neutron-lbaas endless haproxy restarts due to faulty Neutron port fixed_ip field

[Alvaro Uria]

Xenial-Ocata

A LBaaS has a port binding that did not complete (fixed_ip is empty). This causes: https://pastebin.ubuntu.com/26315229/ , making the LB Agent restart every minute (recreating all the LB running on that agent).

"neutron port-show" shows:

{
  "allowed_address_pairs": [],
  "extra_dhcp_opts": [],
  "updated_at": "2018-01-03T18:14:31Z",
  "device_owner": "neutron:LOADBALANCERV2",
  "revision_number": 58905,
  "binding:profile": {},
  "port_security_enabled": true,
  "fixed_ips": [],
  "id": "f28b4089-86f1-408e-9259-778171e3d21a",
  "security_groups": [
    "49316f86-4d84-4ff6-8485-aa0b6b925a21"
  ],
  "binding:vif_details": {
    "port_filter": true,
    "ovs_hybrid_plug": true
  },
  "binding:vif_type": "ovs",
  "mac_address": "fa:16:3e:21:95:4b",
  "project_id": "077055e5a7b74c29a7a7f77769b48a30",
  "status": "BUILD",
  "binding:host_id": "REDACTED",
  "description": null,
  "tags": [],
  "dns_assignment": [],
  "device_id": "08434380-226c-4ccd-be12-34cb6ba5efab",
  "name": "loadbalancer-08434380-226c-4ccd-be12-34cb6ba5efab",
  "admin_state_up": true,
  "network_id": "cce08c18-d734-4d42-a0be-b630fdfde87a",
  "dns_name": "",
  "created_at": "2017-12-05T09:54:40Z",
  "binding:vnic_type": "normal",
  "tenant_id": "077055e5a7b74c29a7a7f77769b48a30"
}

By updating the lbaas with "--admin-state-up False", LB Agent restarts stopped happening. I think this should be the default to avoid affecting other loadbalancers.

[Gregory Thiemonge (gthiemonge)]

Abandoned after re-enabling the Octavia launchpad.