[library] Deploy Haproxy config with Neutron API in Active/Active HA mode

[Oleksii Aleksieiev]

Description of the environment
HA 3 controllers
Neutron+VLAN

Steps to reproduce
Create a new cluster
Select HA mode, “Neutron with VLANs” for network, and choose to install Savanna
Leave defaults the rest of the settings. Run cluster installation

Expected result
085-neutron.cfg should not have backup option for servers.

/etc/haproxy/conf.d/085-neutron.cfg
listen quantum
  bind 10.20.0.10:9696
  balance roundrobin
  option httplog
  server server001 10.20.0.11:9696 check
  server server002 10.20.0.12:9696 check
  server server003 10.20.0.13:9696 check

Actual result

/etc/haproxy/conf.d/085-neutron.cfg
listen quantum
  bind 10.20.0.10:9696
  balance roundrobin
  option httplog
  server server001 10.20.0.11:9696 check backup
  server server002 10.20.0.12:9696 check
  server server003 10.20.0.13:9696 check backup

Workaround
In /etc/puppet/openstack/manifests/ha/neutron.pp
set "define_backups " to false

Impact
Only one neutron server is used at the same time. haproxy will switch connections to another server only if the server is dead. This cause one neutron server to have all the load while all other nodes are idle.
Seems like this is old configuration.

[Bogdan Dobrelya (bogdando)]

I believe this one should be high priority and retargeted for 5.1

[Aleksandr Shaposhnikov (alashai8)]

Behavior checked. To load balance requests to the Neutron-API all of them should be active. My personal experience with such config was pretty fine. I also tested 5.0 release iso with multiple active neutron-api servers.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/113364

[Vladimir Kuklin (vkuklin)]

I think that this bug is not of high priority. We should triage it and find if it is absolutely necessary to turn on active/active for neutron-server. Let's consult with neutron guys.

[Andrew Woodward (xarses)]

duplicate of https://bugs.launchpad.net/fuel/+bug/1276762