Exam Bundle

A second run processes only newly discovered files pertains to a different control (Use Auto Loader with a durable schema location and checkpoint) and a different need (New files must be incrementally processed without listing every historical object each time). It cannot establish that Require producer-side atomic finalization or a stable file arrival protocol met the stated goal.

That observation is a valid indicator for “A job must load a finite set of source files into an existing Delta table without persistent streaming” when using Use supported `COPY INTO` with the source path and format. The scenario instead tests “A source writes partially constructed objects that should not be ingested until finalized”; the relevant result is Only completed object versions enter the accepted ingestion batch.

This signal belongs to the adjacent case “Files arrive repeatedly and source discovery should scale beyond directory listing,” whose remedy is Use Auto Loader managed file-event notification mode where supported. The observed result for the present requirement is Only completed object versions enter the accepted ingestion batch.

The alternative Allow the supported schema evolution update and restart as required has a distinct purpose—An optional field appears for the first time and is approved for schema evolution. It would not establish the required outcome (New JSON attributes should become accepted target columns after a managed restart); the necessary verification concerns The stream detects a new column, updates schema metadata and resumes after restart.

Use Auto Loader managed file-event notification mode where supported belongs to the separate case “Files arrive repeatedly and source discovery should scale beyond directory listing,” where New-file events result in ingest progress without scanning the entire directory would be expected. That differs from the condition in this stem: “New JSON attributes should become accepted target columns after a managed restart.”

Restore the persisted Auto Loader schema directory from backup or safely rebootstrap belongs to the separate case “An existing schema location was deleted and streaming now cannot reliably recover its previous schema,” where The stream resumes with the expected column definitions and history state would be expected. That differs from the condition in this stem: “New JSON attributes should become accepted target columns after a managed restart.”

The restart does not logically duplicate previously committed files pertains to a different control (Keep the durable checkpoint and write to a transactionally safe Delta sink) and a different need (A stream retries processing after a transient worker crash). It cannot establish that Configure Auto Loader `addNewColumns` and a persistent schemaLocation met the stated goal.

The stream retains large numeric values using a widened type pertains to a different control (Use supported type-widening schema evolution and a compatible runtime) and a different need (A new column uses an allowed type widening from int to long). It cannot establish that Configure Auto Loader `addNewColumns` and a persistent schemaLocation met the stated goal.

The foreign table displays the expected new column after refresh pertains to a different control (Refresh the supported foreign catalog metadata or mappings) and a different need (A foreign schema evolves and analysts cannot see a new remote column). It cannot establish that Configure Auto Loader `addNewColumns` and a persistent schemaLocation met the stated goal.

Configure `failOnNewColumns` and enforce the approved schema is appropriate when “A fixed partner schema must stop ingestion when unannounced columns appear”; it would be verified by The stream rejects the unknown column until the contract is revised. It does not directly satisfy “An unexpected JSON attribute must be preserved without expanding production schema,” whose relevant acceptance evidence is The unknown attribute is available in rescued data while target columns stay fixed.

This action solves a different issue: “New files must be incrementally processed without listing every historical object each time.” Its normal signal is A second run processes only newly discovered files. Here the requested behavior is “An unexpected JSON attribute must be preserved without expanding production schema,” so that signal is not sufficient.

Choose supported Auto Loader file-event notification mode is appropriate when “The source container has high object count and low-latency event-based discovery is required”; it would be verified by Files are discovered through events without a costly whole-container listing. It does not directly satisfy “An unexpected JSON attribute must be preserved without expanding production schema,” whose relevant acceptance evidence is The unknown attribute is available in rescued data while target columns stay fixed.

The target schema includes the approved field with existing rows preserved would demonstrate the separate action Allow the supported schema evolution update and restart as required, which addresses “An optional field appears for the first time and is approved for schema evolution.” It does not verify the requirement “An unexpected JSON attribute must be preserved without expanding production schema” following Choose Auto Loader `rescue` mode and configure rescued data column.

This signal belongs to the adjacent case “Analysts must query an external supported database without first copying it into Databricks,” whose remedy is Create a Unity Catalog Lakehouse Federation connection and foreign catalog. The observed result for the present requirement is The unknown attribute is available in rescued data while target columns stay fixed.

The stream rejects the unknown column until the contract is revised would demonstrate the separate action Configure `failOnNewColumns` and enforce the approved schema, which addresses “A fixed partner schema must stop ingestion when unannounced columns appear.” It does not verify the requirement “An unexpected JSON attribute must be preserved without expanding production schema” following Choose Auto Loader `rescue` mode and configure rescued data column.

Use supported type-widening schema evolution and a compatible runtime is appropriate when “A new column uses an allowed type widening from int to long”; it would be verified by The stream retains large numeric values using a widened type. It does not directly satisfy “A fixed partner schema must stop ingestion when unannounced columns appear,” whose relevant acceptance evidence is The stream rejects the unknown column until the contract is revised.

