Can you take a thread dump and provide the output here? ./bin/nifi.sh dump /path/to/output/dump.txt ... View more

Also wanted to mention that site-to-site can only happen with ports on the root canvas, which is done intentionally to make it clear that it is a system decision to send or receive data out of the given instance. This means that the user that sets up the port would need write access to the root canvas, which typically in a multi-tenant environment tenants would be given to a sub-process group, but they would still need some other administrator to setup the site-to-site connection for them. ... View more

You would need to setup an appender in logback.xml that could talk to a Splunk input. One example would be a SocketAppender to send to a TCP input in Splunk. https://github.com/bbende/jsonevent-producer/blob/master/src/main/resources/logback.xml#L8-L11 ... View more

The message about changing the OS buffer is just a warning and not something to usually something to worry about it. What is happening is that there is a connection pool, and each time a connection is made it tries to set the socket send buffer based on the value of the property in the processor, and if it can't set it to that value than it logs a warning to let you know. The latest screenshot you posted is a completely different problem, and it is something we need to look into. What I think is happening is that the processor is holding on to flow files to transfer them in the next execution, but since you cleared the queue they are no longer valid and this is causing an unexpected situation. As Oleg mentioned, we need to see the full stacktrace from the logs for this. ... View more

If you set it to /tmp and set Recurse Sub Directories to true, then it will list both. ... View more

You need to grant your users Read/Write access to the root process group (i.e. the top level canvas). You do this from the context palette on the left, using the the little policy icon. ... View more

The URL property supports expression language so that should be fine. You may want to double check that the flow files have the attributes you expect. You could have a RouteOnAttribute that routes ${latitude:isEmpty():or(longitude:isEmpty())} to somewhere so that by the time it makes it to InvokeHttp you know you have both values, or you could just use a LogAttribute and check the logs. ... View more

Right now you have FetchDistributedMapCache -> ExecuteScript... you could replace this with one custom processor (or maybe one custom scripted processor) that uses DistributedMapCache to fetch the questions, joins it with the response, and writes the whole thing to flow file content, thus avoiding ever sticking the questions in an attribute. ... View more