java.io.IOException: Cannot run program “python” u

问题:

I am trying to write a very simple code using Spark in Pycharm and my os is Windows 8. I have been dealing with several problems which somehow managed to fix except for one. When I run the code using pyspark.cmd everything works smoothly but I have had no luck with the same code in pycharm. There was a problem with SPARK_HOME variable which I fixed using the following code:

import sys
import os
os.environ['SPARK_HOME'] = "C:/Spark/spark-1.4.1-bin-hadoop2.6"
sys.path.append("C:/Spark/spark-1.4.1-bin-hadoop2.6/python")
sys.path.append('C:/Spark/spark-1.4.1-bin-hadoop2.6/python/pyspark')

So now when I import the pyspark and everything is fine:

from pyspark import SparkContext

The problem rises when I want to run the rest of my code:

logFile = "C:/Spark/spark-1.4.1-bin-hadoop2.6/README.md"
sc = SparkContext()
logData = sc.textFile(logFile).cache()
logData.count()

When I receive the following error:

15/08/27 12:04:15 ERROR Executor: Exception in task 0.0 in stage 0.0 (TID 0)
java.io.IOException: Cannot run program "python": CreateProcess error=2, The system cannot find the file specified

I have added the python path as an environment variable and it's working properly using the command line but I could not figure out what my problem is with this code. Any help or comment is much appreciated.

Thanks

回答1:

After struggling with this for two days, I figured what the problem is. I added the followings to the "PATH" variable as windows environment variable:

C:/Spark/spark-1.4.1-bin-hadoop2.6/python/pyspark
C:\Python27

Remember, You need to change the directory to wherever your spark is installed and also the same thing for python. On the other hand, I have to mention that I am using prebuild version of spark which has Hadoop included.

Best of luck to you all.

回答2:

I had the same problem as you, and then I made the following changes: set PYSPARK_PYTHON as environment variable to point to python.exe in Edit Configurations of Pycharm, here is my example:

PYSPARK_PYTHON = D:\Anaconda3\python.exe

SPARK_HOME = D:\spark-1.6.3-bin-hadoop2.6

PYTHONUNBUFFERED = 1

回答3:

I have faced this problem, it's caused by python version conflicts on diff nodes of cluster, so, it can be solved by

export PYSPARK_PYTHON=/usr/bin/python

which are the same version on diff nodes. and then start:

pyspark

回答4:

I had to set SPARK_PYTHONPATH as environment variable to point to python.exe file in addition to PYTHONPATH and SPARK_HOME variables as

SPARK_PYTHONPATH=C:\Python27\python.exe