SnowPro COF-C02 Practice Test: Snowflake Core

COF-C02 is Snowflake’s SnowPro Core certification for candidates who need strong platform fundamentals across architecture, security, performance, cost, and core data workflows. If you are searching for COF-C02 sample questions, a practice test, mock exam, or exam simulator, this is the main IT Mastery page to start on web and continue on iOS or Android with the same IT Mastery account.

Free diagnostic: Try the 100-question SnowPro Core COF-C02 full-length practice exam before subscribing. Use it as one Snowflake Core baseline, then return to IT Mastery for timed mocks, topic drills, explanations, and the full SnowPro Core question bank.

What this COF-C02 practice page gives you

• a direct route into IT Mastery practice for COF-C02
• topic drills and mixed sets across architecture, security, performance, loading, transformations, and sharing
• detailed explanations that show why the best Snowflake answer is correct
• a clear free-preview path before you subscribe
• the same IT Mastery account across web and mobile

COF-C02 exam snapshot

• Vendor: Snowflake
• Official exam name: SnowPro Core Certification (COF-C02)
• Exam code: COF-C02
• Items: 100 total
• Exam time: 115 minutes
• Question types: multiple-choice, multiple-select, and interactive items
• Passing score: 750 scaled

COF-C02 questions usually reward the option that follows Snowflake defaults and platform-native patterns instead of forcing warehouse, security, or data-movement decisions from other data-platform habits.

Topic coverage for COF-C02 practice

How to use the COF-C02 simulator efficiently

1. Start with domain drills so you can isolate whether your misses come from architecture, security, performance, or data workflows.
2. Review every miss until you can explain the Snowflake-native behavior or default that makes the best answer correct.
3. Move into mixed sets once you can connect warehouses, RBAC, loading patterns, Time Travel, cloning, and sharing in one scenario.
4. Finish with timed runs so the 100-question pace feels routine before test day.

SnowPro Core decision filters

Use these filters when several Snowflake features sound close:

• Object hierarchy: identify account, organization, database, schema, table, view, stage, file format, warehouse, role, and share boundaries.
• Compute vs storage: decide whether the issue is warehouse sizing, scaling, clustering, caching, storage, loading, or query design.
• Access control: apply roles, privileges, ownership, grants, masking, row access, network policy, and account-security controls.
• Loading path: distinguish internal and external stages, file formats, COPY, Snowpipe, streams, tasks, and unloading workflows.
• Performance signal: look for warehouse pressure, pruning, clustering, query profile, cache behavior, and micro-partition clues.

Final 7-day SnowPro Core practice sequence

When SnowPro Core practice is enough

If several unseen mixed attempts are above roughly 75% and you can explain the Snowflake object, role, loading, or warehouse behavior behind each answer, you are likely ready. More practice should improve platform judgment, not repeated-stem memorization.

Focused sample questions

Use these child pages when you want focused IT Mastery practice before returning to mixed sets and timed mocks.

• Free SnowPro Core COF-C02 Full-Length Practice Exam
• Architecture and Features
• Security and Governance
• Loading and Transformation
• Query Performance
• Sharing and Marketplace
• Operations and Continuity

Free study resources

Need concept review first? Read the SnowPro Core COF-C02 Cheat Sheet on Tech Exam Lexicon, then return here for timed mocks, topic drills, and full IT Mastery practice.

Free preview vs premium

• Free preview: a smaller web set so you can validate the question style and explanation depth.
• Premium: the full COF-C02 practice bank, focused drills, mixed sets, timed mock exams, detailed explanations, and progress tracking across web and mobile.

24 COF-C02 sample questions with detailed explanations

Question 1

Topic: Domain 4: Query processing and performance optimization

Which Snowflake command correctly lists all files currently stored in an internal named stage called mystage?

Options:

• A. SELECT * FROM @mystage;
• B. SHOW FILES IN STAGE mystage;
• C. DESCRIBE STAGE mystage;
• D. LIST @mystage;

Best answer: D

Explanation: The choice that uses LIST @mystage; is correct because:

• LIST is the dedicated Snowflake command for listing files in stages.
• The @ prefix correctly indicates a stage reference.
• Using just the stage reference (without a path or pattern) lists all files currently stored at the root of that named stage.

