[KAFKA-14457] Inconsistent in quorum controller fenced broker metric

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.1.0, 3.2.0, 3.1.1, 3.3.0, 3.1.2, 3.2.1, 3.2.2, 3.2.3, 3.3.1
• Fix Version/s: 3.4.0, 3.3.2
• Assignee: José Armando García Sancio
• Reporter: José Armando García Sancio
• Created: 10/Dec/22 00:27
• Updated: 27/Dec/22 20:46
• Resolved: 20/Dec/22 19:04

Description

It is possible for controllers to replay a record twice. This happens when the active controller replays an uncommitted record, resigns and replays the same record when it becomes committed.

The controller handles these transition changes by using timeline data structures and reverting to previous in-memory snapshots.

This functionality is not used when computing the fenced and unfenced metrics. Specifically, the metric can over count when executing this code:

           } else if (prevRegistration == null) {
              if (registration.fenced()) {
                  controllerMetrics.setFencedBrokerCount(controllerMetrics.fencedBrokerCount() + 1);
                  log.info("Added new fenced broker: {}", registration.id());
              } else {
                  controllerMetrics.setActiveBrokerCount(controllerMetrics.activeBrokerCount() + 1);
                  log.info("Added new unfenced broker: {}", registration.id());
              }

From ClusterControlManager::updateMetrics.

Issue Links

• links to: GitHub Pull Request #12994