Snowflake COF-C03 - SnowPro Core Certification 2026 Exam

 Directory tables in Snowflake are used to store metadata about data files in a stage. They are not separate database objects but are conceptually similar to external tables. Directory tables do not have grantable privileges of their own

In the Admin area of Snowsight, users can analyze query performance, manage Snowflake warehouses, set up and view details about resource monitors, manage users and roles, and administer Snowflake accounts in their organization2.

Snowflake’s virtual warehouses are responsible for query execution. They are clusters of compute resources that execute SQL statements, perform DML operations, and load data into tables

 In a standard multi-cluster warehouse with auto-scaling, a cluster will shut down when, after 2-3 consecutive checks, the system determines that the load on the least-loaded cluster could be redistributed to other clusters. This ensures efficient resource utilization and cost management. References: [COF-C03] SnowPro Core Certification Exam Study Guide

When creating a virtual warehouse in Snowflake, users have the option to set several properties to manage its behavior and resource usage. Two of these optional properties are Auto-suspend and Resource monitor.

Auto-suspend:This property defines the period of inactivity after which the warehouse will automatically suspend. This helps in managing costs by stopping the warehouse when it is not in use.

CREATE WAREHOUSE my_warehouse

WITH WAREHOUSE_SIZE = ' XSMALL '

AUTO_SUSPEND = 300; -- Auto-suspend after 5 minutes of inactivity

Resource monitor:Users can assign a resource monitor to a warehouse to control and limit the amount of credit usage. Resource monitors help in setting quotas and alerts for warehouse usage.

CREATE WAREHOUSE my_warehouse

WITH WAREHOUSE_SIZE = ' XSMALL '

RESOURCE_MONITOR = ' my_resource_monitor ' ;

A key consideration for Data Clean Rooms is account and region compatibility. Participating accounts generally need to be in supported regions for the clean room configuration and collaboration model.

Important note: Data Clean Rooms have evolved quickly, and regional/cross-cloud support can change. I cannot live-verify the newest Snowflake restrictions here, but this is the best SnowPro-style answer from the provided options.

The useful calculation is based on BYTES_SCANNED and PERCENTAGE_SCANNED_FROM_CACHE. These metrics help estimate how much scanned data came from the warehouse cache. A low cache percentage on a query that scans many bytes may indicate that warehouse cache optimization could help.

Corrected note: the original options had OCR errors. The best matching valid answer is B .

Snowpark provides language-specific APIs for building data-processing and transformation logic. Snowpark operations are translated into SQL and executed by Snowflake’s compute engine. This provides compute pushdown because processing occurs inside Snowflake instead of requiring the complete dataset to be transferred to an external application.

The JDBC and ODBC drivers provide database connectivity. They allow applications to submit SQL statements and retrieve results, but they are not transformation frameworks comparable to Snowpark.

The Snowflake Connector for Kafka is primarily used to ingest streaming data from Apache Kafka into Snowflake.

Correction: The original option A said “Snowflake Connector for Python.” That connector is a Python database driver. The feature that provides DataFrame-based transformations and compute pushdown is the Snowpark API for Python.

Snowflake supports a variety of subquery types, including both correlated and uncorrelated subqueries. The correct answers are B and E, which highlight Snowflake ' s flexibility in handling subqueries within SQL queries.

Uncorrelated Scalar Subqueries:These are subqueries that can execute independently of the outer query. They return a single value and can be used anywhere a value expression is allowed, offering great flexibility in SQL queries.

EXISTS, ANY/ALL, and IN Subqueries:These subqueries are used inWHEREclauses to filter the results of the main query based on the presence or absence of matching rows in a subquery. Snowflake supports both correlated and uncorrelated versions of these subqueries, providing powerful tools for complex data analysis scenarios.

Examples and Usage:

Uncorrelated Scalar Subquery:

SELECT*FROMemployeesWHEREsalary > (SELECTAVG(salary)FROMemployees);

Correlated EXISTS Subquery:

SELECT*FROMorders oWHEREEXISTS(SELECT1FROMcustomer cWHEREc.id = o.customer_idANDc.region = ' North America ' );

Temporary and transient tables in Snowflake share several characteristics, notably, neither table type has a Fail-safe period. Fail-safe is a feature that provides additional data protection beyond the Time Travel period. However, this feature does not apply to temporary or transient tables. Additionally, for both types of tables, the data retention period effectively ends when the tables are dropped. This means that once these tables are deleted, their data is not recoverable, distinguishing them from permanent tables, which benefit from Snowflake ' s Time Travel and Fail-safe features.

The correct answer is A. View .

In Snowflake data sharing, secure objects are normally required when sharing views or functions because secure objects prevent sensitive underlying query details from being exposed. However, when the SECURE_OBJECTS_ONLY property is set to FALSE, non-secure views can be shared.

Why A is correct:

A view can be shared when SECURE_OBJECTS_ONLY = FALSE. This setting relaxes the requirement that only secure views be included.

Why the other options are incorrect:

B. Stored procedures are not standard objects shared through Snowflake secure data sharing in the same way as tables and views.

C. Storage integration access is not shared as a database object through a Snowflake share.

D. UDF sharing has security restrictions, and the safest exam answer here is a view when SECURE_OBJECTS_ONLY is disabled.

Official Snowflake documentation reference:

Snowflake documentation explains that shares can include database objects such as tables and views. Secure object requirements apply to certain shared objects, and the SECURE_OBJECTS_ONLY property controls whether only secure objects are allowed.

When using theSnowflake Kafka connector, the table schema includes two importantvariant columns:

RECORD_TIMESTAMP: This column stores the timestamp from the Kafka record, enabling time-based analysis of incoming data.

RECORDKEY: This captures the unique key of each Kafka message, useful for uniquely identifying records or managing deduplication.

These columns ensure that each message’s metadata and key information are preserved, facilitating data analysis and real-time processing tasks in Snowflake.

For minimizing the cost of large tables that are used to store transitory data, which does not need to be protected against failures because it can be reconstructed outside of Snowflake, the best table type to use isTransient. Transient tables in Snowflake are designed for temporary or transitory data storage and offer reduced storage costs compared to permanent tables. However, unlike temporary tables, they persist across sessions until explicitly dropped.

Why Transient Tables:Transient tables provide a cost-effective solution for storing data that is temporary but needs to be available longer than a single session. They have lower data storage costs because Snowflake does not maintain historical data (Time Travel) for as long as it does for permanent tables.

Creating a Transient Table:

To create a transient table, use theTRANSIENTkeyword in the CREATE TABLE statement:

CREATETRANSIENTTABLEmy_transient_table (...);

Use Case Considerations:Transient tables are ideal for scenarios where the data is not critical, can be easily recreated, and where cost optimization is a priority. They are suitable for development, testing, or staging environments where data longevity is not a concern.

When querying VARIANT columns (such as JSON), the values are often returned as strings . To convert them to usable SQL types, users must apply the CAST() function (e.g., CAST(column:value AS NUMBER)).

USERADMIN : Used for managing users and roles.

SECURITYADMIN : Can manage users, roles, and access control.

SYSADMIN is used primarily for object creation, not user management.

PARSE_DOCUMENT is used to extract information from documents, including text and layout elements. It is part of Snowflake Cortex document processing functionality.

Why the other options are incorrect:

TRANSCRIBE is for audio transcription, COMPLETE is for large language model completion, and EMBED generates vector embeddings.