You uploaded 5 years of logs to Cloud Storage and found out-of-range values indicating errors. You must fix values, keep originals for compliance, and be able to rerun the process. What should you do?

The correct option is to use a Dataflow pipeline to read the data, correct the errors, and write the corrected data to a new dataset in Cloud Storage. Dataflow is a fully managed service ideal for large-scale data processing, offering features like fault tolerance and auto-scaling. Writing to a new dataset ensures that the original, uncorrected data is preserved for compliance, and allows for easy rerunning of the process if needed. Importing into BigQuery and skipping rows would lose the erroneous data, violating the compliance requirement to keep originals. Creating a Compute Engine instance for this task is less efficient and scalable than Dataflow for large datasets and requires more manual management. Writing corrected records back to the same dataset would overwrite the original data, again violating the compliance requirement.