Spark Thrift Server Feature Support

This topic describes the expected behavior and feature support for Spark Thrift Server.

Running Multiple Spark Thrift Server Instances (HA Mode)

Spark Thrift Server on Kubernetes supports HA mode using the native Kubernetes service concept. You use the spec.tenantservices.sparkts.count value to set the quantity of Spark Thrift Server deployments in the application configuration. All deployments are exposed through a single NodePort service. This endpoint should be used by the Spark Thrift Server clients.

Expected Behavior for Spark Thrift Jobs

Executor pods transition to a Completed state but are not removed after stopping or restarting the Spark Thrift Server. This is a normal behavior, and the status of the job can be checked in the completed log.