Great find, Mike! Thanks for sharing.
Rick From: Michael Moser [mailto:[email protected]] Sent: Friday, September 04, 2015 3:24 PM To: [email protected] Subject: Long start time for NiFi on a VM I have experienced some long start times for NiFi while running on a virtual machine. This happens randomly so it has just been a minor annoyance. Well I finally spent some time investigating why, and this is what I learned. In my nifi-app.log I would see log messages like this. Note the 13 minute startup ... 2015-09-01 17:12:02,237 INFO [main] /nifi-api Initializing Spring root WebApplicationContext 2015-09-01 17:25:02,321 INFO [main] o.a.n.p.PersistentProvenanceRepository Recovered 0 records On low entropy environments, such as a VM on cloud services, it can take a long time to generate random numbers using java.security.SecureRandom. This is needed for encryption which NiFi does with encrypting sensitive properties, SSL, and more. Some references to other experiences with this are [1], [2] and [3]. So I took their advice and modified my bootstrap.conf to add this line, and my NiFi startup problems went away! java.arg.15=-Djava.security.egd=file:/dev/./urandom I think this deserves a note in the Admin Guide, perhaps in the Best Practice section. Using /dev/urandom on Linux to generate random numbers doesn't seem to be any less secure than /dev/random [4]. -- Mike [1] - http://stackoverflow.com/questions/137212/how-to-solve-performance-problem-with-java-securerandom [2] - https://github.com/kaazing/gateway/issues/167 [3] - http://stackoverflow.com/questions/26227344/oracle-java-8-x64-for-linux-and-randomsource [4] - http://www.2uo.de/myths-about-urandom/
