Not so much resolved as worked around.  We had a large number of components on the canvas.  And we think we tracked the issue to timeouts in the cluster sync process.  (It was taking too long sometimes to get flow.xml sync'd between the nodes so it would stop.) The default timeout is 30 seconds.  We moved it up to 180 seconds. In Cloudera Manager, for the NiFi Configuration, search for properties which start with `nifi.cluster` to find the ones we changed. I also think things might have gotten better with more recent releases.  We have not seen this happen for quite some time. ... View more