In order to use Spark in Python, environment has to bee set up according to Using Spark in UFAL Environment.

Interactive shell can be started using:

pyspark

Better interactive shell with code completion using ipython3 (do `pip3 install –user ipython` if you do not have it) can be started using:

PYSPARK_DRIVER_PYTHON=ipython3 pyspark

As described in Running Spark on Single Machine or on Cluster, environmental variable MASTER specifies which Spark master to use (or whether to start a local one).

Consider the following simple script computing 10 most frequent words of Czech Wikipedia:

(sc.textFile("/net/projects/spark-example-data/wiki-cs", 3*sc.defaultParallelism)
   .flatMap(lambda line: line.split())
   .map(lambda word: (word, 1))
   .reduceByKey(lambda c1, c2: c1+c2)
   .sortBy(lambda word_count: word_count[1], ascending=False)
   .take(10))

• run interactive shell using existing Spark cluster (i.e., inside spark-srun), or start local Spark cluster using as many threads as there are cores if there is none:

PYSPARK_DRIVER_PYTHON=ipython3 pyspark

• run interactive shell with local Spark cluster using one thread:

MASTER=local PYSPARK_DRIVER_PYTHON=ipython3 pyspark

• start Spark cluster (10 machines, 2GB RAM each) on Slurm and run interactive shell:

PYSPARK_DRIVER_PYTHON=ipython3 spark-srun 10 2G pyspark

Note that PYSPARK_DRIVER_PYTHON variable can be left out or specified in .bashrc (or other configuration files).

Python Spark scripts can be started using:

spark-submit

As described in Running Spark on Single Machine or on Cluster, environmental variable MASTER specifies which Spark master to use (or whether to start a local one).

Consider the following simple word-count script word_count.py:

#!/usr/bin/python
 
import sys
if len(sys.argv) < 3:
    print >>sys.stderr, "Usage: %s input output" % sys.argv[0]
    exit(1)
input = sys.argv[1]
output = sys.argv[2]
 
from pyspark import SparkContext
 
sc = SparkContext()
(sc.textFile(input, 3*sc.defaultParallelism)
   .flatMap(lambda line: line.split())
   .map(lambda token: (token, 1))
   .reduceByKey(lambda x, y: x + y)
   .sortBy(lambda word_count: word_count[1], ascending=False)
   .saveAsTextFile(output))
sc.stop()

• run word_count.py script inside existing Spark cluster (i.e., inside spark-sbatch or spark-srun), or start local Spark cluster using as many threads as there are cores if there is none:

spark-submit word_count.py /net/projects/spark-example-data/wiki-cs outdir

• run word_count.py script with local Spark cluster using one thread:

MASTER=local spark-submit word_count.py /net/projects/spark-example-data/wiki-cs outdir

• start Spark cluster (10 machines, @GB RAM each) using Slurm and run word_count.py script:

spark-sbatch 10 2G spark-submit word_count.py /net/projects/spark-example-data/wiki-cs outdir

If you want to use specific virtual environment in your Spark job, use

PYSPARK_PYTHON=path_to_python_in_venv [pyspark|spark-submit]