Lucas Wang created KAFKA-7350:
---------------------------------
Summary: Improve or remove the PreferredReplicaImbalanceCount
metric
Key: KAFKA-7350
URL: https://issues.apache.org/jira/browse/KAFKA-7350
Project: Kafka
Issue Type: Improvement
Reporter: Lucas Wang
Assignee: Lucas Wang
In KAFKA-6753, we identified that in the ControllerEventManager, updating two
metrics after processing every controller event ends up consuming too much CPU.
The first metric OfflinePartitionCount is resolved in KAFKA-6753, and this
ticket is for tracking progress on the 2nd metric
PreferredReplicaImbalanceCount.
The options we have about this metric include
1. Remove this metric given that if necessary, the value of this metric can be
derived by getting the metadata of all topics in the cluster
2. Piggyback the update of the metric every time the auto leader balancer runs.
The benefit is keeping this metric. However the downside is that this metric
may then get obsolete and incorrect depending on when it's checked.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
