Github user kayousterhout commented on a diff in the pull request:
https://github.com/apache/spark/pull/13603#discussion_r67070153
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala ---
@@ -280,10 +280,54 @@ private[spark] class TaskSchedulerImpl(
}
}
}
+ if (!launchedTask && isTaskSetCompletelyBlacklisted(taskSet)) {
+ taskSet.abort(s"Aborting ${taskSet.taskSet} because it has a task
which cannot be scheduled" +
+ s" on any executor due to blacklists.")
+ }
return launchedTask
}
/**
+ * Check whether the given task set has been blacklisted to the point
that it can't run anywhere.
+ *
+ * It is possible that this taskset has become impossible to schedule
*anywhere* due to the
+ * blacklist. The most common scenario would be if there are fewer
executors than
+ * spark.task.maxFailures. We need to detect this so we can fail the
task set, otherwise the job
+ * will hang.
+ *
+ * The check here is a balance between being sure to catch the issue,
but not wasting
+ * too much time inside the scheduling loop. Just check if the last
task is schedulable
+ * on any of the available executors. So this is O(numExecutors)
worst-case, but it'll
+ * really be fast unless you've got a bunch of things blacklisted. Its
possible it won't detect
+ * the unschedulable task immediately, but if it returns false, there is
at least *some* task
+ * that is schedulable, and after scheduling all of those, we'll
eventually find the unschedulable
+ * task.
+ */
+ private[scheduler] def isTaskSetCompletelyBlacklisted(
--- End diff --
Ah got it -- agree that in the long term, given the new blacklistling, it
might make sense here! But let's put it in the TaskSetManager for now.
---
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]
