We need to manage the service, otherwise each time Puppet will change
the configuration of HAproxy, the service won't restart.
- Set to false only: Service will never start even by Keepalived
- Set to false with tricking init script: Service will start once, only
by Keealived but never restart after each config change
- Set to true: Service will be managed on all nodes, no ERROR, only
warnings on BACKUP nodes, regarding binding.
For now, we switch to true and we plan to patch puppetlabs-haproxy to
improve it.