Ngone51 commented on a change in pull request #30956:
URL: https://github.com/apache/spark/pull/30956#discussion_r549560267
##########
File path:
core/src/test/scala/org/apache/spark/ExecutorAllocationManagerSuite.scala
##########
@@ -1616,19 +1616,17 @@ class ExecutorAllocationManagerSuite extends
SparkFunSuite {
clock.advance(1000)
manager invokePrivate _updateAndSyncNumExecutorsTarget(clock.nanoTime())
assert(numExecutorsTargetForDefaultProfileId(manager) === 1)
- verify(client, never).killExecutors(any(), any(), any(), any())
+ assert(manager.executorMonitor.executorsPendingToRemove().isEmpty)
// now we cross the idle timeout for executor-1, so we kill it. the
really important
// thing here is that we do *not* ask the executor allocation client to
adjust the target
// number of executors down
- when(client.killExecutors(Seq("executor-1"), false, false, false))
- .thenReturn(Seq("executor-1"))
clock.advance(3000)
schedule(manager)
assert(maxNumExecutorsNeededPerResourceProfile(manager, defaultProfile)
=== 1)
assert(numExecutorsTargetForDefaultProfileId(manager) === 1)
// here's the important verify -- we did kill the executors, but did not
adjust the target count
- verify(client).killExecutors(Seq("executor-1"), false, false, false)
+ assert(manager.executorMonitor.executorsPendingToRemove() ===
Set("executor-1"))
Review comment:
In non-testing mode, `client.killExecutors` will be invoked. But in
testing mode, it's no longer invoked. So we now use the
`executorMonitor.executorsPendingToRemove()` to perform the same check.
----------------------------------------------------------------
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]
