Hi,

> I am working on a load balancing system where the upstream providers need
> to change at runtime. The available peer group will be routed either by a
> hash, or by round robin, but the size of the peer group might change
> between any request.

Do you really need this list to be dynamic? Maybe static list of backends
that are dynamically marked as either "up" or "down" is good enough? If so,
please take a look at ngx_supervisord [1] (you will probably want to use it
with "supervisord none;" setting). You can also use it together with
health-check module to automatically mark failed backends as "down".

[1] http://labs.frickle.com/nginx_ngx_supervisord/
[2] http://github.com/cep21/healthcheck_nginx_upstreams

Best regards,
Piotr Sikora < piotr.sikora@frickle.com >



_______________________________________________
nginx mailing list
nginx@nginx.org
http://nginx.org/mailman/listinfo/nginx