Question 2

Topic: Domain 4: Query processing and performance optimization

A team receives small JSON files into external cloud storage every 5 minutes throughout the day. They need new data available in Snowflake within 15 minutes on average, keep daily loading compute at or below 8 credits, and minimize operational management.

They use a virtual warehouse that consumes 0.3 credits for each scheduled COPY INTO run. Snowpipe continuous loading would consume 5 credits per day.

Which approach BEST meets all requirements?

(Note: There are 24 hours in a day. Assume average data latency is half of the load interval.)

Options:

• A. Use Snowpipe continuous loading on the external stage and stop the scheduled COPY INTO jobs.
• B. Schedule an automated COPY INTO job every 15 minutes using the existing warehouse.
• C. Keep a single manual COPY INTO run at the end of each day using the existing warehouse (2 credits per day total).
• D. Schedule an automated COPY INTO job every hour using the existing warehouse.

Best answer: A

Explanation: Using Snowpipe continuous loading on the external stage meets the average latency requirement by ingesting data soon after it lands, stays within the 8-credit budget at 5 credits per day, and reduces operational overhead by eliminating the need to orchestrate and maintain frequent scheduled COPY jobs and warehouse runtimes.

Question 3

Topic: Domain 5: Data sharing, collaboration, and marketplace

A data engineer compares query performance on a structured table and a directory table over unstructured image files.

Exhibit: Query history excerpt

Based on this exhibit, what is the most appropriate expectation when querying unstructured data in Snowflake?

Options:

• A. Unstructured queries are more efficient here because they produce fewer rows in a similar amount of time, so they can be treated as a low-latency replacement for structured tables.
• B. Because the unstructured query scans far fewer bytes, it should usually complete much faster than the structured query; the slower runtime must be caused by a misconfigured virtual warehouse.
• C. Since both queries complete in under 3,000ms, there is effectively no meaningful latency difference between structured and unstructured data access in Snowflake.
• D. Queries that list or filter unstructured files can have noticeably higher latency than comparable structured queries, so they are best suited for occasional metadata-style access rather than tight, low-latency analytics loops.

Best answer: D

Explanation: The choice that states that unstructured file listing and filtering can have noticeably higher latency than structured queries, and that they are best suited for occasional metadata-style access, directly matches the exhibit.

The unstructured query has EXECUTION_TIME_MS of 2,750 while scanning 1,048,576 bytes and producing 10 rows. The structured query has EXECUTION_TIME_MS of 380 while scanning 52,428,800 bytes and producing 12,500,000 rows. This shows that lower scanned bytes and smaller result sets for unstructured data do not translate into lower latency. Instead, there is extra overhead, so it is reasonable to expect higher latency and to use such queries sparingly for metadata and file discovery, not as a main low-latency analytics path.

Question 4

Topic: Domain 6: Operations, monitoring, and business continuity

A team instantly creates full-size test databases from production without using additional storage, because only metadata is duplicated while the underlying data storage is shared. Which Snowflake concept does this practice BEST represent?

Options:

• A. Data replication across regions for disaster recovery
• B. Automatic result caching to speed up repeated queries
• C. Materialized views for accelerating complex aggregations
• D. Zero-copy cloning for fast, cost-efficient environment creation

Best answer: D

Explanation: The option describing zero-copy cloning for fast, cost-efficient environment creation directly matches the stem’s details:

• Metadata-only duplication: Zero-copy cloning creates new objects by copying metadata, not the underlying data files.
• Shared storage: The clone and source reference the same micro-partitions until data diverges.
• Fast creation of test/backup environments: Because only metadata is written, large environments can be cloned almost instantly, which is ideal for test or backup-like use cases.

These behaviors align perfectly with the described practice, making this the only correct principle.

