Neutron port DOWN after restarting neutron services

[Georgy Dyuldin]

Detailed bug description:
 Instance port reach DOWN status after restarting all neutron services on all controllers.

Steps to reproduce:
 1. Boot "vm1" in exists internal net
 2. Vefiry ping public IP from "vm1"
 3. Restart all neutron services on all nodes:
    * get names with `crm_mon -1`
    * restart with with pcs resource restart <name>
    * get names with `initctl list | grep running | grep neutron | awk '{ print $1 }'`
    * restart with restart <name>
 4. Boot "vm2" in same net as "vm1"
 5. Verify pings betwheen vms and to public ip
 6. Create new network "net01" and router between "net01" and external net
 7. Boot "vm3" on "net01"
 8. Vefiry ping public IP from "vm3"
 9. Delete all created servers
 10. Delete created router and network

Expected results:
 All steps passed

Actual result:
 Can't connect to vm1 after step 4. Corresponding port is in DOWN status

Reproducibility:
 Always

Description of the environment:
- Operation system: Ubuntu
- Versions of components: MOS 9.1 snapshot #76
- Network model: VxLAN + L2pop + L3HA

[Denis Meltsaykin (dmeltsaykin)]

Neutron team please take a look and change importance if it's needed.

[Oleg Bondarev (obondarev)]

Please include particular uuids into bug descriptions, like port id in this case. Digging in snapshot trying to find out what was the port with issue is a waste of time.

[Oleg Bondarev (obondarev)]

No errors in logs, need port_id or live repro for further debugging.

[Alexander Koryagin (akoryagin)]

Can not reproduce this bug with MOS 9.1 snapshot #185.
Following test has passed several times:
    mos_tests/neutron/python_tests/
    test_restarts.py::TestRestarts::test_restart_all_neutron_services[(1295480)]

- Network model: VxLAN + L2pop + L3HA