ARA-C01 Question 243
Single answerData corruption impactsA retail company loads daily sales data from cloud object storage into a Snowflake table using a scheduled COPY INTO command. After a deployment, an upstream transformation bug writes incorrect currency conversion values into the source files for the last 36 hours. The files were loaded successfully, and downstream dashboards now show materially overstated revenue. The architecture team must restore the table to a correct state quickly while minimizing data loss and reprocessing effort. Which approach should the architect recommend?
- A
Use Time Travel to create or restore the table to a point before the bad loads, then reload only the corrected files for the affected period.
- B
Run RECLUSTER on the table so Snowflake can rebuild corrupted micro-partitions and restore the original values.
- C
Use FAILSAFE to immediately query the table as it existed before the bad loads, then merge the corrected rows back into the current table.
- D
Resume the warehouse with a larger size and rerun the dashboards so Snowflake recalculates the cached results using the original source data.
Show answer and explanation
Correct answer: A
Explanation
The key architectural distinction is between logical data corruption and platform/storage failure. Here, Snowflake successfully loaded files, but the source content was wrong due to an upstream bug. Snowflake's immutable micro-partition architecture does not automatically infer business-correct values, so the remediation pattern is to use Time Travel to access the table state before the bad load and then reload corrected data for the impacted time range. This is generally the fastest and least disruptive recovery method when the retention period still covers the event. RECLUSTER and warehouse resizing do not correct bad data, and Fail-safe is a last-resort recovery feature with limited customer control, not a primary operational remediation tool. Best practice is to pair recovery with load auditing, validation checks, and controlled replay of corrected source files. Relevant Snowflake documentation areas include Time Travel, Fail-safe, COPY INTO behavior, and zero-copy cloning for point-in-time recovery workflows.
- A. Correct.
Correct. This scenario describes logical data corruption caused by bad upstream data, not physical storage corruption. In Snowflake, Time Travel is the primary feature for accessing or restoring historical table states within the configured retention period. An architect can use CLONE AT/BEFORE or restore data from a point before the incorrect COPY operations, then reload only the corrected files for the affected window. This minimizes data loss and avoids unnecessary full historical reloads.
- B. Incorrect.
Incorrect. RECLUSTER only reorganizes table storage for clustering efficiency and query pruning. It does not repair incorrect business values loaded into a table. The misconception is confusing storage optimization with data recovery.
- C. Incorrect.
Incorrect. Fail-safe is not designed for immediate, self-service querying by customers. It is intended as a last-resort recovery mechanism managed by Snowflake after Time Travel retention has expired. It also does not support selectively querying historical data directly for day-to-day remediation of bad loads.
- D. Incorrect.
Incorrect. Increasing warehouse size or rerunning dashboards affects compute performance, not the correctness of persisted table data. Result cache invalidation or recomputation will still read the incorrect values currently stored in the table. The misconception is assuming the issue is stale cache rather than corrupted loaded data.