i think this bug is unfortunately not to 100 % percent solved, the healthmonitor, not closed the connection to the errored member
So in Detail:
The ERROR operating status of the members did not cause the LB to no longer distribute requests to the ERROR'ed members -- but in a installation released with ovn-octavia-provider we observe in a example, 50%
of the requests fail due to 2/4 ERROR'ed members.
Release of this environment is Zed.
i think this bug is unfortunately not to 100 % percent solved, the healthmonitor, not closed the connection to the errored member
So in Detail:
The ERROR operating status of the members did not cause the LB to no longer distribute requests to the ERROR'ed members -- but in a installation released with ovn-octavia- provider we observe in a example, 50%
of the requests fail due to 2/4 ERROR'ed members.
Release of this environment is Zed.
we tracked this bug in our project here:
https:/ /github. com/SovereignCl oudStack/ issues/ issues/ 268