Question 5

Topic: Domain 1: Snowflake architecture and key features

Which statement correctly describes how Snowflake shares provide controlled read-only access to data for other Snowflake accounts?

Options:

• A. A share exports table data to external cloud storage, where the consumer downloads and loads it into their own Snowflake account.
• B. A data provider defines a share with specific database objects, and consumers create a read-only database from that share without copying the underlying data.
• C. A share creates a fully independent, writable copy of the provider’s database in the consumer account using zero-copy cloning.
• D. A share allows the provider to manage and modify roles, users, and warehouses inside the consumer’s Snowflake account.

Best answer: B

Explanation: The choice stating that a data provider defines a share with specific database objects, and consumers create a read-only database from that share without copying the underlying data is correct because it captures three essential facts:

• The provider controls which specific objects are included by defining the share.
• The consumer creates a database in their own account from that share.
• The shared database is read-only and references the provider’s data without physical data duplication.

These points reflect how secure data sharing works conceptually in the Snowflake AI Data Cloud.

Question 6

Topic: Domain 2: Account setup, security, and governance

Which statement BEST describes a Snowflake account in the Snowflake AI Data Cloud?

Options:

• A. It is the same as a single database and only controls objects inside that database.
• B. It is only a billing identifier and does not contain users, roles, or compute resources.
• C. It represents an entire cloud provider subscription and automatically spans all regions in that subscription.
• D. It is a logically isolated container that holds databases, schemas, virtual warehouses, users, roles, and related objects for that environment.

Best answer: D

Explanation: The statement that an account is a logically isolated container for databases, schemas, virtual warehouses, users, roles, and related objects correctly captures both scope and purpose. It emphasizes that the account is the main isolation boundary and that it includes data objects, compute, and security configuration used by that environment or tenant.

Question 7

Topic: Domain 2: Account setup, security, and governance

A security engineer reviews recent Snowflake login activity after enabling single sign-on with an external identity provider.

Exhibit:

Based on the exhibit, which statement BEST describes what federated authentication is providing for this Snowflake account?

Options:

• A. Virtual warehouses now handle both query processing and user authentication whenever AUTHENTICATION_METHOD is FEDERATED.
• B. User authentication is performed by an external identity provider, and Snowflake trusts those logins, enabling centralized single sign-on and password policies outside Snowflake.
• C. Snowflake validates user passwords directly and then pushes successful logins to the external identity provider for auditing only.
• D. Each user must maintain a separate Snowflake password in addition to using the external identity provider when AUTHENTICATION_METHOD is FEDERATED.

Best answer: B

Explanation: The choice describing that an external identity provider performs authentication, and that Snowflake trusts those logins to enable centralized SSO and external password policies, matches the exhibit’s AUTHENTICATION_METHOD = FEDERATED values. This is exactly what federated authentication provides: centralized control of user sign-in and credentials outside Snowflake while still granting access to Snowflake once the identity is verified.

Question 8

Topic: Domain 1: Snowflake architecture and key features

Which Snowflake interface is most appropriate for analysts who want to run ad-hoc queries, organize worksheets, and view dashboards and usage insights in a modern web browser UI?

Options:

• A. Snowsight web interface
• B. Snowflake SQL API
• C. SnowSQL command-line client
• D. Classic Snowflake web UI

Best answer: A

Explanation: The choice that specifies Snowsight is correct because Snowsight is Snowflake’s modern, browser-based UI that combines worksheets, dashboards, and account insights, which directly matches the needs of analysts running ad-hoc queries and exploring data visually.

Question 9

Topic: Domain 6: Operations, monitoring, and business continuity

Which TWO statements about cost and governance considerations for Snowflake continuous data protection features are INCORRECT? (Select TWO.)

Options:

• A. Zero-copy clones of large databases immediately duplicate all underlying storage, roughly doubling storage costs at creation time.
• B. Replicating databases to another region or account increases overall storage consumption because a full copy of the data is maintained in the target location.
• C. Fail-safe can be turned off for individual tables to avoid the additional storage overhead associated with that protection window.
• D. Secondary databases created by replication still require role-based grants in the target account to control who can query them.
• E. Extending Time Travel retention for a table can increase storage costs because more historical data is retained.

