Hi ML, Once you have created your message flow file as Mark suggests you could use an external method to pass the file from the sending to the receiving group. These may be simpler to manage than S2S depending on your use case.
One simple option would be to use a PutFile and GetFile combination, using the local disk. Another would be to use a messaging system such as Kafka to publish the flow file to a topic that the receiver is consuming. Regards Steve Hindmarch -----Original Message----- From: Mark Payne <marka...@hotmail.com> Sent: 01 November 2024 15:07 To: users@nifi.apache.org; Maxime Lézier <maxime.lez...@ik.me> Subject: Re: Flowfile transfer without relationship Hey ML, There’s not really a straight-forward way to do what you’re asking. You’d essentially need to create Output Ports to allow the FlowFile to move up the stack of Process Groups and then into the neighboring Process Group (and potentially back down that other stack). The alternative, as you noted, would be using Site-to-Site but that does have the overhead of sending the FlowFIle over the network, and potentially to a different node in your cluster. Now, given that you do not need the content of the FlowFile but rather the “fact of” an option that might make sense for you would be to use a ReplaceText processor, configured with a Replacement Strategy of “Always Replace” and a replacement value of an empty string. This would essentially remove the content of the FlowFile, which means that sending over site-to-site would be far less resource intensive. Again, it may push the FlowFile to another node in the cluster, but depending on your requirements may be worth considering. Thanks -Mark > On Nov 1, 2024, at 5:54 AM, Maxime Lézier via users <users@nifi.apache.org> > wrote: > > > > Hello, > On one node from a cluster, I’d like to transfer flowfiles between two > processors which are not in the same process group. > The process group are not localized in the same zone on my canvas and I’d > like to avoid multiple relationship (imbrication of process group) for > sending flowfile from one processor to target processor. > I tried with s2s but the content is pushed, and make not necessary network > pressure on the s2s port. > All I want is to push only the flowfile reference from one processor to > target processor without “physical” relationship. The target processor could > continue to work on the flowfile like if it were the source processor > continuity. > This push is only on the node which execute the pipeline, no transfer between > node of the cluster. > Is there a way to achieve this ? > Thanks > ML
