Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/5964#discussion_r30862163
--- Diff:
core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala ---
@@ -545,9 +546,21 @@ class DAGSchedulerSuite
taskSet.tasks(1).epoch = newEpoch
runEvent(CompletionEvent(
taskSet.tasks(1), Success, makeMapStatus("hostA", 1), null,
createFakeTaskInfo(), null))
+
+ // now we should have a new taskSet for stage 0, which has us retry
partition 0
+ assert(taskSets.size === 2)
+ val newTaskSet = taskSets(1)
+ assert(newTaskSet.stageId === 0)
+ assert(newTaskSet.attempt === 1)
+ assert(newTaskSet.tasks.size === 1)
+ val newTask = newTaskSet.tasks(0)
+ assert(newTask.epoch === newEpoch + 1)
+ runEvent(CompletionEvent(
+ newTask, Success, makeMapStatus("hostB", 1), null,
createFakeTaskInfo(), null))
+
assert(mapOutputTracker.getServerStatuses(shuffleId, 0).map(_._1) ===
Array(makeBlockManagerId("hostB"), makeBlockManagerId("hostA")))
- complete(taskSets(1), Seq((Success, 42), (Success, 43)))
+ complete(taskSets(2), Seq((Success, 42), (Success, 43)))
--- End diff --
this change is only needed if we keep the fix for issues (3) & (4). If we
remove those fixes (and just let it handle itself with one more stage failure),
then we'd keep this test the same as before.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
