pyspark in Ipython notebook raises Py4JNetworkError

Question

I was using IPython notebook to run PySpark with just adding the following to the notebook:

import os
os.chdir('../data_files')
import sys
import pandas as pd
%pylab inline
from IPython.display import Image
os.environ['SPARK_HOME']="spark-1.3.1-bin-hadoop2.6"
sys.path.append( os.path.join(os.environ['SPARK_HOME'], 'python') )
sys.path.append( os.path.join(os.environ['SPARK_HOME'], 'bin') )
sys.path.append( os.path.join(os.environ['SPARK_HOME'], 'python/lib/py4j-0.8.2.1-src.zip') )
from pyspark import SparkContext
sc = SparkContext('local')

This worked fine for one project. but on my second project, after running a couple of lines (not the same every time), I get the following error:

ERROR:py4j.java_gateway:An error occurred while trying to connect to the Java server
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.py", line 425, in start
    self.socket.connect((self.address, self.port))
  File "/usr/lib/python2.7/socket.py", line 224, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 111] Connection refused
---------------------------------------------------------------------------
Py4JNetworkError                          Traceback (most recent call last)
<ipython-input-21-4626925bbe8f> in <module>()
----> 1 words.count()

/home/eee/Desktop/NLP/spark-1.3.1-bin-hadoop2.6/python/pyspark/rdd.pyc in count(self)
    930         3
    931         """
--> 932         return self.mapPartitions(lambda i: [sum(1 for _ in i)]).sum()
    933 
    934     def stats(self):

/home/eee/Desktop/NLP/spark-1.3.1-bin-hadoop2.6/python/pyspark/rdd.pyc in sum(self)
    921         6.0
    922         """
--> 923         return self.mapPartitions(lambda x: [sum(x)]).reduce(operator.add)
    924 
    925     def count(self):

/home/eee/Desktop/NLP/spark-1.3.1-bin-hadoop2.6/python/pyspark/rdd.pyc in reduce(self, f)
    737             yield reduce(f, iterator, initial)
    738 
--> 739         vals = self.mapPartitions(func).collect()
    740         if vals:
    741             return reduce(f, vals)

/home/eee/Desktop/NLP/spark-1.3.1-bin-hadoop2.6/python/pyspark/rdd.pyc in collect(self)
    710         Return a list that contains all of the elements in this RDD.
    711         """
--> 712         with SCCallSiteSync(self.context) as css:
    713             port = self.ctx._jvm.PythonRDD.collectAndServe(self._jrdd.rdd())
    714         return list(_load_from_socket(port, self._jrdd_deserializer))

/home/eee/Desktop/NLP/spark-1.3.1-bin-hadoop2.6/python/pyspark/traceback_utils.pyc in __enter__(self)
     70     def __enter__(self):
     71         if SCCallSiteSync._spark_stack_depth == 0:
---> 72             self._context._jsc.setCallSite(self._call_site)
     73         SCCallSiteSync._spark_stack_depth += 1
     74 

/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.pyc in __call__(self, *args)
    534             END_COMMAND_PART
    535 
--> 536         answer = self.gateway_client.send_command(command)
    537         return_value = get_return_value(answer, self.gateway_client,
    538                 self.target_id, self.name)

/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.pyc in send_command(self, command, retry)
    360          the Py4J protocol.
    361         """
--> 362         connection = self._get_connection()
    363         try:
    364             response = connection.send_command(command)

/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.pyc in _get_connection(self)
    316             connection = self.deque.pop()
    317         except Exception:
--> 318             connection = self._create_connection()
    319         return connection
    320 

/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.pyc in _create_connection(self)
    323         connection = GatewayConnection(self.address, self.port,
    324                 self.auto_close, self.gateway_property)
--> 325         connection.start()
    326         return connection
    327 

/usr/local/lib/python2.7/dist-packages/py4j-0.8.2.1-py2.7.egg/py4j/java_gateway.pyc in start(self)
    430                 'server'
    431             logger.exception(msg)
--> 432             raise Py4JNetworkError(msg)
    433 
    434     def close(self):

Py4JNetworkError: An error occurred while trying to connect to the Java server

Once this happens, other lines working before now raise the same problem, any ideas?

Answer 1

Specifications for:

• pyspark 1.4.1

• ipython 4.0.0

• [OSX / homebrew]

If you want to launch pyspark within a Jupyter (ex-iPython) Notebook using the iPython kernel, I advise you to launch your notebook directly with the pyspark command:

>>>pyspark

But in order to do that, you need to add three lines in your bash .profile or zsh .zshrc profile to set these environment variables:

export SPARK_HOME=/path/to/apache-spark/1.4.1/libexec
export PYSPARK_DRIVER_PYTHON=ipython2 # remember that Apache-Spark only works with pyhton2.7
export PYSPARK_DRIVER_PYTHON_OPTS='notebook'

In my case, given that I'm on OSX , an installed apache-spark with Homebrew, this is:

export SPARK_HOME=/usr/local/Cellar/apache-spark/1.4.1/libexec
export PYSPARK_DRIVER_PYTHON=ipython2
export PYSPARK_DRIVER_PYTHON_OPTS='notebook'

Then, when you execute the command 'pyspark' in your terminal, your terminal will automatically open a Jupyter (ex-iPython) notebook in your default Browser.

>>>pyspark
I 17:51:00.209 NotebookApp] Serving notebooks from local directory: /Users/Thibault/code/kaggle
[I 17:51:00.209 NotebookApp] 0 active kernels
[I 17:51:00.210 NotebookApp] The IPython Notebook is running at: http://localhost:42424/
[I 17:51:00.210 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[I 17:51:11.980 NotebookApp] Kernel started: 53ad11b1-4fa4-459d-804c-0487036b0f29
15/09/02 17:51:15 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable