I deploy clusters of machines to the cloud (Amazon EC2), and would like to use zabbix for monitoring these clusters. In a cloud environment, adding/removing machine instances is pretty dynamic and happens fairly frequently, sometimes at large scales (tens to hundreds of instances at a time). As a result, auto-discovery becomes a lot more important. The problem is, as far as I know, the EC2 IP range is very large (10...*), which makes the IP scan not very effective.

What I would like to see:
1) When agentd starts up, it tries to contact the server
2) When server sees this connection from an unknown machine, it triggers a Discovery event.
3) One can write an action for this event to add the instance to their configuration in the same way as ping based auto discovery.

One other nice to have would be some way for the new instance to communicate the roles it is in to the server as part of this discovery event. This would make it easy to pick templates for the instances based on what roles it was configured for.

This works well because typically the agent has to be configured with the server's address, and so always knows how to get to it, but the reverse is not true before the host is added to the server's configuration.