Jackeyzhe commented on code in PR #28535:
URL: https://github.com/apache/flink/pull/28535#discussion_r3537746634
##########
flink-runtime/src/test/java/org/apache/flink/runtime/jobmaster/slotpool/DeclarativeSlotPoolBridgeTest.java:
##########
@@ -224,8 +231,15 @@ void
testNoConcurrentModificationWhenSuspendingAndReleasingSlot() throws Excepti
@TestTemplate
void testAcceptingOfferedSlotsWithoutResourceManagerConnected() throws
Exception {
+ final ManuallyTriggeredScheduledExecutorService scheduledExecutor =
Review Comment:
I checked the production path around DeclarativeSlotPoolBridge. I do not
think this race can occur in the main code in the same way.
The failing window requires pendingRequests to be mutated while the close
path is copying/failing it. In production, slot-pool operations and scheduled
callbacks are expected to be serialized through ComponentMainThreadExecutor.
The copy-then-clear in cancelPendingRequests() then protects the later
fail-request loop from reentrant modifications after the snapshot is created.
So this looks like a test-only flakiness caused by relying on real scheduled
timing while closing the bridge from the test thread. Using
ManuallyTriggeredScheduledExecutorService here looks reasonable to me.
--
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]
