[hadoop]MAPREDUCE-4888 Report

1. Symptom

NLineInputFormat drops data

1.1 Severity


1.2 Was there exception thrown?

No Exception

1.2.1 Were there multiple exceptions?


1.3 Scope of the failure

Affects all version beyond MR 1.1 while calling NLineInputFormat function

2. How to reproduce this failure

Call NLineInputFormat function

2.0 Version


2.1 Configuration

- map reduce library

2.2 Reproduction procedure

1. call NLineInputFormat function (feature start)

2.2.1 Timing order

Timing is not relevant here. As long as the code path goes through NLIneInputFormat, MR will drop data

2.2.2 Events order externally controllable?

Yes, but not applicable

2.3 Can the logs tell how to reproduce the failure?


2.4 How many machines needed?


3. Diagnosis procedure

3.1 Detailed Symptom (where you start)

 NLIneInputFormat drops data

3.2 Backward inference

Took a look at code and found the version included the wrong version

4. Root cause

Wrong merge of different code versions.

4.1 Category:

semantic (used wrong version of the code)

5. Fix

5.1 How?

Fix the right implementation in 1.0.2.