Correct answers: A and C

Explanation: The statements claiming that zero-copy clones immediately duplicate all storage and that Fail-safe can be turned off per table are incorrect.

Zero-copy cloning initially references existing micro-partitions, so storage is not doubled at creation. Storage only grows as changes are made to the source or clone.

Fail-safe is not configurable on a per-table basis and cannot simply be disabled to remove its storage overhead; it is a built-in safety net after Time Travel. These inaccuracies make those two statements the correct choices in a negative-polarity question.

Question 10

Topic: Domain 5: Data sharing, collaboration, and marketplace

You store unstructured documents in an internal stage and expose them through a directory table. Compliance requires that text extraction and keyword detection run entirely inside Snowflake to minimize data movement. Which approach is the BEST fit?

Options:

• A. Use Snowsight to export file metadata as CSV, run a local script to analyze the documents, and manually load the keyword list back into a Snowflake table.
• B. Unload the staged files to external cloud storage, process them with a separate compute cluster to extract keywords, and then reload the results into a Snowflake table.
• C. Create a JavaScript UDF that accepts document text returned by a SQL file function and outputs extracted keywords, then call it in a query over the directory table.
• D. Generate presigned URLs from the directory table so an external service can download each file, extract keywords, and write results back via a Snowflake connector.

Best answer: C

Explanation: The choice that creates a JavaScript UDF on top of text returned by a SQL file function is best because it keeps the entire pipeline inside Snowflake. The documents remain in the internal stage, the file function exposes their contents to SQL, and the UDF encapsulates custom logic for text extraction or keyword detection. This directly satisfies the single deciding factor in the scenario: minimizing data movement by performing all processing within Snowflake.

Question 11

Topic: Domain 5: Data sharing, collaboration, and marketplace

Which statement correctly describes how scalar functions and table functions differ in Snowflake transformations?

Options:

• A. Scalar functions return multiple rows of data, while table functions return a single aggregated value for the entire input set.
• B. Both scalar functions and table functions always return exactly one row with multiple columns, but table functions can be used only in JOIN clauses.
• C. Scalar functions can only be used in the FROM clause, whereas table functions can only be used in the SELECT list to compute derived columns.
• D. A scalar function returns a single value for each input row, while a table function returns a set of rows (and columns) that can be used like a table in the FROM clause.

Best answer: D

Explanation: The option that states a scalar function returns a single value per input row, while a table function returns a set of rows and columns usable like a table in the FROM clause is correct because it focuses exactly on the return shape: scalar = single value, table function = tabular result set. This aligns with how queries in Snowflake integrate scalar functions into expressions and table functions into the FROM/JOIN portions of a SELECT statement.

Question 12

Topic: Domain 1: Snowflake architecture and key features

A data engineering team currently exports large tables from the Snowflake AI Data Cloud to an external processing cluster to run complex transformation code. They want to redesign the pipeline using the requirements shown below.

Based on the exhibit, which Snowflake feature is MOST appropriate to implement the data processing logic?

Options:

• A. Create external tables on files in cloud storage and process them with an external Spark cluster.
• B. Rewrite all transformation logic as pure SQL in Snowsight worksheets and schedule it with tasks.
• C. Continue using the external processing cluster but replace exports with a Snowflake ODBC connection for reading data.
• D. Use Snowpark to build data pipelines in a supported programming language that execute inside Snowflake compute.

Best answer: D

Explanation: The choice to use Snowpark to build data pipelines in a supported programming language that execute inside Snowflake compute directly satisfies every line of the exhibit:

• R1: Snowpark supports familiar general-purpose languages, allowing teams to reuse existing language skills.
• R2: Snowpark code runs inside Snowflake virtual warehouses, so processing is pushed to Snowflake rather than external clusters.
• R3: Pipelines are implemented as code, which can be version-controlled and tested like other application code.
• R4: Snowpark provides DataFrame-like operations that operate on Snowflake tables, matching the requirement.

