[KAFKA-8444] If broker sense that it is unhealthy, broker should remove itself as leader from all partitions that it is hosting as partition leader.

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.1.1
• Component/s: core
• Labels: LeaderElector broker replication
• Flags: Important
• Assignee: Unassigned
• Reporter: bijaya
• Created: 28/May/19 20:33
• Updated: 28/May/19 20:33

Description

A broker can have partition leader for more than one partition.

If broker sense any panic with the host like disk write, memory (crashing), network. 

It should send eject me ( like ejecting from fighter jet) from cluster message to cluster controller/zookeeper. Which means its not a part of cluster anymore.

This makes the next leader election quicker and can prevent many issues before it happens.