This is a clone of issue OCPBUGS-33048. The following is the description of the original issue:
—
Description of problem:

When you delete a cluster, or just a BMH, before the installation starts (Assisted Service takes the control), the metal3-operator tries to generate a PreprovisioningImage.

In previous versions, it was created a fix that, during some first installation phases the creation of the PreprovisioningImage was not invoked:

https://github.com/openshift/baremetal-operator/pull/262/files#diff-a69d9029388ab766ed36b32180145f52785a9d4a153775510dbddfa928a72e1cR787

it was based on the status "StateDeleting".

Recently, it was added a new status "StatePoweringOffBeforeDelete":

https://github.com/openshift/baremetal-operator/commit/6f65d8e75ef6ed921863ebaf793cccda61de8bcb#diff-eeed3703d04e4c23a7d7af8cd0b7931b6b7990f23d826c49bdbc31c5f0a50291

but this status is not covered on the previous fix. And during this new phase there should not be tried to create the image.

The problem of trying create the PreprovisioningImage, when it should not, it is that create problems on ZTP. Where the BMH and all the objects are deleted at the same time. And the operator cannot create the image because the NS is been deleted.

Version-Release number of selected component (if applicable):

4.14    

How reproducible:

Steps to Reproduce:

    1.Create a cluster
    2.Wait until the provisioing phase
    3.Delete the cluster
    4.The metal3-operator tries to create the PreprovisioningImage wrongly. 

Actual results:

Expected results:

Additional info: