Stanislav Kozlovski created KAFKA-9866:
------------------------------------------
Summary: Do not attempt to elect preferred leader replicas which
are outside ISR
Key: KAFKA-9866
URL: https://issues.apache.org/jira/browse/KAFKA-9866
Project: Kafka
Issue Type: Improvement
Reporter: Stanislav Kozlovski
The controller automatically triggers a preferred leader election every N
minutes. It tries to elect all preferred leaders of partitions without doing
some basic checks like whether the leader is in sync.
This leads to a multitude of errors which cause confusion:
{code:java}
April 14th 2020, 17:01:11.015 [Controller id=0] Partition TOPIC-9 failed to
complete preferred replica leader election to 1. Leader is still 0{code}
{code:java}
April 14th 2020, 17:01:11.002 [Controller id=0] Error completing replica
leader election (PREFERRED) for partition TOPIC-9
kafka.common.StateChangeFailedException: Failed to elect leader for partition
TOPIC-9 under strategy PreferredReplicaPartitionLeaderElectionStrategy {code}
It would be better if the Controller filtered out some of these elections, not
attempt them at all and maybe log an aggregate INFO-level log
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
