Scala Spark行级错误处理

Question

I'm having some trouble figuring out how to do some row-level error handling with a Scala Spark program. In the code below, I'm reading in a CSV text file, parsing it, and creating a Row using a mapSchema method (not shown; basically, it takes the Array of strings that result from the CSV and uses a schema to convert the strings into ints, doubles, dates, etc.). It works great when the data is all formatted appropriately. However, if I have a bad row -- for example, one with fewer fields than expected -- I want to perform some error handling.

val rddFull = sqlContext.sparkContext.textFile(csvPath).map {
  case(txt) =>
    try {
      val reader = new CSVReader(new StringReader(txt), delimiter, quote, escape, headerLines)
      val parsedRow = reader.readNext()
      Row(mapSchema(parsedRow, schema) : _*)
    } catch {
      case err: Throwable =>
        println("a record had an error: "+ txt)
        throw new RuntimeException("SomeError")
    }

The problem is that the try/catch expressions don't seem to be working. When I give it bad row, I don't ever get the "SomeError" RuntimeException. Instead, I get the same error that I get when I don't use try/catch.

Any ideas about what could be going wrong here?

Answer 1

You need to look in the correct place for the logs . To start with: the catch does work. Here is an example from the spark-shell:

val d = sc.parallelize(0 until 10)
val e = d.map{ n =>
  try {
   if (n % 3==0) throw new IllegalArgumentException("That was a bad call")
   println(n)
 } catch {
    case e:  IllegalArgumentException =>  throw new UnsupportedOperationException("converted from Arg to Op except")
 }
}
e.collect

Here is the result: notice the exception were properly caught and converted:

org.apache.spark.SparkException: Job aborted due to stage failure: Task 5 in
stage 0.0 failed 1 times, most recent failure: Lost task 5.0 in   
stage 0.0 (TID 5, localhost): 
java.lang.UnsupportedOperationException: converted from Arg to Op except
    at $anonfun$1.apply$mcVI$sp(<console>:29)
    at $anonfun$1.apply(<console>:24)
    at $anonfun$1.apply(<console>:24)

Try looking in the stderr logs of one or more of the workers.