How can I include a python package with Hadoop streaming job?

Question

I am trying include a python package (NLTK) with a Hadoop streaming job, but am not sure how to do this without including every file manually via the CLI argument, \"-file\"

Answer 1

If you are using much more complex libs such as numpy、pandas, virtualenv is a better way. You can add -archives to send the env to cluster.

Refer to the writing: https://henning.kropponline.de/2014/07/18/virtualenv-hadoop-streaming/

Updated:

I tried above virtualenv in our online env, and find some problems.In the cluster，there is some errors like "Could not find platform independent libraries "。Then i tried the conda to create python env, it worked well.

If you are Chinese, you can look this:https://blog.csdn.net/Jsin31/article/details/53495423

If not, i can translate it briefly:

1. create an env by conda：

   conda create -n test python=2.7.12 numpy pandas

2. Go to the conda env path.You can find it by cmd:

   conda env list

   Then,you can pack it:

   tar cf test.tar test

3. submit the job through hadoop stream：

hadoop jar /usr/lib/hadoop/hadoop-streaming.jar \
-archives test.tar \
-input /user/testfiles \
-output /user/result \ 
-mapper "test.tar/test/bin/python mapper.py" \
-file mapper.py \
-reducer"test.tar/test/bin/python reducer.py" \
-file reducer.py