I'm having same issue with CentOS 8 stream / Wallaby / kolla-ansible.
I have tried following, added ovsdb_timeout=30 in ovn_meta config file. but still i am seeing issue time to time. I have also setup cronjob to hide this issue but look like this is real issue. why don't we keep haproxy running all the time and not spawning when vm get created? I believe openstack-ansible doing that approach where dedicated haproxy namespace running for this task.
I'm having same issue with CentOS 8 stream / Wallaby / kolla-ansible.
I have tried following, added ovsdb_timeout=30 in ovn_meta config file. but still i am seeing issue time to time. I have also setup cronjob to hide this issue but look like this is real issue. why don't we keep haproxy running all the time and not spawning when vm get created? I believe openstack-ansible doing that approach where dedicated haproxy namespace running for this task.
# cat /etc/neutron/ neutron_ ovn_metadata_ agent.ini
[ovs]
ovsdb_connection = tcp:127.0.0.1:6640
ovsdb_timeout = 30