As a sort of preoptimization we changed our Druid load jobs to truncate the timestamp column before loading it into Druid. The idea is that if we know we’re going to be querying by hour we could load data where the timestamp is truncated by hour for each record. This allows us to save a ton on file size and reduce the amoun of time it takes to load the data into Druid.
Unfortunately, it seems that the data is no longer queryable. It shows up as if it was loaded when looking at the coordinator node (# of segments are expected, file size is reasonable) but when we query Druid the responses make it seem as if there wasn’t any data loaded.
Has anyone run into this?