Snowflake SnowPro Core COF-C03 Practice Test 99
Question 980
[Production migration readiness] A reviewer recommends: Use Snowpark for Python with an appropriately provisioned execution context and pushdown-compatible dataframe operations. The requirement remains: A production migration team is preparing a controlled cutover; the solution must preserve existing account isolation. A data engineer wants Python dataframe transformations to execute near governed Snowflake data without moving the full dataset to a workstation. The release needs an operationally defensible choice without accidentally changing unrelated access or data. Which explanation best justifies this recommendation?
- Snowpark-optimized warehouses provide a higher memory profile that can help memory-intensive Snowpark computations, subject to the workload's actual bottleneck.
- Unqualified object resolution depends on active session context; the warehouse alone does not specify which schema is targeted.
- The Spark connector integrates Snowflake data access with Spark distributed processing and can use supported query pushdown.
- Snowpark builds logical plans executed in Snowflake, reducing unnecessary client-side transfers for supported operations.
Explanation
Correct Answer
D:In this production migration readiness context, Snowpark builds logical plans executed in Snowflake, reducing unnecessary client-side transfers for supported operations. The recommendation is linked to observable evidence for this exact case: Query history shows the intended Snowflake-executed statements and the Python client retrieves only intended final results. Merely choosing a nearby feature would not establish the needed capability.
Incorrect Answers
A:For the production migration readiness conditions, Snowpark-optimized warehouses provide a higher memory profile that can help memory-intensive Snowpark computations, subject to the workload's actual bottleneck. That explanation applies to a different snowpark optimized workload isolation problem: A Python Snowpark pipeline uses large in-memory intermediate processing and repeatedly runs out of memory on a standard virtual warehouse. It does not justify the specific snowpark python pushdown action or its validation path: Query history shows the intended Snowflake-executed statements and the Python client retrieves only intended final results.
B:That option concerns session database and role context, not the requirement in this case. It does not justify the specific snowpark python pushdown action or its validation path: Query history shows the intended Snowflake-executed statements and the Python client retrieves only intended final results.
C:The Spark connector integrates Snowflake data access with Spark distributed processing and can use supported query pushdown.
Question 981
[Production compliance review] A compliance reviewer is checking which safeguard actually enforces the stated boundary. Regional managers share a table but must see only the geographic rows permitted by their assigned region. The conclusion must be justified with observable verification rather than an undocumented assumption. Which Snowflake action best meets this specific requirement?
- Define and attach a row access policy that evaluates approved regional role or mapping criteria.
- Use a managed access schema and put the relevant grant authority under the schema owner or a role with MANAGE GRANTS.
- Evaluate hybrid tables for the transactional access pattern and verify feature and account availability.
- Attach a dynamic data masking policy to the sensitive column with conditions appropriate for the approved role model.
Explanation
Correct Answer
A:Row access policies constrain which rows are returned at query time; a column masking policy does not filter rows. In this production compliance review setting, the intended validation is: Separate user-role tests return only the authorized region's row keys while total source data remains unchanged.
Incorrect Answers
B:That action is relevant when A governance team wants object owners inside a schema to be unable to distribute arbitrary grants without review by the schema's grant administrator. Row access policies constrain which rows are returned at query time; a column masking policy does not filter rows.
C:That action is relevant when A customer application requires low-latency keyed row updates with enforced transactional constraints that fit a supported hybrid-table design. Evaluate hybrid tables for the transactional access pattern and verify feature and account availability.
D:That action is relevant when A customer table contains national identifiers; privileged reviewers may see full values but ordinary analysts should receive redacted values. It is not the primary action for the stated row access policy filtering need.
Question 982
[Security control assessment] A reviewer recommends: Configure appropriate future grants for the required schema scope while preserving narrowly defined access roles. The requirement remains: A quarterly access-control audit requires an evidence-backed answer. Analysts repeatedly lose approved read access whenever the ingestion team creates another table in a governed reporting schema. The reviewer must choose a supported mechanism for the precise requirement, not a generic product feature. Which explanation best justifies this recommendation?
- System-defined administrative roles are broad; custom roles can reflect separation of duties without giving each analyst infrastructure authority.
- Future grants apply designated privileges to newly created objects and avoid uncontrolled ad hoc grant repetition.
- Warehouses provide compute; they are not parents of schemas or tables within the Snowflake object catalog.
- Roles can inherit privileges from other roles through deliberate grants, making functional privilege layering auditable.
