holdenk commented on a change in pull request #27636:
[SPARK-30873][CORE][YARN]Handling Node Decommissioning for Yarn cluster manger
in Spark
URL: https://github.com/apache/spark/pull/27636#discussion_r385902131
##########
File path: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
##########
@@ -1888,6 +1922,18 @@ private[spark] class DAGScheduler(
clearCacheLocs()
}
+ /**
+ * Remove shuffle data mapping when node is decomissioned.
+ *
+ * @param host host of the node that is decommissioned
+ */
+ private[scheduler] def handleNodeDecommissioned(host: String) {
Review comment:
So if we remove the shuffle files on decommissioning then we're essentially
forcing a recompute even if the files are still present and the node hasn't
gone away which doesn't seem ideal.
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
