Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/5636#discussion_r35692981
--- Diff: core/src/main/scala/org/apache/spark/scheduler/Stage.scala ---
@@ -76,6 +76,37 @@ private[spark] abstract class Stage(
*/
private var _latestInfo: StageInfo = StageInfo.fromStage(this,
nextAttemptId)
+ /**
+ * Spark is resilient to executors dying by retrying stages on
FetchFailures. Here, we keep track
+ * of unique stage failures (per stage attempt) triggered by fetch
failures to prevent endless
+ * stage retries. Specifically, per stage we wish to only record a
failure when the following
+ * holds:
+ *
+ * A) A fetch failure was observed
+ * B) A failure has not yet been registered for this stage attempt.
There may be multiple
+ * concurrent failures for a sinlge stage since we may have multiple
tasks executing at the same
+ * time, one or many of which may fail. Also, even though there may only
be one non-zombie stage
+ * attemp, zombie stages may still have running tasks.
+ */
+ private val attemptsFailedFromFetch = new HashSet[Int]
+
+ private[scheduler] def clearFailures() : Unit = {
+ attemptsFailedFromFetch.clear()
+ }
+
+ /**
+ * Check whether we should abort the failedStage due to multiple
failures.
+ * This method updates the running set of failures for a particular
stage and returns
--- End diff --
can you change this to " ... running set of failed stage attempts ...",
since its not clear if "failure" refers to task failures or stage attempt
failures
---
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
