`targetCompactionSizeBytes` is similar to that of automatic compaction and 
means the size of the result segments after compaction. This is especially 
useful to compact a large interval partition into a few segments of the optimal 
size. A large interval partition means that the total size of segments in that 
partition is larger than `targetCompactionSizeBytes` no matter what the number 
of segments is.

Since indexTask doesn't support a sort configuration for target segment size, 
compactionTask assumes that the segment size is proportional to the number of 
rows in the segment and automatically computes `targetPartitionSize` based on 
the given `targetCompactionSizeBytes`. This assumption is pretty naive, so we 
can improve it later.

`targetCompactionSizeBytes` cannot be used together with `targetPartitionSize`, 
`maxTotalRows`, or `numShards` in `IndexTuningConfig`. If both are set, 
compactionTask will throw an exception and fail. If both are not set, 
`targetCompactionSizeBytes` is set to default (400MB).

[ Full content available at: 
https://github.com/apache/incubator-druid/pull/6203 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to