Securing Spark connections

Communication between Spark applications and transactional nodes, masters and workers, and intercommunication between Spark drivers and executors can be encrypted.

Communication between Spark applications and transactional nodes, masters and workers, and intercommunication between Spark drivers and executors can be encrypted. Encryption must be configured individually on each node in the cluster.

Encryption between the Spark processes and DSE is configured by enabling client-to-node encryption in cassandra.yaml.

Encryption between Spark applications, including between the Spark master and worker, is configured by enabling Spark security in dse.yaml.

Encryption between the Spark driver and executors in client applications is configured by enabling Spark security in the application configuration properties, or by default in spark-defaults.conf and spark-daemon-defaults.conf in the Spark configuration directory.

The default location of the Spark configuration files depends on the type of installation:
  • Package installations: /etc/dse/spark/
  • Tarball installations: installation_location/resources/spark/conf

dse.yaml

The location of the dse.yaml file depends on the type of installation:
Package installations /etc/dse/dse.yaml
Tarball installations installation_location/resources/dse/conf/dse.yaml

cassandra.yaml

The location of the cassandra.yaml file depends on the type of installation:
Package installations /etc/dse/cassandra/cassandra.yaml
Tarball installations installation_location/resources/cassandra/conf/cassandra.yaml

Procedure

  1. Enable mutual authentication and encryption between Spark master and worker nodes in dse.yaml.
    Tip: In DSE 6.7.4 and later, when DSE authentication is enabled with in dse.yaml, Spark security is enabled regardless of these settings.
    1. Enable mutual authentication by setting spark_security_enabled to true.
    2. Enable encryption by setting spark_security_encryption_enabled to true.
    spark_security_enabled: true
    spark_security_encryption_enabled: true
    
    Note: To enable encryption, you must also enable mutual authentication.
  2. To encrypt communication between the Spark application and master, DSE inherits the client to cluster connection encryption options.
  3. The Spark web UI by default uses client-to-cluster encryption settings to enable SSL security in the web interface. To enable SSL security separately from DSE client-to-cluster encryption, change the settings in dse.yaml under spark_ui_options. Set encryption to custom, then set the keystore settings in encryption_options.
    The Spark web UI server automatically discards all cipher algorithms that end with a suffix SHA, SHA1 or MD5. If you manually specify cipher suites for encryption, make sure the specified cipher suites are not ignored by the server and are supported by the web browsers used to access the Spark web UI. For example, two 256 bit cipher suites that are supported by both the server and current web browsers are TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 and TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384.
  4. The AlwaysOn SQL server can optionally use SSL to encrypt client connections.
  5. Restart the node to recognize the changes.
  6. For each Spark application, set the following client encryption options to true in the spark-defaults.conf file to enable encryption between the Spark driver and executors.
    Option Description
    spark.authenticate Enables or disables mutual authentication using a secret key. The default is false.
    spark.authenticate.enableSaslEncryption Enables or disables SASL encryption between a Spark driver and its executors. The default is false.
    spark.io.encryption.enabled Enables or disables encryption for data stored by drivers directly on disk.

    To enable encryption by default for all Spark applications, modify the options in the spark-defaults.conf file in the Spark configuration directory. To encrypt data stored on the server by default, modify the options in the spark-daemon-defaults.conf file.

    Option Description
    spark.io.encryption.enabled Enables or disables encryption for data stored by executors and shuffle services directly on disk.