I was trying out the simplistic word count example for hadoop pipes. Unfortunately it is erroring out with java.lang.NullPointerException and /usr/lib64/libstdc++.so.6: no version information available errors when I try to run it as:
$ bin/hadoop pipes -Dhadoop.pipes.java.recordreader=true \
-Dhadoop.pipes.java.recordwriter=true \
-input <inputHDFSPath> \
-output <outputHDFSPath> \
-program <cppBinaryHDFSPath>
The error details are here.
Any suggestions on how to fix this would be greatly appreciated. Thanks.
Edit1: Added command used to run.
Edit2: It turns out that the C++ binary wasn’t statically linked. After rectifying that I am now seeing this error.
Edit3: Added hadoop version being used. I have also added relevant portions of the jobtracker logs here. Also, stderr for each of the failed attempts has this:
$more /usr/local/dp/hadoop/userlogs/job_201107211831_0013/\
attempt_201107211831_0013_m_000000_0/stderr
Hadoop Pipes Exception: Aborted by driver
So I gave it another shot and started from scratch. This time round I had better success. Here is the code and the config files I used.
Also, here’s the command I used to run the job:
The input files in the
/path/to/inputwere:And the output was:
(Apologies for the wonky paths. My actual paths were different – the paths here are just my lazy attempts at anonymizing them.)