No other option in the list simultaneously meets all four requirements from the table.

Question 13

Topic: Domain 5: Data sharing, collaboration, and marketplace

A legal team stores signed contract PDFs as unstructured files in an internal stage and wants a Snowflake dashboard showing file name, size, last modified time, and a searchable text summary from each PDF. They insist on staying fully inside Snowflake SQL, avoiding external services and extra pipelines, and want a simple, maintainable design. Which approach best meets these requirements?

Options:

• A. Clone the internal stage into a new database, enable Time Travel on the database, and have analysts download the PDF files locally to inspect metadata and content on their laptops.
• B. Create a directory table on the stage and query it with SQL file functions in a view to derive file metadata and text content, then use that view as the dashboard source.
• C. Set up an external ETL tool to scan the stage storage, extract PDF metadata and text externally, and bulk load the results into a Snowflake table each night.
• D. Configure Snowpipe on the stage to continuously ingest the PDFs into a VARIANT column, then use an external function to parse the files and write results into a separate table.

Best answer: B

Explanation: The choice that creates a directory table on the stage and queries it with SQL file functions satisfies all constraints:

• It uses Snowflake’s built-in unstructured data support (directory tables) to expose file metadata.
• It uses SQL file functions to read metadata and contents directly in queries to derive text summaries.
• It stays entirely within Snowflake SQL, avoiding external services or extra ETL tools.
• It is simple to maintain because the logic lives in views or SQL transformations instead of additional infrastructure.

Question 14

Topic: Domain 2: Account setup, security, and governance

Which TWO statements about Snowflake network policies for restricting client IP addresses are correct? Assume a standard Snowflake AI Data Cloud deployment. (Select TWO.)

Options:

• A. They define which client IP addresses are allowed (or blocked) from connecting to a Snowflake account or specific users.
• B. They are configured only in external cloud networking components, not within Snowflake account settings.
• C. They determine which databases, schemas, and tables a user can access once connected to Snowflake.
• D. They transparently redirect traffic from disallowed IP addresses through a secure proxy instead of blocking the connection.
• E. They provide an additional perimeter control so that even if credentials are compromised, logins from unauthorized IP ranges can still be blocked.

Correct answers: A and E

Explanation: The statement that network policies define which client IP addresses are allowed or blocked is correct because their core purpose is to enforce IP-based connection restrictions at the account and/or user level.

The statement that they provide an additional perimeter control against compromised credentials is also correct, as limiting access to trusted IP ranges helps prevent successful logins from unexpected locations even when credentials are known.

Question 15

Topic: Domain 1: Snowflake architecture and key features

A small analytics team uses a single, always-on Large virtual warehouse for all workloads. Their data volume will double next quarter, but the compute credit budget must stay flat. They want to maintain performance and control costs by leveraging Snowflake’s separation of storage and compute. Which change is MOST appropriate?

Options:

• A. Increase the size of the existing warehouse to X-Large and keep it always running to handle the expected data growth.
• B. Manually limit storage growth by deleting historical micro-partitions so that the current warehouse can continue to run without change.
• C. Request a larger fixed storage cluster and disable auto-suspend so that compute and storage scale together and are easier to track.
• D. Create separate, smaller virtual warehouses for ETL and BI, each with auto-suspend and auto-resume enabled, while allowing Snowflake storage to grow automatically as data volume increases.

Best answer: D

Explanation: The choice that creates separate, smaller ETL and BI warehouses with auto-suspend and auto-resume correctly applies Snowflake’s separation of storage and compute:

• Storage is allowed to grow elastically as data volume increases, which is how Snowflake storage is designed.
• Compute is split into dedicated warehouses, improving concurrency and potentially performance.
• Auto-suspend and auto-resume ensure that compute credits are only consumed when workloads run, helping keep the overall compute budget flat even as data grows.

This meets both goals: maintaining performance (via workload isolation) and controlling compute costs (via right-sized, auto-suspending warehouses) while relying on independent storage scaling.

