Github user vakshorton commented on the issue:
https://github.com/apache/nifi/pull/2310
@pvillard31 You are likely seeing a lot of dropped events because the
timestamps in the events are out of synch with the server clock. If your
segment granularity is set to 10 minutes, late window set to 1 minute and the
event timestamp is more than 11 minutes out of synch with the server time, that
data is considered "late" (since the target index task will have closed). What
kind of ingestion rates on what input volume? Are you seeing a lot of queuing?
Can you check your middle manager configuration/resource allocation?
---
