Hi everyone Here is a JIRA discussing about improving DN HB lock https://issues.apache.org/jira/browse/HDFS-7060 , it seems folks who participated the discussion both agreed on reducing the use of dataset log but the ticket goes to stale for a while. Recently we hit the issue in a 450 nodes cluster which has very heavy write load, after applying the patch, it improves the HB latency a lot. So I appreciate if someone can help to review and get this move forward.
Appreciate your comments on this. Thanks! -- Weiwei