Question 16

Topic: Domain 1: Snowflake architecture and key features

Which statement best describes how query filters and the way data is ordered in a table affect micro-partition pruning in Snowflake?

Options:

• A. Filtering on columns whose values are well-clustered within micro-partitions lets Snowflake skip entire micro-partitions using min/max metadata, reducing the amount of data scanned.
• B. Micro-partition pruning happens only when a table is created with an explicit sort key; otherwise every micro-partition must always be scanned.
• C. Including an ORDER BY clause in a query automatically reorganizes table micro-partitions to improve pruning for future queries on those columns.
• D. Filters can improve pruning only for numeric columns; micro-partitions containing string or date columns are always fully scanned.

Best answer: A

Explanation: The statement about filtering on well-clustered columns allowing Snowflake to skip entire micro-partitions is correct because it directly describes how Snowflake uses per-column min/max metadata to prune micro-partitions and reduce scanned data when filters align with the stored order/distribution of values.

Question 17

Topic: Domain 3: Data loading, unloading, and transformation

Which condition most likely causes Snowflake to recompute a query instead of using a previously populated result cache?

Options:

• A. The data in one or more queried tables changed after the last time the query ran.
• B. The query is run on a different but similarly sized virtual warehouse.
• C. The virtual warehouse auto-suspended between the first and second execution of the query.
• D. The query is submitted from Snowsight instead of a JDBC-based client.

Best answer: A

Explanation: The choice stating that the data in one or more queried tables changed is correct because Snowflake’s result cache is only valid as long as the underlying data remains unchanged. Once DML modifies the tables referenced by the query, the cached result no longer represents the current data, so Snowflake discards it and recomputes the query to return up-to-date results.

Question 18

Topic: Domain 1: Snowflake architecture and key features

In Snowflake, when is a temporary table the most appropriate choice instead of a permanent or transient table?

Options:

• A. When storing business-critical data that must be recoverable for the longest possible Time Travel and Fail-safe durations
• B. When holding reference data that many users and sessions will query over time
• C. When creating a table that must be visible to all sessions in the account but automatically drops after 24 hours
• D. When storing intermediate query results needed only within the current session and safe to discard afterward

Best answer: D

Explanation: The choice that stores intermediate query results needed only within the current session and safe to discard afterward directly matches the design of temporary tables. They exist only for the duration of the session, are not shared across sessions, and are ideal for scratch or working data that does not need long-term persistence or recovery.

Question 19

Topic: Domain 3: Data loading, unloading, and transformation

A large fact table has about 5 billion rows. Most queries filter on ORDER_DATE, CUSTOMER_ID, or both. The team plans to use clustering to improve micro-partition pruning. Which approach is the LEAST appropriate and should NOT be chosen?

Options:

• A. Define a composite clustering key on (ORDER_DATE, CUSTOMER_ID) to support frequent filters on both columns.
• B. Define a clustering key on ORDER_DATE so partitions align with the most common time-based filters.
• C. Define a clustering key on high-cardinality ORDER_ID plus five additional low-selectivity columns so that “all possible filters are covered.”
• D. Initially leave the table unclustered, relying on natural clustering by ingestion order on ORDER_DATE, and monitor query performance before adding a key.

Best answer: C

Explanation: The choice that defines a clustering key on high-cardinality ORDER_ID plus five additional low-selectivity columns is incorrect because it violates several best practices:

• ORDER_ID is typically unique or near-unique, so micro-partitions will not group many related rows together; pruning benefits are minimal.
• Adding several extra columns “to cover all possible filters” produces a very wide clustering key, increasing the cost of reclustering and metadata maintenance.
• The key does not directly focus on the main filter columns (ORDER_DATE and CUSTOMER_ID), so it fails the primary performance goal.

This is a classic clustering anti-pattern: high complexity and cost with little improvement in pruning.

Question 20

Topic: Domain 1: Snowflake architecture and key features

A provider account has created an outbound share named SALES_SHARE and wants to give a consumer account controlled access to sales data.