Explanation
Correct Answer
B:In this security control assessment context, Future grants apply designated privileges to newly created objects and avoid uncontrolled ad hoc grant repetition. The recommendation is linked to observable evidence for this exact case: A newly created sample table receives the expected SELECT grant without adding unrelated elevated ownership. Merely choosing a nearby feature would not establish the needed capability.
Incorrect Answers
A:That option concerns custom roles versus system roles, not the requirement in this case. It does not justify the specific future grants in new schema action or its validation path: A newly created sample table receives the expected SELECT grant without adding unrelated elevated ownership.
C:That option concerns account and schema object hierarchy, not the requirement in this case. It does not justify the specific future grants in new schema action or its validation path: A newly created sample table receives the expected SELECT grant without adding unrelated elevated ownership.
D:That option concerns inherited functional role grants, not the requirement in this case. It does not justify the specific future grants in new schema action or its validation path: A newly created sample table receives the expected SELECT grant without adding unrelated elevated ownership.
Question 983
[Ingestion design acceptance] The team proposes: Create a named CSV file format with the actual FIELD_DELIMITER, SKIP_HEADER, and FIELD_OPTIONALLY_ENCLOSED_BY settings. Its actual requirement is: A data engineering team is checking the design before promoting a pipeline. One source feed uses pipe-delimited UTF-8 files with a header and quoted embedded delimiters, and every daily load must interpret them consistently. The reviewer must choose a supported mechanism for the precise requirement, not a generic product feature. Which observation most directly validates success?
- The exported object files have the expected schema, row counts, and authorized destination path.
- A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY.
- A staged preview and validation load reconcile embedded delimiter strings and expected column counts across sample rows.
- Load history and target counts confirm expected files were not ingested twice after simulated network retry.
Explanation
Correct Answer
C:A staged preview and validation load reconcile embedded delimiter strings and expected column counts across sample rows. This is an outcome test of the specified reusable csv file format action under the ingestion design acceptance conditions. A file format controls how COPY parses file bytes; mismatched delimiter and quoting rules cause shifted or rejected columns.
Incorrect Answers
A:In this ingestion design acceptance context, The exported object files have the expected schema, row counts, and authorized destination path. This is an outcome for a different requirement: A downstream finance application requires an export of curated query results to an approved cloud storage path for a partner batch job. It cannot by itself validate the proposed reusable csv file format action; the appropriate test is: A staged preview and validation load reconcile embedded delimiter strings and expected column counts across sample rows.
B:A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY. This is an outcome for a different requirement: Incoming small object-storage files appear several times per hour and should load automatically soon after the object store sends events. It cannot by itself validate the proposed reusable csv file format action; the appropriate test is: A staged preview and validation load reconcile embedded delimiter strings and expected column counts across sample rows.
D:In this ingestion design acceptance context, Load history and target counts confirm expected files were not ingested twice after simulated network retry. This is an outcome for a different requirement: A nightly pipeline may retry an earlier bulk load after a network timeout and must avoid inadvertently duplicating files that were already committed. It cannot by itself validate the proposed reusable csv file format action; the appropriate test is: A staged preview and validation load reconcile embedded delimiter strings and expected column counts across sample rows.
Question 984
[Optimization regression test] A team is validating a targeted tuning plan against documented workloads and costs. A table contains several rows per customer and an analyst wants exactly the latest row for each customer after computing ROW_NUMBER. The conclusion must be justified with observable verification rather than an undocumented assumption. Which specific design mistake would most directly defeat the intended result?
- Assume every syntactically similar SELECT always reuses cached results even after source tables are modified.
- Assume adding a conventional user-managed B-tree index is the only way Snowflake can prune native tables.
- CAST the entire purchases array to VARCHAR and assume one row per array element appears automatically.
- Put a window function directly inside a WHERE predicate and assume WHERE executes after window evaluation.
Explanation
Correct Answer
D:The supportable direction is: Use ROW_NUMBER with PARTITION BY and an ordered window, then filter the window result with QUALIFY. Put a window function directly inside a WHERE predicate and assume WHERE executes after window evaluation.
Incorrect Answers
A:That is a mistake for a different persisted result reuse conditions task, where A team reruns an unchanged SELECT but sometimes receives cached results and sometimes spends compute again after underlying data changes. Under optimization regression test conditions, Assume every syntactically similar SELECT always reuses cached results even after source tables are modified.
