When ingesting data, if the source table has no primary key, is it mandatory to externally define primary key columns in the target, or can we ingest without them?

It is not mandatory to define a primary key in the target if the source has none, but it is highly recommended. Without a key, you risk duplicate rows during incremental loads and cannot efficiently perform upserts or deletes. The typical solution is to add a surrogate key in the target (e.g., an auto‑incrementing sequence or UUID) and, if possible, create a unique constraint on a combination of business columns that uniquely identify a record. This allows you to perform idempotent loads and maintain referential integrity. If the downstream system does not require a key, you can skip it, but you should document the decision and ensure that any downstream processes can handle duplicates or lack of uniqueness.