How to start Spark session on Vertex AI workbench Jupyterlab notebook?

Question

Can you kindly show me how do we start the Spark session on Google Cloud Vertex AI workbench Jupyterlab notebook?
This is working fine in Google Colaboratory by the way.
What is missing here?

# Install Spark NLP from PyPI
!pip install -q spark-nlp==4.0.1 pyspark==3.3.0

import os
import sys

# https://github.com/jupyter/jupyter/issues/248
os.environ["JAVA_HOME"] = "C:/Program Files/Java/jdk-18.0.1.1"
os.environ["PATH"] = os.environ["JAVA_HOME"] + "/bin:" + os.environ["PATH"]

import sparknlp

from sparknlp.base import *
from sparknlp.common import *
from sparknlp.annotator import *

from pyspark.ml import Pipeline
from pyspark.sql import SparkSession

import pandas as pd

spark=sparknlp.start() 

print("Spark NLP version: ", sparknlp.version())
print("Apache Spark version: ", spark.version)

spark

UPDATE_2022-07-21:
Hi @Sayan. I am still not able to start Spark session on Vertex AI workbench Jupyterlab notebook after running the commands =(

# Install Spark NLP from PyPI
!pip install -q spark-nlp==4.0.1 pyspark==3.3.0

import os
# Included else "JAVA_HOME is not set"
# https://github.com/jupyter/jupyter/issues/248
os.environ["JAVA_HOME"] = "C:/Program Files/Java/jdk-18.0.1.1"
os.environ["PATH"] = os.environ["JAVA_HOME"] + "/bin:" + os.environ["PATH"]

import sparknlp
spark = sparknlp.start()

print("Spark NLP version: {}".format(sparknlp.version()))
print("Apache Spark version: {}".format(spark.version))

The error:

/opt/conda/lib/python3.7/site-packages/pyspark/bin/spark-class: line 71: C:/Program Files/Java/jdk-18.0.1.1/bin/java: No such file or directory
/opt/conda/lib/python3.7/site-packages/pyspark/bin/spark-class: line 96: CMD: bad array subscript
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_5831/489505405.py in <module>
      6 
      7 import sparknlp
----> 8 spark = sparknlp.start()
      9 
     10 print("Spark NLP version: {}".format(sparknlp.version()))

/opt/conda/lib/python3.7/site-packages/sparknlp/__init__.py in start(gpu, m1, memory, cache_folder, log_folder, cluster_tmp_dir, real_time_output, output_level)
    242         return SparkRealTimeOutput()
    243     else:
--> 244         spark_session = start_without_realtime_output()
    245         return spark_session
    246 

/opt/conda/lib/python3.7/site-packages/sparknlp/__init__.py in start_without_realtime_output()
    152             builder.config("spark.jsl.settings.storage.cluster_tmp_dir", cluster_tmp_dir)
    153 
--> 154         return builder.getOrCreate()
    155 
    156     def start_with_realtime_output():

/opt/conda/lib/python3.7/site-packages/pyspark/sql/session.py in getOrCreate(self)
    267                         sparkConf.set(key, value)
    268                     # This SparkContext may be an existing one.
--> 269                     sc = SparkContext.getOrCreate(sparkConf)
    270                     # Do not update `SparkConf` for existing `SparkContext`, as it's shared
    271                     # by all sessions.

/opt/conda/lib/python3.7/site-packages/pyspark/context.py in getOrCreate(cls, conf)
    481         with SparkContext._lock:
    482             if SparkContext._active_spark_context is None:
--> 483                 SparkContext(conf=conf or SparkConf())
    484             assert SparkContext._active_spark_context is not None
    485             return SparkContext._active_spark_context

/opt/conda/lib/python3.7/site-packages/pyspark/context.py in __init__(self, master, appName, sparkHome, pyFiles, environment, batchSize, serializer, conf, gateway, jsc, profiler_cls, udf_profiler_cls)
    193             )
    194 
--> 195         SparkContext._ensure_initialized(self, gateway=gateway, conf=conf)
    196         try:
    197             self._do_init(

/opt/conda/lib/python3.7/site-packages/pyspark/context.py in _ensure_initialized(cls, instance, gateway, conf)
    415         with SparkContext._lock:
    416             if not SparkContext._gateway:
--> 417                 SparkContext._gateway = gateway or launch_gateway(conf)
    418                 SparkContext._jvm = SparkContext._gateway.jvm
    419 

/opt/conda/lib/python3.7/site-packages/pyspark/java_gateway.py in launch_gateway(conf, popen_kwargs)
    104 
    105             if not os.path.isfile(conn_info_file):
--> 106                 raise RuntimeError("Java gateway process exited before sending its port number")
    107 
    108             with open(conn_info_file, "rb") as info:

RuntimeError: Java gateway process exited before sending its port number

Answer 1

One possible reason is that Java is not installed. When you create a Python-3 Vertex AI Workbench you can have either Debian or Ubuntu as an OS and it does not come with Java pre-installed. You need to install it manually. To install you can use

sudo apt-get update
sudo apt-get install default-jdk

You can follow this tutorial to install Open JDK.

All your problems lie with installing JDK and setting its path in the environment. Once you do this properly you don't need to set path in python also. Your code should look something like this

# Install Spark NLP from PyPI
!pip install -q spark-nlp==4.0.1 pyspark==3.3.0

#no need to set the environment path

import sparknlp
#all other imports

import pandas as pd

spark=sparknlp.start() 

print("Spark NLP version: ", sparknlp.version())
print("Apache Spark version: ", spark.version)

spark

EDIT: I have tried your code and had the same error.All I did was Open the terminal inside JupyterLab of the workbench and installed java there.

Opened the JupyterLab from Workbench

Notebook instance.

Opening the terminal from File->New->Terminal

From here I downloaded and installed the Java.

You can check whether it has been installed and added to your path by running java --version it will return the current version.