I've been running a logstash framework for a while now that had 5 instances of logstash reading from a kafka queue, processing that data, and pushing it to elasticsearch. I discovered recently that that kafka queue only had 2 partitions, and 2 machines had taken those, and that my other 3 machines had essentially been sitting idle for quite a while. I never suspected this because, they were pretty consistently running at 70-100% all of the time EVEN THOUGH they were processing no records.
Why might this possibly be??