[#HADOOP-3875] Fix TaskTracker's heartbeat timer to note the time the hearbeat RPC returned to decide next heartbeat time

Hadoop Common

Fix TaskTracker's heartbeat timer to note the time the hearbeat RPC returned to decide next heartbeat time

Created: 31/Jul/08 01:19 AM Updated: 08/Jul/09 04:52 PM

Return to search

Component/s:

None

Affects Version/s:

0.18.0

Fix Version/s:

0.19.0

Time Tracking:

Not Specified

File Attachments:

	File Name	Date Attached	Attached By	Size
	HADOOP-3875_0_20080730.patch	2008-07-31 02:08 AM	Arun C Murthy	2 kB

Hadoop Flags:	Reviewed
Resolution Date:	08/Aug/08 11:36 PM

Description

« Hide

The tasktracker notes the time the heartbeat RPC was fired, rather than when it returned to decide when to send the next one. If the RPC spends a lot of time in the JobTracker's queue then the above exacerbates the load on the JobTracker.

All

Comments

Work Log

Change History

Subversion Commits

Sort Order:

Repository	Revision	Date	User	Message
ASF	#684142	Fri Aug 08 23:35:46 UTC 2008	omalley	~~HADOOP-3875~~. Change the time period between heartbeats to be relative to the end of the heartbeat rpc, rather than the start. This causes better behavior if the JobTracker is overloaded. (acmurthy via omalley)
				Files Changed
				MODIFY /hadoop/core/trunk/src/mapred/org/apache/hadoop/mapred/TaskTracker.java MODIFY /hadoop/core/trunk/CHANGES.txt