Github user mosermw commented on the issue:
https://github.com/apache/nifi/pull/1240
I tested this a lot, and based on the conversation in the JIRA ticket, I
adjusted the purge low water mark to 88%. This allows the prov_repo to purge
old events quicker after a large spike in events, whereas before it could take
a long time to bring the prov_repo back down to 90% usage. It also doesn't
delete too many events. I also adjusted the rollover high water mark to 99%,
which allows space for journals and indexes. This keeps prov_repo usage much
closer to the nifi.provenance.repository.max.storage.size.
+1 will merge.
---
