efeg
commented
2 years ago @mohitpali The metric AnomalyDetector.GOAL_VIOLATION-has-unfixable-goals is updated whenever a GOAL_VIOLATION anomaly with unfixable goals is detected.
Based on the log you shared, you manually used the REST API of Cruise Control and got an OptimizationFailureException due to failure to satisfy ReplicaCapacityGoal.
https://github.com/linkedin/cruise-control/blob/4e5927b48bf2581ab76acbbecbf42b355b871b65/cruise-control/src/main/java/com/linkedin/kafka/cruisecontrol/servlet/KafkaCruiseControlServletUtils.java#L260
In this case, it is not expected for the AnomalyDetector.GOAL_VIOLATION-has-unfixable-goals metric to be updated. So what you observed is the correct behavior.
- For definition of all sensors, please see https://github.com/linkedin/cruise-control/wiki/Sensors.
In a scenario where the replica counts exceed on all brokers and the goals are unfixable, I would expect the metric
AnomalyDetector.GOAL_VIOLATION-has-unfixable-goalsto go to 1.However, I noticed that the suggestion is to also remove excluded_topics, which is not an option for me. I would expect this to be reported as unfixable goals.
Error processing GET request '/proposals' due to: 'com.linkedin.kafka.cruisecontrol.exception.OptimizationFailureException: [ReplicaCapacityGoal] Replica count (513) in broker 8 exceeds the maximum allowed number of replicas per broker: 480. Add at least 1 broker. || Tips: [1] There are 5 topics excluded from replica move. Potential mitigation: Remove selected topics from exclusion using excluded_topics parameter. Then, re-run your original request. Add at least 1 broker.'.