B:That is a mistake for a different selective filter micro-partition pruning task, where A daily operational report filters on a narrow event-time range yet scans nearly the entire multi-terabyte event table. Under optimization regression test conditions, Assume adding a conventional user-managed B-tree index is the only way Snowflake can prune native tables.
C:That is a mistake for a different flatten nested array rows task, where An event JSON record contains a purchases array and analysts need one relational row per purchase item while retaining the parent event identifier. Under optimization regression test conditions, CAST the entire purchases array to VARCHAR and assume one row per array element appears automatically.
Question 985
[Collaboration operating audit] The team proposes: Configure supported database or account replication and plan promotion or failover procedures for the desired recovery objectives. Its actual requirement is: A platform owner must demonstrate that the proposed sharing model respects its access and recovery boundary. A multinational organization requires a recoverable approved dataset in another supported Snowflake account region for resilience planning. The conclusion must be justified with observable verification rather than an undocumented assumption. Which observation most directly validates success?
- Replication status and a controlled secondary-account query confirm the expected dataset and documented lag.
- The recipient discovers the listing in its region and can read the expected version after fulfillment completes.
- The consumer creates a database from the share and can SELECT only granted objects, with provider revocation reflected in access.
- The reader account signs in and queries the intended shared database while billing and warehouse usage are monitored by the provider.
Explanation
Correct Answer
A:Replication status and a controlled secondary-account query confirm the expected dataset and documented lag. This is an outcome test of the specified replication across regions for availability action under the collaboration operating audit conditions. Replication moves supported data and metadata across account or region boundaries and differs from a local zero-copy clone.
Incorrect Answers
B:The recipient discovers the listing in its region and can read the expected version after fulfillment completes. This is an outcome for a different requirement: A provider and a consumer operate in different supported regions and want catalog-based access without assuming a same-region direct share works unchanged. It cannot by itself validate the proposed replication across regions for availability action; the appropriate test is: Replication status and a controlled secondary-account query confirm the expected dataset and documented lag.
C:In this collaboration operating audit context, The consumer creates a database from the share and can SELECT only granted objects, with provider revocation reflected in access. This is an outcome for a different requirement: A Snowflake data provider wants a business partner who already has a Snowflake account to query governed datasets without copying daily CSV exports. It cannot by itself validate the proposed replication across regions for availability action; the appropriate test is: Replication status and a controlled secondary-account query confirm the expected dataset and documented lag.
D:In this collaboration operating audit context, The reader account signs in and queries the intended shared database while billing and warehouse usage are monitored by the provider. This is an outcome for a different requirement: A small research partner needs curated data but has no existing Snowflake account and the provider wants a controlled hosted-consumer path. It cannot by itself validate the proposed replication across regions for availability action; the appropriate test is: Replication status and a controlled secondary-account query confirm the expected dataset and documented lag.
Question 986
[Pipeline incident investigation] An on-call ingestion engineer needs a targeted fix that preserves source-to-target reconciliation. An orders table changes frequently and a downstream merge process must apply inserts and updates since its last committed checkpoint. The release needs an operationally defensible choice without accidentally changing unrelated access or data. Which specific design mistake would most directly defeat the intended result?
- Use a plain view with an identical SELECT and assume it stores refreshed aggregate rows permanently at no cost.
- Run a daily full-table INSERT without keys and assume it is equivalent to applying the stream's change history.
- Run unconditional INSERT for every retry and rely on order-by formatting to prevent duplicate customer keys.
- Call a standard read-only view a continuously materialized table with guaranteed five-second refresh history.
Explanation
Correct Answer
B:The discriminating technical reason is: A stream tracks table-change records and an offset, while the downstream DML determines how those changes are materialized. The supportable direction is: Create a stream on the supported source and consume its delta into the target within a transactionally safe workflow.
Incorrect Answers
A:That is a mistake for a different materialized view precomputation task, where A dashboard repeatedly calculates the same expensive aggregate over a large base table while the source changes incrementally. Under pipeline incident investigation conditions, Use a plain view with an identical SELECT and assume it stores refreshed aggregate rows permanently at no cost.
C:That is a mistake for a different merge keyed upsert task, where A CRM staging feed has new customers and changed addresses and needs an idempotent update-or-insert into the curated customer table. Under pipeline incident investigation conditions, Run unconditional INSERT for every retry and rely on order-by formatting to prevent duplicate customer keys.
