[HBASE-9471] htrace synchronized on getInstance

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.98.0, 0.96.0
• Fix Version/s: 0.98.0, 0.96.0
• Component/s: regionserver
• Hadoop Flags: Reviewed
• Assignee: Elliott Neil Clark
• Reporter: Nicolas Liochon
• Created: 09/Sep/13 16:29
• Updated: 20/Nov/15 11:56
• Resolved: 13/Sep/13 18:12

Description

When doing tests on cached data, one of the bottleneck is the getInstance() on HTrace, called in RequestContext#set() --> Trace.isTracing()

When it's fixed, we see threads blocked in sendResponse and in the metrics (with hadoop 1).

The difference is not huge (it's in the range 0-5%), but there is no reason to keep this.

I'm sending a pull request to htrace.