The provider runs:

SHOW GRANTS TO SHARE sales_share;

Result:

Based on this exhibit, which statement BEST describes what a consumer account can do with this shared data?

Options:

• A. The consumer can create new tables directly inside PROD_SALES_DB because USAGE is granted on the database.
• B. The consumer cannot query SALES_FACT until grant_option is set to true for the SELECT privilege on the table.
• C. The consumer can query PROD_SALES_DB.SALES_FACT as read-only data but cannot modify the provider’s underlying table.
• D. The consumer can load new data into SALES_FACT using COPY INTO because SELECT on the table is granted to the share.

Best answer: C

Explanation: The choice describing that the consumer can query PROD_SALES_DB.SALES_FACT as read-only but cannot modify the provider’s table aligns with both the exhibit and Snowflake’s sharing model. The privilege column lists only USAGE on the database and SELECT on the table, which are inherently non-DML privileges, and shares expose those privileges as read-only access in consumer accounts. Since no write privileges are present, the consumer’s interaction with the provider’s data is limited to querying.

Question 21

Topic: Domain 6: Operations, monitoring, and business continuity

A data engineering team uses the Snowflake AI Data Cloud and wants protection against accidental table drops and incorrect updates. They plan to rely on Snowflake’s data protection features for day-to-day recovery. Which of the following statements about using Snowflake Time Travel for data recovery is INCORRECT?

Options:

• A. The team should primarily rely on Fail-safe instead of Time Travel for routine recovery from accidental data changes because Fail-safe is faster and directly accessible by users.
• B. The team can create clones of tables or schemas at a specific timestamp in the Time Travel window to produce test copies that reflect a prior state of the data.
• C. Analysts can run SELECT queries using AT or BEFORE clauses (for example, AT(TIMESTAMP =>...)) to view historical versions of data without actually restoring or changing the current table.
• D. Engineers can use Time Travel with commands such as UNDROP TABLE to restore tables that were accidentally dropped, as long as they are within the configured Time Travel retention period.

Best answer: A

Explanation: The option that recommends primarily relying on Fail-safe instead of Time Travel for routine recovery is incorrect. Fail-safe is not a user-driven, fast recovery mechanism; it is a Snowflake-managed safety net for disaster scenarios and is accessed with Snowflake Support involvement. Normal, self-service recovery from accidental deletes or drops should use Time Travel, making the statement about using Fail-safe as the main tool for routine recovery clearly wrong.

Question 22

Topic: Domain 5: Data sharing, collaboration, and marketplace

An analytics team repeatedly queries the same set of large image files stored as unstructured data in an internal named stage. Queries are slower than expected and warehouse usage is rising. Which of the following actions/solutions will meet these requirements? (Select TWO.)

Options:

• A. Create and use a directory table on the stage, and filter by file path or metadata so each query touches only the required files.
• B. Run the queries on a dedicated, always-on virtual warehouse so the same warehouse cache can serve repeated reads of the unstructured files.
• C. Upgrade to a multi-cluster warehouse with auto-scale so each query runs on a different cluster, increasing total cache capacity.
• D. Increase the Time Travel retention period on the database that contains the directory table so repeated queries can reuse cached file contents.
• E. Split each large file into many small chunks and store them as separate objects in the stage to increase parallelism.

Correct answers: A and B

Explanation: The choice to run queries on a dedicated, always-on virtual warehouse directly leverages Snowflake’s remote file cache. Because the same warehouse continues running between queries, cached contents of frequently read unstructured files can be reused, lowering latency and compute effort.

The choice to create and use a directory table with filters on file path or metadata optimizes access patterns. It ensures that queries only enumerate and access the subset of files required, which reduces unnecessary remote reads and improves both performance and cost efficiency when working with unstructured data.

Question 23

Topic: Domain 1: Snowflake architecture and key features

An organization is migrating several 10TB transactional tables into Snowflake. The DBA team previously managed table partitions and compression manually and wants to minimize ongoing operational work while still achieving efficient storage and good query performance. Which of the following actions/solutions will meet these requirements? (Select TWO.)

