staged Parquet files
- D. CREATE TABLE ... USING TEMPLATE over INFER_SCHEMA results

**Answer:** D

**Explanation:** INFER_SCHEMA reads staged files and returns column names and types, and CREATE TABLE ... USING TEMPLATE turns that result into a table definition. Cloning works only between Snowflake objects, an external table with auto refresh keeps data outside Snowflake rather than creating a matching table, and COPY needs an existing table with columns to match against.

### Question 8 of 10

Analysts run highly selective point lookups such as WHERE SESSION_ID = 'abc123' on a large table, where SESSION_ID has very high cardinality and is not used for clustering. Which feature is designed to speed up these queries?

- A. A materialized view grouped by SESSION_ID
- B. The search optimization service
- C. The query acceleration service
- D. A multi-cluster warehouse in auto-scale mode

**Answer:** B

**Explanation:** Search optimization builds a persistent search access path that lets point lookups and selective filters on high-cardinality columns skip most micro-partitions. Query acceleration offloads parts of large scan-heavy queries, a materialized view grouped by the lookup column would roughly duplicate the table, and extra clusters improve concurrency rather than single-query pruning.

### Question 9 of 10

A warehouse runs a mix of short dashboard queries and occasional ad hoc queries that scan huge amounts of data with selective filters. The outliers slow down the warehouse. Which feature lets these outlier queries borrow serverless compute without resizing the warehouse?

- A. A Snowpark-optimized warehouse assigned to the dashboard users
- B. The query acceleration service enabled on the warehouse
- C. Automatic Clustering on the tables used by the outlier queries
- D. The search optimization service enabled on the scanned tables

**Answer:** B

**Explanation:** Query acceleration offloads eligible portions of large scan and filter work to shared serverless compute, bounded by a scale factor, so outliers finish faster without upsizing. Search optimization targets point lookups, clustering improves pruning over time rather than adding compute, and Snowpark-optimized warehouses give more memory per node for different workloads.

### Question 10 of 10

An organization has accounts in several business units in different regions, and data owners want to share curated datasets internally across those accounts with discovery and request workflows. Which feature fits BEST?

- A. Reader accounts created by each data owner
- B. Public listings on the Snowflake Marketplace
- C. Direct shares created manually in each account by every data owner
- D. Internal listings in the organization's Internal Marketplace

**Answer:** D

**Explanation:** Internal listings published to the organization's Internal Marketplace let business units discover and request curated datasets across accounts and regions with governance. Public listings expose data outside the organization, reader accounts are for non-Snowflake consumers, and manual direct shares do not provide discovery and do not span regions.

Ready for a timed run? Take the free 15-question COF-C03 mini exam: https://devopscodelab.com/app/exams/snowflake-cof-c03/mini/

DevOpsCode Lab is not affiliated with Snowflake. COF-C03 and related names are trademarks of their owners. We write every question ourselves; these are not real exam questions.
