In Monitoring Console, under Distributed Search: Instance, the average times for "Time to Reap Knowledge Bundle Directory" and "Time to Reap Dispatch Directory" are showing very long times. 593,038 ms for the Knowledge bundle and 1,151,252 for the Dispatch Directory. The notes at the bottom of say that this caused by storage performance issues or excessive number of bundles or artifacts.
My question is how do I find the number of bundles/artifacts and clear out any excess?
I don't know if this is a best practice, but I simply go to the directories and search for any files that have not been modified in the past week - then delete them. This seems to work and you could probably even use a more recent cutoff. I used 7 days because that is the longest time that someone could save their search results (which are kept in the dispatch directory) - at least in my particular case.
I've scripted this and it runs nightly.
Looking closer at the dashboard for this, there are values for MAX values for these instances. The MAX values are a lot lower, 2145 ms max for the "Time to Reap Knowledge Bundle Directory" and 3085 ms max for the "Time to Reap Dispatch Directory". I seem to remember from math class that the average should be lower than the maximum value. Could there be some calculation problem in the underlying search for the dashboard?