我对集群上的执行流有一个奇怪的问题。
方法a调用方法b，方法b在foreach中调用方法c
执行流程应该是

Method A --> Method B --> Method C

但它的工作原理是这样的：

1) Method A --> Method B (skips Method C) and continues to rest of Method B.
2) Method C is executed later separately.

因为流量不正确， accum1.value 在方法b中显示为 blank/null .

**CLASS A::METHOD A:**

object TakeDFs {

    def takeDFs(df: DataFrame): Unit = {
        println("---------------- takeNettedDFs::START ---------------- ")

        for(i <- 0 until bySecurityArray.length) {
            allocProcessDF = bySecurityArray(i).toDF()
            ....

            //WORKS
            AllocOneProcess.getAllocOneDFs(allocProcessDF)

            }
        println("---------------- takeNettedDFs::END ---------------- ")

    }
}

**CLASS B::METHOD B:**

object AllocOneProcess {

    def getAllocOneDFs(df: DataFrame): Unit = {
        println("---------------- getAllocOneDFs::START ---------------- ")

        df.coalesce(1).sort($"PRIORITY" asc).foreach( {
        row => AllocOneTest.allocProcessTest(row)
        })

        println("------------- getAllocOneDFs::accum1.value -------------" + accum1.value)

        println("---------------- getAllocOneDFs::END ---------------- ")

    }
}

**CLASS C::METHOD C:**

object AllocOneTest {

    def allocProcessTest(row: Row): Unit =  {
        println("---------------- AllocOneTest::allocProcessTest::START ---------------- ")

        accum1.add(RegRptPilotConstants.PairProcessCaseClass(row(0).asInstanceOf[String], row(1).asInstanceOf[String], row(2).asInstanceOf[String]))

        println("---------------- AllocationOneTest::allocProcessTest::END ---------------- ")

    }
}

**CLASS D::**

object RegRptPilotConstants {
    var pairedOneSeq = Seq[PairProcessCaseClass]()
    val accum1 = new ProcessAccumulator[ProcessCaseClass]()

}