PXE booting of servers fails in Spine/leaf environment after deleting the stack in Red Hat OpenStack Platform 13

Issue

In a spine/leaf environment, the undercloud's /var/lib/neutron/dhcp/<UUID>/leases contains old / stale entries that lead to a failure when ironic runs its automatic node cleanup. When the node PXE boots, it never gets a DHCP lease from neutron's dnsmasq. This happens when running heat stack-delete <stack name>.

A restart of all neutron* services fixes this issue and cleanup works.

sudo systemctl restart neutron*

Then, this works:

openstack baremetal node abort <node>
openstack baremetal node manage <node>
openstack baremetal node provide <node>