To Reproduce Steps to reproduce the behavior:
After https://review.opendev.org/c/openstack/tripleo-common/+/717798 was merged we fail early with node deletion: if Ironic is unable to return BM node data for requested instance UUID, then we do not proceed with deletion.

However, there are situations when we need to delete semi-deployed overcloud nodes (Heat and Nova resource were created and in FAILED/Error state, but Ironic server wasn't reserved). In this cases we can't delete overcloud nodes after subset of nodes failed provisioning during scale-out.

We currently have VIP customer affected by this, so I am kindly asking for some workaround. Bug itself could be processed later.

Expected behavior
If records for overcloud node exist in Nova and Heat, then it should be possible to use openstack overcloud command to delete it

Bug impact
Cleanup after semi-failed scale-out is blocked

Known workaround
If scale-out failed completely, we can repeat deployment command with original node count. But in this case 3/20 nodes failed provisioning