Description of problem: If keepalived is stopped uncleanly, it is possible for a VIP to be left on a node that shouldn't have it. Even after keepalived is restarted, the improper VIP is not cleaned up.

This was originally reported in https://bugzilla.redhat.com/show_bug.cgi?id=1931505 but as we are working around it in OpenShift I wanted a separate report open against keepalived itself.


Version-Release number of selected component (if applicable): 2.0.10


How reproducible: Always


Steps to Reproduce:
1. Run keepalived with a configuration such as:

vrrp_instance ostest_API {
    state BACKUP
    interface eth0
    virtual_router_id 14
    priority 70
    advert_int 1
    nopreempt
    
    unicast_src_ip 12.1.1.122
    unicast_peer {
        12.1.1.111
    }
    
    authentication {
        auth_type PASS
        auth_pass ostest_api_vip
    }
    virtual_ipaddress {
        12.2.2.2/32
    }
}

2. On the node holding the VIP, run "killall -9 keepalived".
3. Restart keepalived on the node where it was killed.

Actual results: The VIP fails over to another node, but also remains on the one where keepalived was killed.


Expected results: The VIP should be removed from the node where keepalived was killed at some point in the process, either immediately after it is killed or when it restarts.


Additional info: This behavior does not reproduce on Fedora 33 with keepalived 2.1.5. It correctly unconfigures the VIP when keepalived is killed. This is triggered in our product by a failing liveness probe removing the keepalived container, but as noted above I have also reproduced it outside of containers in a VM environment.