...

• The segments are managed differently in Pinot’s realtime table and offline table. The realtime segments are grouped using Kafka offsets, whereas the offline segments are split based on the time range. As a result, The realtime segments cannot be replaced if the job reruns.
• Also, it’s less efficient to use Kafka the intermediate storage for the batch processing, comparing to the direct segment creation and uploads.

Document

Widget Connectorurlhttps https://docs.google.com/document/d/1GVoFHOHSDPs1MEDKEmKguKwWMqM1lwQKj2e64RAKDf8/edit#heading=h.uvocz0dwkepo