Fix error when deleting LB with broken amp

If a load balancer has some broken amps (amphora was unsuccessfully
created, entry exists in the DB but with missing fields), the deletion
of the load balancer fails because octavia is calling delete_port with
None.

Closes-Bug: #2037951
Change-Id: I020ded6e1735ea56da46c5170292ba88ad639103

octavia/network/drivers/neutron/allowed_address_pairs.py

@@ -364,12 +364,13 @@ class AllowedAddressPairsDriver(neutron_base.BaseNeutronDriver):
         """
         try:
             for amphora in vip.load_balancer.amphorae:
-                try:
-                    self.network_proxy.delete_port(amphora.vrrp_port_id)
-                except os_exceptions.ResourceNotFound:
-                    LOG.debug(
-                        'VIP instance port %s already deleted. Skipping.',
-                        amphora.vrrp_port_id)
+                if amphora.vrrp_port_id:
+                    try:
+                        self.network_proxy.delete_port(amphora.vrrp_port_id)
+                    except os_exceptions.ResourceNotFound:
+                        LOG.debug(
+                            'VIP instance port %s already deleted. Skipping.',
+                            amphora.vrrp_port_id)
         except AttributeError as ex:
             LOG.warning(f"Cannot delete port from amphorae. Object does not "
                         f"exist ({ex!r})")

octavia/tests/unit/network/drivers/neutron/test_allowed_address_pairs.py

@@ -122,6 +122,28 @@ class TestAllowedAddressPairsDriver(base.TestCase):
         delete_port.assert_has_calls(calls, any_order=True)
         delete_sec_grp.assert_called_once_with(sec_grp_id)
 
+    def test_deallocate_vip_no_vrrp_port(self):
+        lb = dmh.generate_load_balancer_tree()
+        lb.vip.load_balancer = lb
+        # amphora 0 doesn't have a vrrp_port_id
+        lb.amphorae[0].vrrp_port_id = None
+        vip = lb.vip
+        sec_grp_id = 'lb-sec-grp1'
+        show_port = self.driver.network_proxy.get_port
+        show_port.return_value = Port(
+            device_owner=allowed_address_pairs.OCTAVIA_OWNER)
+        delete_port = self.driver.network_proxy.delete_port
+        delete_sec_grp = self.driver.network_proxy.delete_security_group
+        list_security_groups = self.driver.network_proxy.find_security_group
+        list_security_groups.return_value = SecurityGroup(id=sec_grp_id)
+        self.driver.deallocate_vip(vip)
+        # not called for lb.amphorae[0]
+        calls = [mock.call(vip.port_id),
+                 mock.call(lb.amphorae[1].vrrp_port_id)]
+        delete_port.assert_has_calls(calls, any_order=True)
+        self.assertEqual(2, delete_port.call_count)
+        delete_sec_grp.assert_called_once_with(sec_grp_id)
+
     def test_deallocate_vip_no_port(self):
         lb = dmh.generate_load_balancer_tree()
         lb.vip.load_balancer = lb

releasenotes/notes/fix-error-on-delete-with-broken-amp-10d7f4e85754d7ee.yaml

@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    Fixed a potential issue when deleting a load balancer with an amphora that
+    was not fully created, the deletion may have failed when deallocating the
+    VIP port, leaving the load balancer in ERROR state.