Github user jskora commented on the issue:
https://github.com/apache/nifi/pull/1241
@joewitt Since 0.x is end-of-life, this is OBE and should be closed.
IIRC, the root problem was writing records first and then checking
afterward to see if storage limits were exceeded and a rollover needed. If the
storage limits don't leave a margin before reaching the physical capacity,
writes can fill the disk and crash the NiFi and possible the system. I looked
at the 1.x repositories in depth since they were re-written, but if the same
write-then-check behavior exists this could still be a problem.
---
