Apache Druid

Sorry for late reply. Yes, we used this feature to do minor compaction previously. At that time, we ingested events from a large stream with slow data, which results in many small segments in each time chunk. In our use case, there were ~50 GB data and 1000 segments per hour (HOUR segment gran), so it's heavy to compact them in usual way with time chunk lock

While later we deprecated the minor compaction tweak and partitioned the large stream by region.

:+1:  So you are not using segment locking anymore? Btw We are planning to bring in some enhancements that lets one perform such compactions more cheaply.

Yes, we're not using it anymore. Looking forward to the enhancements!