[KAFKA-2913] GroupMetadataManager unloads all groups in removeGroupsForPartitions

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.9.0.1
• Assignee: Jason Gustafson
• Reporter: Jason Gustafson
• Created: 01/Dec/15 05:01
• Updated: 01/Dec/15 05:34
• Resolved: 01/Dec/15 05:34

Description

In GroupMetadataManager.removeGroupsForPartition, we currently remove all groups from the cache, but we should only be removing groups for the respective partition. This will cause an unneeded rebalance for groups mistakenly removed. And since the generation would reset to 0 in this case, the offsets from the current generation would be rejected and the consumer would see duplicates.