We should cleanup ipv4 address if keepalived is dead

[Yang Li]

If a router's keepalived is dead(kill -9 <pid>, or kill -HUP <pid> too many times that cause process terminated), and the original role in this node is master, this will cause brain split. Then when we restart neutron-l3-agent, the ipv6 will be cleanup, but the ipv4 still exists, I think we should alse cleanup ipv4 before enable keepalived.
The current state:
original master:
[root@node-1 ~]# ip netns exec qrouter-a8f9d4e4-622d-4ebf-b7a9-4818f63ef502 ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
28: ha-66734f93-5e: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:fd:49:a1 brd ff:ff:ff:ff:ff:ff
    inet 169.254.192.1/18 brd 169.254.255.255 scope global ha-66734f93-5e
       valid_lft forever preferred_lft forever
    inet 169.254.0.1/24 scope global ha-66734f93-5e
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fefd:49a1/64 scope link
       valid_lft forever preferred_lft forever
29: qr-4f77a86c-c2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:26:98:8f brd ff:ff:ff:ff:ff:ff
    inet 192.168.100.1/24 scope global qr-4f77a86c-c2
       valid_lft forever preferred_lft forever
46: qg-e28d50e4-ba: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:c4:3e:43 brd ff:ff:ff:ff:ff:ff
    inet 172.16.10.133/24 scope global qg-e28d50e4-ba
       valid_lft forever preferred_lft forever
    inet 172.16.10.134/32 scope global qg-e28d50e4-ba
       valid_lft forever preferred_lft forever

currnet master:
[root@node-2 ~]# ip netns exec qrouter-a8f9d4e4-622d-4ebf-b7a9-4818f63ef502 ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
32: ha-606f2f23-5f: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:8e:dc:c4 brd ff:ff:ff:ff:ff:ff
    inet 169.254.192.2/18 brd 169.254.255.255 scope global ha-606f2f23-5f
       valid_lft forever preferred_lft forever
    inet 169.254.0.1/24 scope global ha-606f2f23-5f
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fe8e:dcc4/64 scope link
       valid_lft forever preferred_lft forever
33: qr-4f77a86c-c2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:26:98:8f brd ff:ff:ff:ff:ff:ff
    inet 192.168.100.1/24 scope global qr-4f77a86c-c2
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fe26:988f/64 scope link nodad
       valid_lft forever preferred_lft forever
50: qg-e28d50e4-ba: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN qlen 1000
    link/ether fa:16:3e:c4:3e:43 brd ff:ff:ff:ff:ff:ff
    inet 172.16.10.133/24 scope global qg-e28d50e4-ba
       valid_lft forever preferred_lft forever
    inet 172.16.10.134/32 scope global qg-e28d50e4-ba
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fec4:3e43/64 scope link nodad
       valid_lft forever preferred_lft forever

The command line output showed that the original master's ipv6 were cleanup.

[YAMAMOTO Takashi (yamamoto)]

while i'm not sure if the suggested cleanup on a restart is good enough, it seems something needs to be done.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/667071

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Rodolfo Alonso <email address hidden>" on branch: master
Review: https://review.opendev.org/c/openstack/neutron/+/667071

[Rodolfo Alonso (rodolfo-alonso-hernandez)]

Bug closed due to lack of activity, please feel free to reopen if needed.