Killed Between Writes

An order service commits the order row to Postgres and then publishes an OrderPlaced event to Kafka. During a rolling deploy the process is killed between those two steps, and downstream services never learn about roughly forty orders. What is the standard fix?

Retry the publish in a finally block with exponential backoff so the event survives any failure of the produce call

Insert the event into an outbox table in the same transaction as the order row, and let a relay publish and mark it sent

Publish to Kafka first and commit the order row only after the broker acknowledges the write, so no event can ever be lost

Wrap the database commit and the Kafka produce in an XA two-phase commit so the two resources reach the same outcome atomically