I have a docker image containing various bits, including Spark. Here is my Dockerfile:
FROM docker-dev.artifactory.company.com/centos:7.3.1611
# set proxy
ENV http_proxy http://proxyaddr.co.uk:8080
ENV HTTPS_PROXY http://proxyaddr.co.uk:8080
ENV https_proxy http://proxyaddr.co.uk:8080
RUN yum install -y epel-release
RUN yum install -y gcc
RUN yum install -y krb5-devel
RUN yum install -y python-devel
RUN yum install -y krb5-workstation
RUN yum install -y python-setuptools
RUN yum install -y python-pip
RUN yum install -y xmlstarlet
RUN yum install -y wget java-1.8.0-openjdk
RUN pip install kerberos
RUN pip install numpy
RUN pip install pandas
RUN pip install coverage
RUN pip install tensorflow
RUN wget http://d3kbcqa49mib13.cloudfront.net/spark-1.6.0-bin-hadoop2.6.tgz
RUN tar -xvzf spark-1.6.0-bin-hadoop2.6.tgz -C /opt
RUN ln -s spark-1.6.0-bin-hadoop2.6 /opt/spark
ENV VERSION_NUMBER $(cat VERSION)
ENV JAVA_HOME /etc/alternatives/jre/
ENV SPARK_HOME /opt/spark
ENV PYTHONPATH $SPARK_HOME/python/:$PYTHONPATH
ENV PYTHONPATH $SPARK_HOME/python/lib/py4j-0.9-src.zip:$PYTHONPATH
I can build then run that docker image, connect to it, and successfully import the pyspark libraries:
$ docker run -d -it sse_spark_build:1.0
09e8aac622d7500e147a6e6db69f806fe093b0399b98605c5da2ff5e0feca07c
$ docker exec -it 09e8aac622d7 python
Python 2.7.5 (default, Nov 6 2016, 00:28:07)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-11)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from pyspark import SparkContext
>>>import os
>>> os.environ['PYTHONPATH']
'/opt/spark/python/lib/py4j-0.9-src.zip:/opt/spark/python/:'
>>>
Note the value of PYTHONPATH!
Problem is that the behaviour in PyCharm is different if I use this same docker image as the interpreter. Here's how I have set up the interpreter:
If I then run Python console in PyCharm this happens:
bec0b9189066:python /opt/.pycharm_helpers/pydev/pydevconsole.py 0 0
PyDev console: starting.
import sys; print('Python %s on %s' % (sys.version, sys.platform))
sys.path.extend(['/home/cengadmin/git/dhgitlab/sse/engine/fs/programs/pyspark', '/home/cengadmin/git/dhgitlab/sse/engine/fs/programs/pyspark'])
Python 2.7.5 (default, Nov 6 2016, 00:28:07)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-11)] on linux2
import os
os.environ['PYTHONPATH']
'/opt/.pycharm_helpers/pydev'
As you can see PyCharm has changed PYTHONPATH meaning that I can no longer use the pyspark libraries that I want to use:
from pyspark import SparkContext
Traceback (most recent call last):
File "<input>", line 1, in <module>
ImportError: No module named pyspark
OK, I could change PATH from the console to make it work:
import sys
sys.path.append('/opt/spark/python/')
sys.path.append('/opt/spark/python/lib/py4j-0.9-src.zip')
but its tedious to have to do that every time I open a console. I can't believe there isn't a way of telling PyCharm to append to PYTHONPATH rather than overwriting it but if there is I can't find it. Can anyone offer any advice? How can I use a docker image as PyCharm's remote interpreter and keep the value of PYTHONPATH?