wypoon commented on a change in pull request #28848:
URL: https://github.com/apache/spark/pull/28848#discussion_r452593445
##########
File path: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
##########
@@ -1933,29 +1951,45 @@ private[spark] class DAGScheduler(
maybeEpoch = None)
}
+ /**
+ * Handles removing an executor from the BlockManagerMaster as well as
unregistering shuffle
+ * outputs for the executor or optionally its host.
+ *
+ * @param execId executor to be removed
+ * @param fileLost If true, indicates that we assume we've lost all shuffle
blocks associated
+ * with the executor; this happens if the executor serves its own blocks
(i.e., we're not
+ * using external shuffle), the Standalone worker (which serves the
shuffle data) is lost,
Review comment:
I'm not sure I understand the point you're making here. The 3 conditions
when `fileLost=true` are: (1) we're not using the external shuffle service
(`!env.blockManager.externalShuffleServiceEnabled`); or (2) the Standalone
worker is lost (as I explained above); or (3) a FetchFailed occurred.
In the case of case of YARN + external shuffle service, `fileLost=true`
happens only with a FetchFailed.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
