If there are not enough resources when promoting, the scheduler can not work

[spiritNO1]

When promoting, the scheduler should calculate the member cluster, this requires sufficient resources to be available when calculating scheduling. Since the member cluster already has the corresponding workload, the scheduler should skip the calculation process and write the result directly to rb.

[chaosi-zju]

Judging from the scheduling logic, there seems to be a problem. May I ask whether you measured it in the actual test or inferred it?

[spiritNO1]

Judging from the scheduling logic, there seems to be a problem. May I ask whether you measured it in the actual test or inferred it?

We referred to this part of the logic to automatically create pp of member cluster workloads, but found that some workloads could not create work because there was no scheduling result.

We did not use the promote command directly, but the logic is the same.
We observed the following error in log.

I think if promoting, we should directly patch the schedule result to rb and not need to caculate.

[chaosi-zju]

I think if promoting, we should directly patch the schedule result to rb and not need to caculate.

How does the scheduler identify whether this is promote or normal scheduling

[spiritNO1]

I think if promoting, we should directly patch the schedule result to rb and not need to caculate.

How does the scheduler identify whether this is promote or normal scheduling

Maybe use an annotation or a label, delete it after scheduling