[GitHub] [spark] akpatnam25 commented on a diff in pull request #36601: [SPARK-38987][SHUFFLE] Throw FetchFailedException when merged shuffle blocks are corrupted and spark.shuffle.detectCorrupt is set to true

[GitBox <gi...@apache.org>]

akpatnam25 commented on code in PR #36601:
URL: https://github.com/apache/spark/pull/36601#discussion_r876437163


##########
core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala:
##########
@@ -4342,6 +4342,56 @@ class DAGSchedulerSuite extends SparkFunSuite with TempLocalSparkContext with Ti
     assertDataStructuresEmpty()
   }
 
+  test("SPARK-38987: corrupted shuffle block FetchFailure should unregister merge result") {
+    initPushBasedShuffleConfs(conf)
+    DAGSchedulerSuite.clearMergerLocs()
+    DAGSchedulerSuite.addMergerLocs(Seq("host1", "host2", "host3", "host4", "host5"))
+
+    scheduler = new MyDAGScheduler(
+      sc,
+      taskScheduler,
+      sc.listenerBus,
+      mapOutputTracker,
+      blockManagerMaster,
+      sc.env,
+      shuffleMergeFinalize = false,
+      shuffleMergeRegister = false)
+    dagEventProcessLoopTester = new DAGSchedulerEventProcessLoopTester(scheduler)
+
+    val parts = 2
+    val shuffleMapRdd = new MyRDD(sc, parts, Nil)
+    val shuffleDep = new ShuffleDependency(shuffleMapRdd, new HashPartitioner(parts))
+    val reduceRdd = new MyRDD(sc, parts, List(shuffleDep), tracker = mapOutputTracker)
+
+    // Submit a reduce job that depends which will create a map stage
+    submit(reduceRdd, (0 until parts).toArray)
+
+    // Pass in custom bitmap so that the mergeStatus can store
+    // the correct mapIndex.
+    val bitmap = new RoaringBitmap()
+    bitmap.add(-1)
+
+    val shuffleMapStage = scheduler.stageIdToStage(0).asInstanceOf[ShuffleMapStage]
+    scheduler.handleRegisterMergeStatuses(shuffleMapStage,
+      Seq((0, MergeStatus(makeBlockManagerId("hostA"), shuffleDep.shuffleMergeId, bitmap, 1000L))))
+    scheduler.handleShuffleMergeFinalized(shuffleMapStage,
+      shuffleMapStage.shuffleDep.shuffleMergeId)
+    scheduler.handleRegisterMergeStatuses(shuffleMapStage,
+      Seq((1, MergeStatus(makeBlockManagerId("hostA"), shuffleDep.shuffleMergeId, bitmap, 1000L))))
+

Review Comment:
   I tried using that, but `makeMergeStatus` uses a mock bitmap - so the mapIndex is never present inside the `MergeStatus`. Hence, I opted to use the actual class, and it seemed like this was the convention being used for triggering `MapOutputTracker.unregisterMergeResult` inside `MapOutputTrackerSuite`.  



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org