@domR    I see no issues with using a publish and consume design for triggering follow on flows. That provide a slightly more robust setup than using the postHTTP to ListenHTTP example I provided since any NiFi node can consume the trigger file. When using Kafka you will want to make sure you have the same number of partitions as you have consumer in the same consumer group.  When you add a consumeKafka processor to the NiFi canvas it is really being added to every node in your cluster and is configured the same on every node. Let's assume you have a 3 node NiFi cluster, that means you have at a minimum 3 consumers in that consumer group so your Kafka topic should have 3 partitions; otherwise, you may see a lot of rebalancing happen.  To improve performance even more you can increase the concurrent task on your consumeKafka processor.  With a 3 node NiFi cluster and a consumeKafka configured with 2 concurrent tasks, you now have 6 total consumers in the consumer group; therefore, your Kafka topic should have 6 partitions.   If a NiFi node goes down Kafka will assign multiple partitions to the same consumer, so no need to worry about messages not being consumed. Hope this information was useful in finding a solution, Matt ... View more