[KAFKA-4319] AbstractFetcherManager shutdown speedup

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.10.2.0
• Assignee: Unassigned
• Reporter: Alexey Ozeritskiy
• Created: 19/Oct/16 11:59
• Updated: 19/Oct/16 16:52
• Resolved: 19/Oct/16 16:51

Description

While a shutdown proccess, AbstractFetcherManager closed all worker-threads sequentially which slows the final time of shutdown dramatically on huge clusters (approximately 15 minutes for 100 nodes, for example).

This can be improved by parallel mode. On the first stage AbstractFetcherManager can send the stop signal and then join all the workers to the thread

Issue Links

• links to: GitHub Pull Request #2023