ingestion past data

druid does not recommend to change rejection policy to anything except serverTime. But what if past data corrupted and i want to update it, what should i do?

Kind regards,


Hey Yunus,

The recommended way to handle past data is to use batch ingestion.

To add more flavor here: Fixing up data is part of the Lambda architecture used by Druid. In order to fix up data a batch job is run to re-index ALL the data for the time range of interest. So the complete set of “believed to be good data” should be included in the indexing job. The result of the batch indexing job will replace the prior data in the Druid cluster upon completion.