wilfred-s commented on code in PR #745:
URL: https://github.com/apache/yunikorn-core/pull/745#discussion_r1418042106
##########
pkg/scheduler/objects/application.go:
##########
@@ -936,6 +938,8 @@ func (sa *Application) tryAllocate(headRoom
*resources.Resource, allowPreemption
userHeadroom := ugm.GetUserManager().Headroom(sa.queuePath,
sa.ApplicationID, sa.user)
// get all the requests from the app sorted in order
for _, request := range sa.sortedRequests {
+ request.SetSchedulingAttempted(true)
+
Review Comment:
Can we postpone setting this to a later point in the cycle?
At the point that we now check we have not done some fits and we do not want
to trigger scale up if it does not fit in the users quota as it will not be
able to run.
A replacement during gang scheduling should also not trigger scale up.
##########
pkg/scheduler/objects/application.go:
##########
@@ -898,13 +898,15 @@ func (sa *Application) getOutstandingRequests(headRoom
*resources.Resource, user
return
}
for _, request := range sa.sortedRequests {
- if request.GetPendingAskRepeat() == 0 {
+ if request.GetPendingAskRepeat() == 0 ||
!request.IsSchedulingAttempted() {
continue
}
// ignore nil checks resource function calls are nil safe
if headRoom.FitInMaxUndef(request.GetAllocatedResource()) &&
userHeadRoom.FitInMaxUndef(request.GetAllocatedResource()) {
- // if headroom is still enough for the resources
- *total = append(*total, request)
+ if !request.HasTriggeredScaleUp() {
Review Comment:
Should we handle daemon set pods here too?
Daemonset pods will not be helped by autoscaling. The Daemonset controller
only creates a pod for an existing node.
--
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]
