Installing Modules for SPARK on worker nodes

Question

I am running SPARK 1.3 in standalone mode in a cloudera environment. I can run pyspark from ipython notebook, however as soon as I add a second worker node my code stops running and returns an error. I am pretty sure this is because modules on my master are not visible to the worker node. I tried importing numpy but it didn't work even though I have numpy installed on my worker through anaconda. I have anaconda installed on both master and worker in the same way.

However, following Josh Rosen's advice I made sure that I installed the libraries on the worker nodes.

https://groups.google.com/forum/#!topic/spark-users/We_F8vlxvq0

However, I still seem to be getting issues. Including the fact that my worker does not recognize the command abs. which is standard in python 2.6

The code I am running is from this post:

https://districtdatalabs.silvrback.com/getting-started-with-spark-in-python

def isprime(n):
    """
    check if integer n is a prime
    """
    # make sure n is a positive integer
    n = abs(int(n))
    # 0 and 1 are not primes
    if n < 2:
        return False
    # 2 is the only even prime number
    if n == 2:
        return True
    # all other even numbers are not primes
    if not n & 1:
        return False
    # range starts with 3 and only needs to go up the square root of n
    # for all odd numbers
    for x in range(3, int(n**0.5)+1, 2):
        if n % x == 0:
            return False
    return True

# Create an RDD of numbers from 0 to 1,000,000
nums = sc.parallelize(xrange(1000000))

# Compute the number of primes in the RDD
print nums.filter(isprime).count()

Answer 1

I often use the anaconda distribution with PySpark as well and find it useful to set the PYSPARK_PYTHON variable, pointing to the python binary within the anaconda distribution. I've found that otherwise I get lots of strange errors. You might be able to check with python is being used by running rdd.map(lambda x: sys.executable).distinct().collect(). I suspect it's not pointing to the correct location.

In any case, I recommend wrapping the configuration of your path and environment variables in a script. I use the following.

def configure_spark(spark_home=None, pyspark_python=None):
    spark_home = spark_home or "/path/to/default/spark/home"
    os.environ['SPARK_HOME'] = spark_home

    # Add the PySpark directories to the Python path:
    sys.path.insert(1, os.path.join(spark_home, 'python'))
    sys.path.insert(1, os.path.join(spark_home, 'python', 'pyspark'))
    sys.path.insert(1, os.path.join(spark_home, 'python', 'build'))

    # If PySpark isn't specified, use currently running Python binary:
    pyspark_python = pyspark_python or sys.executable
    os.environ['PYSPARK_PYTHON'] = pyspark_python

When you point to your anaconda binary, you should also be able to import all the packages installed in its site-packages directory. This technique should work for conda environments as well.