[KAFKA-1170] ISR can be inconsistent during partition reassignment for low throughput partitions

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.8.1
• Fix Version/s: 0.8.1
• Component/s: controller
• Assignee: Neha Narkhede
• Reporter: Neha Narkhede
• Created: 06/Dec/13 22:18
• Updated: 13/Jan/14 05:06
• Resolved: 13/Jan/14 05:06

Description

During reassignment of partitions, if the reassigned replicas already has the current leader, then the ISR of the partition can have the old replicas. This is particularly true for partitions that don't have further incoming data, since the leader does not remove replicas that have stopped sending fetch requests from the ISR if their log end offsets match (no more data comes in).