Comment 6 for bug 1826644

Thx for checking that.
So if 2 routers in same tenant have got different vr_id configured than it's definitely not the same bug as something what https://review.openstack.org/651495 is trying to fix.

Can You then check in logs of L3 agents which handles those routers if there are any errors there? Also checking logs and configs of keepalived and neutron-keepalived-state-change may help to understand the issue here.

Please also check in qrouter namespaces on network nodes if all other IPs from HA network are reachable from every namespace.