craigcondit commented on code in PR #458:
URL: https://github.com/apache/yunikorn-core/pull/458#discussion_r1035280632
##########
pkg/scheduler/objects/application.go:
##########
@@ -1742,3 +1754,10 @@ func (sa *Application)
SetTimedOutPlaceholder(taskGroupName string, timedOut int
sa.placeholderData[taskGroupName].TimedOut = timedOut
}
}
+
+// test only
+func (sa *Application) SetPreemptionAttemptInterval(interval time.Duration) {
Review Comment:
Again, I think we need to reconsider this. It's the RM's responsibility to
kill tasks when we ask it to, and we need to trust that it will occur. However,
expecting that this happens in some predetermined time is asking for trouble.
By preempting once (only) and reserving the node, we are reserving space until
the requisite tasks are terminated.
The only thing we might want to do is keep track of whether or not all the
tasks have been preempted yet for a node. If they have, and we are still
unschedulable, that likely indicates that another scheduler got in line ahead
of us and we need to try preempting again.
--
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]
