Read Hadoop Credential in PySpark

In one of my previous articles about Password Security Solution for Sqoop, I mentioned creating credential using hadoop credential command. The credentials are stored in JavaKeyStoreProvider. Credential providers are used to separate the use of sensitive tokens, secrets and passwords from the details of their storage and management.

The following command lines create a credential named mydatabase.password in both local JCEKS file and also in HDFS.

#Store the password in HDFS

hadoop credential create mydatabase.password -provider jceks://hdfs/user/hue/mypwd.jceks

# Store the password locally

hadoop credential create mydatabase.password -provider jceks://file/home/user/mypwd.jceks

For running jobs in clusters like YARN, it is important to create the credential in HDFS so that it can be accessed by all worker nodes in the cluster.

Once the credential is created, you can easily use it in Sqoop by passing in the credential name as parameter. However, if you want to access the credential in Spark, what should you do? If you are using Scala, you can easily reference the Hadoop java libraries for credential. However, if you use Python as programming language, it won’t be that straightforward.

Sample code to retrieve Hadoop credential in PySpark

from pyspark.sql import SparkSession

appName = "PySpark Hadoop Credential Example"
master = "local"

# Create Spark session
spark = SparkSession.builder \
    .appName(appName) \
    .master(master) \
    .getOrCreate()

# Replace the credential provider path accordingly
credential_provider_path = 'jceks://hdfs/user/hue/.jceks' 
credential_name = 'mydatabase.password'

# Retrive credential/password from Hadoop credential
conf = spark.sparkContext._jsc.hadoopConfiguration()
conf.set('hadoop.security.credential.provider.path',credential_provider_path)
credential_raw = conf.getPassword(credential_name)
credential_str = ''
for i in range(credential_raw.__len__()):
    credential_str = credential_str + str(credential_raw.__getitem__(i))

# Now you can use credential_str, for example, use it as database password in JDBC to load data from databases into Spark data frame.

Access to the credential provider file

Anyone who has access to your credential provider file can also use the same approach to retrieve the credential value from the provider. So it is important to manage the access to the credential file so that only allowed users can access it.

More details about Hadoop credential API

Refer to the official page to learn more about Hadoop credential APIs: CredentialProvider API Guide.