D:That is a mistake for a different dynamic table downstream freshness task, where A pipeline repeatedly joins newly ingested transactions to reference data and needs a declarative freshness objective instead of hand-managed CDC steps. Under pipeline incident investigation conditions, Call a standard read-only view a continuously materialized table with guaranteed five-second refresh history.
Question 987
[Peak traffic performance incident] The team proposes: Use a deterministic MERGE condition on a true business key with MATCHED UPDATE and NOT MATCHED INSERT branches. Its actual requirement is: A BI engineer is investigating the exact bottleneck observed under production-like load. A CRM staging feed has new customers and changed addresses and needs an idempotent update-or-insert into the curated customer table. The release needs an operationally defensible choice without accidentally changing unrelated access or data. Which observation most directly validates success?
- A validated report has one region row and expected quarter columns with reconciled totals.
- After controlled source updates, the stream returns the expected delta, and a committed MERGE advances the consumption checkpoint.
- A repeated staging batch produces stable row counts and correctly updates changed addresses without duplicated business keys.
- A regression test with tied timestamps always chooses the same event according to the documented tie-breaker.
Explanation
Correct Answer
C:A repeated staging batch produces stable row counts and correctly updates changed addresses without duplicated business keys. This is an outcome test of the specified merge keyed upsert action under the peak traffic performance incident conditions. MERGE reconciles source and target records under explicit key conditions and avoids blind duplicate inserts on retries.
Incorrect Answers
A:A validated report has one region row and expected quarter columns with reconciled totals. This is an outcome for a different requirement: A quarterly metrics report stores one row per region and quarter but the executive view needs a separate column for each quarter. It cannot by itself validate the proposed merge keyed upsert action; the appropriate test is: A repeated staging batch produces stable row counts and correctly updates changed addresses without duplicated business keys.
B:In this peak traffic performance incident context, After controlled source updates, the stream returns the expected delta, and a committed MERGE advances the consumption checkpoint. This is an outcome for a different requirement: An orders table changes frequently and a downstream merge process must apply inserts and updates since its last committed checkpoint. It cannot by itself validate the proposed merge keyed upsert action; the appropriate test is: A repeated staging batch produces stable row counts and correctly updates changed addresses without duplicated business keys.
D:In this peak traffic performance incident context, A regression test with tied timestamps always chooses the same event according to the documented tie-breaker. This is an outcome for a different requirement: A churn dataset has multiple events with identical timestamps per customer, and selecting the latest without a tie-breaker gives inconsistent winners. It cannot by itself validate the proposed merge keyed upsert action; the appropriate test is: A repeated staging batch produces stable row counts and correctly updates changed addresses without duplicated business keys.
Question 988
[Production migration readiness] The team proposes: Configure an external table and its stage/integration and maintain the external file metadata as required. Its actual requirement is: A production migration team is preparing a controlled cutover; the solution must preserve existing account isolation. A data scientist needs governed SQL reads over files remaining in external object storage, without an ordinary Snowflake-managed load step. The release needs an operationally defensible choice without accidentally changing unrelated access or data. Which observation most directly validates success?
- Queries return the same committed rows after a warehouse resumes, while compute credit consumption stops during suspension.
- Directory metadata contains the expected file paths, sizes and update timestamps after a controlled stage refresh.
- The reader account signs in and queries the intended shared database while billing and warehouse usage are monitored by the provider.
- SELECT against the external table returns staged records after location metadata has been refreshed appropriately.
Explanation
Correct Answer
D:SELECT against the external table returns staged records after location metadata has been refreshed appropriately. This is an outcome test of the specified external table over data lake action under the production migration readiness conditions. External tables expose externally stored data through Snowflake-managed metadata rather than copying all rows into a normal native table.
Incorrect Answers
A:In this production migration readiness context, Queries return the same committed rows after a warehouse resumes, while compute credit consumption stops during suspension. This is an outcome for a different requirement: An administrator suspends every reporting warehouse overnight and the operations manager fears that previously committed tables will be erased. It cannot by itself validate the proposed external table over data lake action; the appropriate test is: SELECT against the external table returns staged records after location metadata has been refreshed appropriately.
B:In this production migration readiness context, Directory metadata contains the expected file paths, sizes and update timestamps after a controlled stage refresh. This is an outcome for a different requirement: A document processing workflow keeps PDFs on a Snowflake stage and needs a queryable inventory of staged files before extraction. It cannot by itself validate the proposed external table over data lake action; the appropriate test is: SELECT against the external table returns staged records after location metadata has been refreshed appropriately.
