Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/3518#discussion_r21062406
--- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
---
@@ -788,6 +793,63 @@ class DAGScheduler(
}
}
+ /**
+ * Helper function to check whether an RDD is serializable.
+ *
+ * Note: This function is defined seperately from the
SerializationHelper.isSerializable()
+ * since DAGScheduler.isSerialiazble() is passed as a parameter to the
RDDWalker class's graph
+ * traversal, which would otherwise require knowledge of the
closureSerializer
+ * (which was undesirable).
+ *
+ * @param rdd - Rdd to attempt to serialize
+ * @return - An output string qualifying success or failure.
+ */
+ private def isSerializable(rdd: RDD[_]): String = {
+ SerializationHelper.isSerializable(closureSerializer,rdd)
+ }
+
+ /**
+ * Use the RDDWalker class to execute a graph traversal of an RDD and
its dependencies to help
+ * identify which RDDs are not serializable. In short, attempt to
serialize the RDD and catch
+ * any Exceptions thrown (this is the same mechanism used within
submitMissingTasks() to deal with
+ * serialization failures).
+ *
+ * Note: This is defined here since it uses the isSerializale function
which in turn uses
+ * the closure serializer. Although the better place for the serializer
would be in the
+ * SerializationHelper, the Helper is not guaranteed to run in a single
thread unlike the
+ * DAGScheduler.
+
+ *
+ * @param rdd - The rdd for which to print the serialization trace to
identify unserializable
+ * components
+ * @return - String - The serialization trace
+ *
+ */
+ def getSerializationTrace(rdd : RDD[_]): String = {
+ // Next, if there are dependencies, attempt to serialize those
+ val results: util.ArrayList[String] = RDDWalker.walk(rdd,
isSerializable)
+
+ var trace = "Serialization trace:\n"
+
+ val it = results.iterator()
+ while(it.hasNext){
+ trace += it.next() + "\n"
+
+ }
+
+ trace
+ }
+
+ /**
+ * Use the RDD toDebugString function to print a formatted dependency
trace for an RDD
+ * @param rdd - The RDD for which to print the dependency graph
+ * @return
+ */
+ def getDependencyTrace(rdd: RDD[_]): String ={
--- End diff --
Why not just call `getDebugString`? What does this level of indirection
add?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