The alternative Allow the supported schema evolution update and restart as required has a distinct purpose—An optional field appears for the first time and is approved for schema evolution. It would not establish the required outcome (A fixed partner schema must stop ingestion when unannounced columns appear); the necessary verification concerns The stream rejects the unknown column until the contract is revised.

The alternative Normalize the key field during bronze-to-silver ingestion has a distinct purpose—Several operational feeds use differing column names for the same business key. It would not establish the required outcome (A fixed partner schema must stop ingestion when unannounced columns appear); the necessary verification concerns The stream rejects the unknown column until the contract is revised.

This signal belongs to the adjacent case “An unexpected JSON attribute must be preserved without expanding production schema,” whose remedy is Choose Auto Loader `rescue` mode and configure rescued data column. The observed result for the present requirement is The stream rejects the unknown column until the contract is revised.

This signal belongs to the adjacent case “A source writes partially constructed objects that should not be ingested until finalized,” whose remedy is Require producer-side atomic finalization or a stable file arrival protocol. The observed result for the present requirement is The stream rejects the unknown column until the contract is revised.

The physical scan reads only selected required columns where possible pertains to a different control (Read it as Parquet or another supported columnar format) and a different need (A large columnar export should support efficient Spark column pruning). It cannot establish that Configure `failOnNewColumns` and enforce the approved schema met the stated goal.

Configure `failOnNewColumns` and enforce the approved schema is appropriate when “A fixed partner schema must stop ingestion when unannounced columns appear”; it would be verified by The stream rejects the unknown column until the contract is revised. It does not directly satisfy “A new column uses an allowed type widening from int to long,” whose relevant acceptance evidence is The stream retains large numeric values using a widened type.

The alternative Use Databricks-to-Databricks sharing for the authorized recipient has a distinct purpose—A recipient also uses Databricks with compatible Unity Catalog sharing. It would not establish the required outcome (A new column uses an allowed type widening from int to long); the necessary verification concerns The stream retains large numeric values using a widened type.

The alternative Refresh the supported foreign catalog metadata or mappings has a distinct purpose—A foreign schema evolves and analysts cannot see a new remote column. It would not establish the required outcome (A new column uses an allowed type widening from int to long); the necessary verification concerns The stream retains large numeric values using a widened type.

The foreign table displays the expected new column after refresh would demonstrate the separate action Refresh the supported foreign catalog metadata or mappings, which addresses “A foreign schema evolves and analysts cannot see a new remote column.” It does not verify the requirement “A new column uses an allowed type widening from int to long” following Use supported type-widening schema evolution and a compatible runtime.

This signal belongs to the adjacent case “A large columnar export should support efficient Spark column pruning,” whose remedy is Read it as Parquet or another supported columnar format. The observed result for the present requirement is The stream retains large numeric values using a widened type.

The stream detects a new column, updates schema metadata and resumes after restart pertains to a different control (Configure Auto Loader `addNewColumns` and a persistent schemaLocation) and a different need (New JSON attributes should become accepted target columns after a managed restart). It cannot establish that Use supported type-widening schema evolution and a compatible runtime met the stated goal.

Configure Auto Loader `addNewColumns` and a persistent schemaLocation belongs to the separate case “New JSON attributes should become accepted target columns after a managed restart,” where The stream detects a new column, updates schema metadata and resumes after restart would be expected. That differs from the condition in this stem: “An existing schema location was deleted and streaming now cannot reliably recover its previous schema.”

The alternative Use Auto Loader with a durable schema location and checkpoint has a distinct purpose—New files must be incrementally processed without listing every historical object each time. It would not establish the required outcome (An existing schema location was deleted and streaming now cannot reliably recover its previous schema); the necessary verification concerns The stream resumes with the expected column definitions and history state.

This action solves a different issue: “A connector was removed but its replication slot still exists.” Its normal signal is The orphaned slot disappears and WAL retention decreases. Here the requested behavior is “An existing schema location was deleted and streaming now cannot reliably recover its previous schema,” so that signal is not sufficient.

The stream detects a new column, updates schema metadata and resumes after restart would demonstrate the separate action Configure Auto Loader `addNewColumns` and a persistent schemaLocation, which addresses “New JSON attributes should become accepted target columns after a managed restart.” It does not verify the requirement “An existing schema location was deleted and streaming now cannot reliably recover its previous schema” following Restore the persisted Auto Loader schema directory from backup or safely rebootstrap.

The target schema includes the approved field with existing rows preserved pertains to a different control (Allow the supported schema evolution update and restart as required) and a different need (An optional field appears for the first time and is approved for schema evolution). It cannot establish that Restore the persisted Auto Loader schema directory from backup or safely rebootstrap met the stated goal.

This signal belongs to the adjacent case “A foreign schema evolves and analysts cannot see a new remote column,” whose remedy is Refresh the supported foreign catalog metadata or mappings. The observed result for the present requirement is The stream resumes with the expected column definitions and history state.