[ https://issues.apache.org/jira/browse/NIFI-11388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17709035#comment-17709035 ]
Joe Witt commented on NIFI-11388: --------------------------------- Any processor could exceed the limit by basically any amount. It is not a hard cap. It purely is used to restrict giving threads to the processor in most cases before it could even be scheduled. Or, it is also available to processors to check during their processing if they should do more. The fundamentals are there but we deliberately did not *force* the hard limit but doing things such as failing a session if it would result in exceeding the back pressure value. > Backpressure queue settings loosely followed > -------------------------------------------- > > Key: NIFI-11388 > URL: https://issues.apache.org/jira/browse/NIFI-11388 > Project: Apache NiFi > Issue Type: Improvement > Reporter: Nissim Shiman > Priority: Major > > The backpressure settings on connections between processors are only loosely > followed. More flowfiles can end up on queues than configured via > backpressure setting. > For example, set up flow: > GenerateFlowFile -> UpdateAttribute -> (some processor) > where > GenerateFlowFile has _Custom Text_ set to be _hello_ > and Run Schedule is set to be _0 min_ > and > the connection following UpdateAttribute has _Back Pressure Object Threshold_ > set to be _100_ > Start GenerateFlowFile. > Wait a few moments until outgoing connection fills up. > Start UpdateAttribute > OutGoing conection will have more than 100 flowfiles on it > (I had over 1000 on mine when running these steps) -- This message was sent by Atlassian Jira (v8.20.10#820010)