spark-hive-streaming-sink | save Spark Structured Streaming DataFrame

There are two ways to solve your problem:

1. Delete/clear your checkpoint: /checkpoints/employee/checkpoint on your machine

2. Using the other source that maintains offset, like Kafka

The reason you meet this issue is that the socket doesn't maintain offset information.

When you restart your job which receives input data from socket 9999, the first thing that your job does is trying to recover the state from /checkpoints/employee/checkpoint, and it finds that your current offset that has been recorded is 1. Then you input other messages to socket 9999, your job finds that the offset from socket 9999 is 0. So it throws this exception.