[
https://issues.apache.org/jira/browse/HADOOP-4011?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12625062#action_12625062
]
Jothi Padmanabhan commented on HADOOP-4011:
-------------------------------------------
To prevent premature re-execution of maps, the back off between two retries is
currently set to MAP_RUN_TIME/2 and the reducer will intimate JT of failure on
three READ_TIME_OUT FAILURES. The back off logic, as it exists now,
unfortunately does not take the load of the system into consideration.
> shuffling backoff 5 minutes despite the whole cluster is idle
> -------------------------------------------------------------
>
> Key: HADOOP-4011
> URL: https://issues.apache.org/jira/browse/HADOOP-4011
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Reporter: Runping Qi
>
> A reducer of a mapred job got a read time out exception during fetching the
> last map output segment.
> The reducer waited 5 minutes before re-fetch the map output, even though the
> whole cluster was idle during that period.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