Options:

• A. Load the data into standard Snowflake tables using COPY and rely on Snowflake to automatically create and maintain micro-partitions and data compression.
• B. Periodically run VACUUM and ANALYZE commands on large tables to keep micro-partitions compact and statistics up to date.
• C. Disable Snowflake’s automatic micro-partition maintenance and configure custom compression algorithms per table to reduce storage usage.
• D. Pre-partition the source data into separate physical files and directory structures that exactly match anticipated query filters, then manually manage these layouts over time.
• E. Avoid manually reorganizing or compressing table storage, and instead monitor query performance and only consider defining clustering keys if specific query patterns require it.

Correct answers: A and E

Explanation: The option that loads data with COPY and relies on automatic micro-partitions and compression is correct because it uses Snowflake’s default behavior: data is transparently stored in compressed micro-partitions without any manual tuning. The option that avoids manual reorganization and only considers clustering keys when query patterns demand it is also correct because it acknowledges that Snowflake already manages storage, and that additional tuning should be exception-based, preserving low operational overhead.

Question 24

Topic: Domain 1: Snowflake architecture and key features

Which statement BEST describes an external stage in the Snowflake AI Data Cloud?

Options:

• A. A special schema type used to organize external tables and views that are created from shared databases.
• B. A named object that stores staged files inside Snowflake-managed storage, typically used for temporary data loading and unloading operations.
• C. A virtual warehouse type optimized for reading data directly from external cloud storage without using stages.
• D. A named object that references files in external cloud storage, used for loading and unloading data without storing the files inside Snowflake-managed storage.

Best answer: D

Explanation: The choice describing a named object that references files in external cloud storage and is used for loading and unloading data is correct because it captures the key properties of an external stage: it is a Snowflake object, it points to an external location, and it is used as the source or target for COPY and related operations without storing data inside Snowflake-managed stage storage.

SnowPro Core COF-C02 platform map

Use this map after the sample questions to connect individual items to the Snowflake architecture, loading, security, performance, and cost decisions these practice samples test.

    flowchart LR
	  S1["Analytics platform requirement"] --> S2
	  S2["Choose database schema warehouse and role design"] --> S3
	  S3["Load and transform data"] --> S4
	  S4["Secure share and govern access"] --> S5
	  S5["Monitor performance cost and usage"] --> S6
	  S6["Optimize storage compute and operations"]

Quick Cheat Sheet

Mini Glossary

• Warehouse: Snowflake compute cluster used to execute queries and workloads.
• Stage: Location for files before loading into Snowflake.
• RBAC: Role-based access control model.
• Micro-partition: Snowflake storage unit used for pruning and query efficiency.
• Time Travel: Snowflake feature for accessing historical data within retention.

In this section

• SnowPro Core COF-C02: Architecture and Features
  Try 10 focused SnowPro Core COF-C02 questions on Architecture and Features, with explanations, then continue with IT Mastery.
• SnowPro Core COF-C02: Security and Governance
  Try 10 focused SnowPro Core COF-C02 questions on Security and Governance, with explanations, then continue with IT Mastery.
• SnowPro Core COF-C02: Loading and Transformation
  Try 10 focused SnowPro Core COF-C02 questions on Loading and Transformation, with explanations, then continue with IT Mastery.
• SnowPro Core COF-C02: Query Performance
  Try 10 focused SnowPro Core COF-C02 questions on Query Performance, with explanations, then continue with IT Mastery.
• SnowPro Core COF-C02: Sharing and Marketplace
  Try 10 focused SnowPro Core COF-C02 questions on Sharing and Marketplace, with explanations, then continue with IT Mastery.
• SnowPro Core COF-C02: Operations and Continuity
  Try 10 focused SnowPro Core COF-C02 questions on Operations and Continuity, with explanations, then continue with IT Mastery.
• Free SnowPro Core COF-C02 Full-Length Practice Exam: 100 Questions
  Try 100 free SnowPro Core COF-C02 questions across the exam domains, with explanations, then continue with full IT Mastery practice.