I have a ES 2.4.x cluster that has many indices and some of them have too many docs.deleted. Some even have 80% of docs.count as docs.deleted.
This looks like ES is not keeping up with enough merges to cleanup deleted docs automatically, while there is normally not much activity going on in the cluster.
I found some documented options on ES < 2.x that can be tweaked but nothing on ES >= 2.x.
Could anyone point me to a good document on what to tweak to make ES be more aggressive on doing GC on docs.deleted?