[NIFI-8466] Offloading the Single Node load balancing target causes ignored backpressure

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.14.0
• Fix Version/s: 1.14.0
• Assignee: Joe Gresock
• Reporter: Joe Gresock
• Created: 22/Apr/21 20:17
• Updated: 19/May/21 18:40
• Resolved: 19/May/21 18:40

Description

In a 3-node cluster, it is possible to get a Single Node load balanced queue into the following state by offloading the node that receives the load balanced flow files:

1. The load balanced queue ignores backpressure
2. The processor following the load balanced queue acts like it has nothing to process, even when running

Steps to reproduce:

1. Add a Single Node load balanced queue on a multi-node cluster (see attached Template)
2. Discover which node receives the flow files in the queue by running only the first processor, then viewing the Connections tab of the Summary view, and finally clicking the Cluster button for the load balanced queue. The node that has all the flow files is the one you will need to Offload in the next step
3. Disconnect and then Offload the node that gets the flow files in Single Node load balancing
4. Reconnect the node
5. Run both processors surrounding the load balanced queue

Issue Links

• links to: GitHub Pull Request #5025