When starvation heppens:
- a lot of unschedulable pods exists in the head of queue
- because condition.LastTransitionTime is updated only when condition.Status changed
- (this means that once a pod is marked unschedulable, the field never updated until the pod successfuly scheduled.)
What was changed:
- condition.LastProbeTime is updated everytime when pod is determined
unschedulable.
- changed sort function so to use LastProbeTime to avoid starvation
described above
Consideration:
- This changes increases k8s API server load because it updates Pod.status whenever scheduler decides it as
unschedulable.
Signed-off-by: Shingo Omura <everpeace@gmail.com>
With the alpha scheduling queue we move pods from unschedulable to
active on certain events without a backoff. As a result we can cause
starvation issues if high priority pods are in the unschedulable queue.
Implement a backoff mechanism for pods being moved to active.
Closes#56721
The Heap data structure is useful for our backoff system in addition to
scheduling queue. Move it to somewhere it can be consumed by both
systems and properly export needed names. Also adding unit tests
from client-go/tools/cache/heap.go.
- don't update nominatedMap cache when Pop() an element from activeQ
- instead, delete the nominated info from cache when it's "assumed"
- unit test behavior adjusted
- expose SchedulingQueue in factory.Config
