You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For RWO backup pods allow to set the priorityClassName to allow other pods on the node to be preempted in case the node reached its resource capacity.
Context
With RWO backups pods, a node selector is set on the pod to make sure that the Backup pod is scheduled on the same node as the pod with the mounted volume. This is working fine. However if the resources requested by the backup pod are higher than the remaining resources on the node, the pod is not scheduled on the specified node. Other pods on the node are not preempted, because the backup pod has the default priority set (priority 0). The backup pod would stay in pending state.
Allowing to set a priorityClassName for the backup pod with a PriorityClass higher than the default priority would cause the scheduler to preempt pods on the node and schedule the backup pod instead.
Summary
For RWO backup pods allow to set the
priorityClassName
to allow other pods on the node to be preempted in case the node reached its resource capacity.Context
With RWO backups pods, a node selector is set on the pod to make sure that the Backup pod is scheduled on the same node as the pod with the mounted volume. This is working fine. However if the resources requested by the backup pod are higher than the remaining resources on the node, the pod is not scheduled on the specified node. Other pods on the node are not preempted, because the backup pod has the default priority set (priority 0). The backup pod would stay in
pending
state.Allowing to set a
priorityClassName
for the backup pod with a PriorityClass higher than the default priority would cause the scheduler to preempt pods on the node and schedule the backup pod instead.Out of Scope
No response
Further links
Acceptance Criteria
No response
Implementation Ideas
priorityClassName
in a schedule and on a backup objectpriorityClassName
is set, therefore the pod is scheduled with a default priority of 0The text was updated successfully, but these errors were encountered: