sunchao commented on code in PR #58137:
URL: https://github.com/apache/spark/pull/58137#discussion_r3816816181
##########
core/src/test/scala/org/apache/spark/scheduler/CoarseGrainedSchedulerBackendSuite.scala:
##########
@@ -606,6 +606,22 @@ class CoarseGrainedSchedulerBackendSuite extends
SparkFunSuite with LocalSparkCo
assert(mockEndpointRef.decommissionReceived)
}
+ test("SPARK-58886: requestExecutors should saturate instead of overflowing a
huge" +
+ " requested total") {
+ val conf = new SparkConf()
+ .setMaster("local-cluster[0, 3, 1024]")
+ .setAppName("test")
+
+ sc = new SparkContext(conf)
+ val backend =
sc.schedulerBackend.asInstanceOf[CoarseGrainedSchedulerBackend]
+
+ sc.requestTotalExecutors(Int.MaxValue, 0, Map.empty)
+ backend.requestExecutors(1)
+
+ val defaultProf = sc.resourceProfileManager.defaultResourceProfile
+ assert(backend.getRequestedTotalExecutors()(defaultProf) === Int.MaxValue)
Review Comment:
[P3] Non-blocking: could we also start at Int.MaxValue - 1, request 2
executors, and verify both the capped target and an applied request-time
increment of 1? The current assertion would still pass if updateExecRequestTime
accidentally received numAdditionalExecutors again, so it does not protect the
bookkeeping half of this fix.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
