Kerberos support for Apache Spark

[Jimvin]

Implement Kerberos support for Apache Spark jobs.

[zsvr-pkruse]

Kerberos support would help connecting to MSSQL databases in a save manner removing the need for database specific users.

flowchart LR
  app["SparkApplication"]
  id["Service Account"]
  db[(MSSQL)]

  app -.keytab.- id
  app -.token.- id
  app --read-from--> db

Loading

As far as I know either a keytab file or service account credentials could be provided via k8s secret. The SparkApplication would need a Kerberos-able image and an appropriate kinit call before a spark-submit happens, but if set up this way, the application itself wouldn't need to handle credentials anymore to interact with databases.