After migrating from liberty to mikata using:
https://docs.openstack.org/openstack-ansible/mitaka/upgrade-guide/manual-upgrade.html
all previous loadbalancers (using lbaasv2) do not start. No errors in neutron-lbaasv2-agent log and everything looks normal in cli and horizon. New loadbalancers start and work normal.
It appears that the hostname translation done in:
https://github.com/openstack/openstack-ansible/blob/13.3.18/scripts/upgrade-utilities/playbooks/rfc1034_1035-cleanup.yml
does not apply to lbaasv2 ports. So currently we have ports:
root@control1-utility-container-cd796ab4:~# neutron port-list -c device_owner -c binding:host_id | grep LOADBA
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network2-neutron-agents-container-4c4e8121 |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network2_neutron_agents_container-4c4e8121 |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
| neutron:LOADBALANCERV2 | network1_neutron_agents_container-1728f97f |
<snip>
where you can see a single entry has underscores replaced:
| neutron:LOADBALANCERV2 | network2-neutron-agents-container-4c4e8121 |
and that's the one created after migration and one that works.
I'm happy to look at a patch to rfc playbook if you can confirm that's correct and provide some guidance on best way to fix (manual sql change, or use neutron command)?
Ran:
update ml2_port_bindings set host=replace(host, "_","-"); binding_ levels set host=replace(host, "_","-");
update ml2_port_
on neutron db to fixup hosts, but haproxy is still not launching on old ones. Did a comparison of output of loadbalancer, pool, members, port, subnet of a working and non working and only difference that stands out is working now has created_at and updated_at fields on the port (new in mikata).
Still digging.