[
https://issues.apache.org/jira/browse/MAPREDUCE-3815?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13202685#comment-13202685
]
Siddharth Seth commented on MAPREDUCE-3815:
-------------------------------------------
Looked at this a little more.
This shows up when a split spans across multiple blocks.
{{getFileBlockLocations}} always returns hostnames. In case of multiple blocks,
mapred.FileInputFormat ends up using {{BlockLocations.getTopologyPaths}}
instead of getFileBlockLocations - which returns an IP address.
Will open a MR / HDFS jira once I can find out how this API behaves in the 1.0
line. Anyone happen to know ?
Meanwhile, changing the description and posting a patch to have the AM resolve
IPs if they show up.
> Data Locality suffers if HDFS returns IPs in getFileBlockLocations
> ------------------------------------------------------------------
>
> Key: MAPREDUCE-3815
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-3815
> Project: Hadoop Map/Reduce
> Issue Type: Sub-task
> Components: mrv2
> Affects Versions: 0.23.0
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Priority: Critical
> Attachments: MR3815.txt
>
>
> BlockLocation.getHosts() returns IP addresses occasionally. Data locality is
> affected - since the RM requires hostnames.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
