Details
-
Type:
Improvement
-
Status: Open
-
Priority:
Critical
-
Resolution: Unresolved
-
Affects Version/s: CDH 4.7.1
-
Fix Version/s: None
-
Component/s: MapReduce
-
Labels:None
Description
Background:
I am working on the POC of SolrHadoop MapreduceIndexer Tool using CDH 4.x version for the past one year.
Problem:
During my last phase of comparing the index count, I found there were considerable indices missing.
Reason:
The cause is due to the fix for MAPREDUCE-4782 was not back ported to Mapreduce package.
This leads to the last input split in NLineInputFormat missing the first input path.
My Action:
I tried to migrate to CDH 5 and respective CM, but the UI and the internals seems to be different and my custom AMIs and security groups stopped working
I have already spent enough time in the past to freeze my ports and AMIs to spin up the hadoop cluster and do not want to move to YARN and fix the issues at this moment.
Action Needed:
It would be helpful if the fix is backported to Mapreduce MR v1 Framework as well on CDH 4.7 so that I can continue my works without delay.