Github user JonathanTaws commented on a diff in the pull request:
https://github.com/apache/spark/pull/15405#discussion_r82673637
--- Diff: core/src/main/scala/org/apache/spark/deploy/master/Master.scala
---
@@ -637,6 +637,16 @@ private[deploy] class Master(
}
freeWorkers = freeWorkers.filter(canLaunchExecutor)
}
+
+ val numExecutorsScheduled = assignedExecutors.sum
+ val numExecutorsLaunched = app.executors.size
+ // Check to see if we managed to launch the requested number of
executors
+ if(numUsable != 0 && numExecutorsLaunched != app.executorLimit &&
+ numExecutorsScheduled != app.executorLimit) {
--- End diff --
`numExecutorsLaunched` corresponds to the actual number of executors that
have been launched so far (literally that have been registered in the executors
list in the ApplicationInfo), whereas `numExecutorsScheduled` corresponds to
the number of executors that have been scheduled/allocated by
`scheduleExecutorsOnWorkers`. This is needed because
`scheduleExecutorsOnWorkers` is called multiple times when setting up the
executors, and if we don't check the condition we will log repeatedly the same
message but with incorrect information (such as "0 executors launched" even
though the executors have been launched previously).
Tell me if that doesn't make sense, I did a lot of trial and error until
coming up with this condition.
---
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]
