[
https://issues.apache.org/jira/browse/HADOOP-3333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12593832#action_12593832
]
Amar Kamat commented on HADOOP-3333:
------------------------------------
Yes even I saw these things in the logs.
bq. However a lost tasktracker leads to tasks being marked KILLED.
As this is different from FAILED, we should probably keep it as it is.
Tasktracker can be lost because of some transient issues. Blacklisting such
trackers for the tips that are local might not be good. But a tracker getting
lost too frequently can be considered for blacklisting.
- Can we do something regarding the machines where the trackers (on different
ports) are failing or getting lost too often?
bq. We also have to track hostnames rather than 'trackernames', trackername
includes the host:port... (#2)
+1. What can be done is that the TIP can be scheduled to trackers on different
machines in the first go. Then consider scheduling it to TaskTrackers sharing
the machines.
----
One thing I felt was the system was loaded. That could be the possible reason
for job failures. Wondering under what conditions running multiple TaskTrackers
is better than running one tracker.
> job failing because of reassigning same tasktracker to failing tasks
> --------------------------------------------------------------------
>
> Key: HADOOP-3333
> URL: https://issues.apache.org/jira/browse/HADOOP-3333
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Affects Versions: 0.16.3
> Reporter: Christian Kunz
> Assignee: Arun C Murthy
> Priority: Blocker
>
> We have a long running a job in a 2nd atttempt. Previous job was failing and
> current jobs risks to fail as well, because reduce tasks failing on marginal
> TaskTrackers are assigned repeatedly to the same TaskTrackers (probably
> because it is the only available slot), eventually running out of attempts.
> Reduce tasks should be assigned to the same TaskTrackers at most twice, or
> TaskTrackers need to get some better smarts to find failing hardware.
> BTW, mapred.reduce.max.attempts=12, which is high, but does not help in this
> case.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
