Hello, I am having a scala unit function which use scala

groupBy

API running in flink job, and it has different results when I run this function in a unit test with normal jvm context and run this function in the flink job (with parallelism 1 and only 1 taskmanager and 1 task slot ). Anyone knows something we should be careful to configure in our flink cluster so the computation result are the same in unit test and flink run time ?