We are running druid 0.6.171 and seeing double counts in one of our segments. We are running a realtime node on our cluster on java 1.7.0_60 with the following command:
java -Xmx4096m -Duser.timezone=UTC -Dfile.encoding=UTF-8 -Ddruid.realtime.specFile=/usr/local/druid/config/realtime/volagg_kafka8_realtime_task.json -classpath /usr/local/druid/lib/*:/usr/local/druid/config/realtime io.druid.cli.Main server realtime
Other segment data looks good and we have been able to parse the same file that (mostly) created data in the bad segment with good counts on a separate cluster so we know it is not our data. From the log, it looks as if maxRowsInMemory was hit and it flushed the same data twice. Please advise.
Realtime log pertaining to the segment