Druid query behavior for older data that has been dropped from historical nodes

I will appreciate any help with the following question on a specific druid usage pattern.

The pattern is as follows.

Deep storage is setup as S3; druid coordinator rules are setup to load data for the latest one month and drop older data; a query is then made for data that is older than one month.

In this scenario how will the druid broker know which historical nodes to forward the query to and further how will it take into account the current capacity of the historical nodes?

I have gone through the documentation, the druid white paper and previous postings in this google group, but am not clear about the behavior.

Much thanks,


Hi Rahul,

I think you are missing one thing: data that is not loaded in advance on historical nodes cannot be queried. So in your scenario, where a query is made for a time range that has been dropped, what happens is you will just get no results at all.