attilapiros commented on a change in pull request #32114:
URL: https://github.com/apache/spark/pull/32114#discussion_r639185928
##########
File path:
core/src/main/scala/org/apache/spark/storage/BlockManagerMasterEndpoint.scala
##########
@@ -273,12 +282,12 @@ class BlockManagerMasterEndpoint(
}
}
bmIdsExecutor.foreach { bmId =>
- blockManagerInfo.get(bmId).foreach { bmInfo =>
+ blockManagerInfo.get(bmId).filter(_.isAlive).foreach { bmInfo =>
Review comment:
@Ngone51, @mridulm So the feature was introduced by
https://github.com/apache/spark/pull/24499.
It can be enabled by setting `spark.shuffle.service.fetch.rdd.enable` to
true then instead of the executor's block manager ID an artificial ID will be
used by where the port part is coming from the external shuffle service.
About the leak:
I cannot see leak in `blockStatusByShuffleService` regarding this method. As
the clean is done a few lines above:
https://github.com/apache/spark/blob/c0a4102acbeee20abd69e1b9c98f4bc308b9caea/core/src/main/scala/org/apache/spark/storage/BlockManagerMasterEndpoint.scala#L279-L281
And there is test for it:
https://github.com/apache/spark/blob/c0a4102acbeee20abd69e1b9c98f4bc308b9caea/core/src/test/scala/org/apache/spark/storage/BlockManagerInfoSuite.scala#L129-L144
--
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]