C:In this production migration readiness context, The reader account signs in and queries the intended shared database while billing and warehouse usage are monitored by the provider. This is an outcome for a different requirement: A small research partner needs curated data but has no existing Snowflake account and the provider wants a controlled hosted-consumer path. It cannot by itself validate the proposed external table over data lake action; the appropriate test is: SELECT against the external table returns staged records after location metadata has been refreshed appropriately.
Question 989
[Provider and consumer workshop] A data collaboration team is defining the permitted provider and consumer relationship. A small research partner needs curated data but has no existing Snowflake account and the provider wants a controlled hosted-consumer path. The reviewer must choose a supported mechanism for the precise requirement, not a generic product feature. Which specific design mistake would most directly defeat the intended result?
- Grant access to a nonexistent consumer account identifier and assume Snowflake creates a working user interface automatically.
- Assume CLONE DATABASE in one account silently creates an always-current replica in every other geography.
- Assume a database is automatically replicated to every account merely because all accounts share the same organization.
- Email unencrypted copies of all source tables nightly and claim that this is the only supported Snowflake collaboration model.
Explanation
Correct Answer
A:The discriminating technical reason is: Reader accounts can let consumers without their own full Snowflake account access provider-shared data, subject to provisioning and cost controls. The supportable direction is: Evaluate and provision a Snowflake reader account with authorized shares and documented provider billing responsibilities.
Incorrect Answers
B:That is a mistake for a different replication across regions for availability task, where A multinational organization requires a recoverable approved dataset in another supported Snowflake account region for resilience planning. It does not directly defeat the present reader account for external consumer requirement, whose proper treatment is: Evaluate and provision a Snowflake reader account with authorized shares and documented provider billing responsibilities.
C:Under provider and consumer workshop conditions, Assume a database is automatically replicated to every account merely because all accounts share the same organization. It does not directly defeat the present reader account for external consumer requirement, whose proper treatment is: Evaluate and provision a Snowflake reader account with authorized shares and documented provider billing responsibilities.
D:Under provider and consumer workshop conditions, Email unencrypted copies of all source tables nightly and claim that this is the only supported Snowflake collaboration model.
Question 990
[Pipeline incident investigation] The team proposes: Configure Snowpipe with a pipe definition and supported auto-ingest cloud event notifications. Its actual requirement is: An on-call ingestion engineer needs a targeted fix that preserves source-to-target reconciliation. Incoming small object-storage files appear several times per hour and should load automatically soon after the object store sends events. The release needs an operationally defensible choice without accidentally changing unrelated access or data. Which observation most directly validates success?
- End-to-end event timestamps show data arriving within the required streaming latency under peak event load.
- A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY.
- LIST on the stage shows the expected files and COPY INTO loads exactly the reconciled input rows.
- Load history and target counts confirm expected files were not ingested twice after simulated network retry.
Explanation
Correct Answer
B:A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY. This is an outcome test of the specified snowpipe file event auto ingest action under the pipeline incident investigation conditions. Snowpipe uses Snowflake-managed compute to ingest staged files automatically, distinct from warehouse-driven scheduled COPY batches.
Incorrect Answers
A:End-to-end event timestamps show data arriving within the required streaming latency under peak event load. This is an outcome for a different requirement: An IoT product emits row events continuously and needs lower-latency ingestion than file-batched loading provides. It cannot by itself validate the proposed snowpipe file event auto ingest action; the appropriate test is: A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY.
C:In this pipeline incident investigation context, LIST on the stage shows the expected files and COPY INTO loads exactly the reconciled input rows. This is an outcome for a different requirement: A team needs a stable staging location where several engineers can PUT source files and run repeatable bulk loads for the same project. It cannot by itself validate the proposed snowpipe file event auto ingest action; the appropriate test is: A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY.
D:In this pipeline incident investigation context, Load history and target counts confirm expected files were not ingested twice after simulated network retry. This is an outcome for a different requirement: A nightly pipeline may retry an earlier bulk load after a network timeout and must avoid inadvertently duplicating files that were already committed. It cannot by itself validate the proposed snowpipe file event auto ingest action; the appropriate test is: A newly uploaded event-notified file appears in pipe load history and target rows without running a manual warehouse COPY.