PutEmail is definitely good for specific parts of the flow. As you mentioned it can get complex quickly trying to route all the failures to a single PutEmail. The ReportingTask is definitely a good idea. When a ReportingTask executes it gets access to a ReportingContext which has access to BulletinRepository, which then gets you access to any of the bulletins you see in the UI. You could have one that got all the error bulletins and sent them somewhere or emailed them. Along the lines of monitoring the logs, you could probably configure NiFi's logback.xml to do UDP or TCP forwarding of all log events at the ERROR level, and then have a ListenUDP/ListenTCP processor in NiFi receive them and send an email. If you are in a cluster I guess you would have all nodes forward to only one of the nodes. This introduces possibility for circular logic, meaning if the ListenUDP/ListenTCP had problems that would generate more ERROR logs which would get sent back to ListenUDP/ListenTCP, and this produce more errors and keep doing this until the problem was resolved, but that is probably rare.
... View more
