You were right, there were more HAProxy processes running in the server, thats why it kept sending connections to the OLD_NAME_FRONTEND. I killed all the other processes and everything is working fine now.
Thanks a lot. De: PiBa-NL [mailto:piba.nl....@gmail.com] Enviado el: sábado, 28 de noviembre de 2015 07:39 Para: Mauricio Cacho Gutiérrez; haproxy@formilux.org Asunto: Re: Configuring Load Balance HAProxy There is no cache to delete.. Can you check there is only 1 active haproxy process running? Depending on how you restart haproxy it could be that old existing connections are still served by the old process that should shutdown once all connections are closed. The old stopping process should not serve new connections if it properly received the shutdown signal from the new process. Op 27-11-2015 om 20:24 schreef Mauricio Cacho Gutiérrez: Hi, Ive configured HAProxy to load balance between servers running Postgresql. At first, I setup 3 servers, one of which was the master and the other two slaves. It worked fine; now I want only the slaves to be available to HAProxy, so I deleted the line with the master configuration in the haproxy.cfg, stopped the service and started it again, but Ive notice in the logs that HAProxy is still sending connections to the master server, even though theres no line in the haproxy.cfg with that IP. So I tried changing the name in the front_end to ensure the changes where taking effect, so I changed the name, stopped the service haproxy and started it again; in the logs it displays Proxy new_name_frontend started Proxy new_name_backend started So nothing about the old frontend name or backend But when I connect via terminal to the server running HAProxy, it displays something like: [ip_client]:[port] [date] OLD_NAME_FRONTEND . And its using the master server Is there some caché or something like that that I should erase so HAProxy has no clue of the master server anymore? PS: Im using the same port. Thanks