Re: [Netdisco] discovery subnet instead of individual IP

[Simon H. <li...@th...>]

Victor Zhong <vz...@wa...> wrote:

> We just start using NetDisco v2 lately. Still got quite a few fancy stuff to "disco", perhaps 'discovering devices using subnet instead of individual IP" is one of them.
> 
> just using "~netdisco/bin/netdisco-do discover -d x.x.x.x " would be a nightmare for us to discover the whole "ocean".

I think the point is that for the sort of network it's designed to handle, you only need to discover one device - then NetDisco will automatically find out about all devices that are connected to it (neighbours). When it knows about those close neighbours, it will find their neighbours, and so on until it knows about every device.

This will only break down if you have devices/links that don't support neighbour discovery. E.g. in my case I have a VPN tunnel between my network here and one I manage on a customer site - and because I've not installed/configured everything needed for the endpoints (GNU/Linux) to report the topology via SNMP it means I've had to manually tell NetDisco about the link and "seed" the far end by discovering one device there. Similarly, I've had to manually configure around a couple of switches/firewalls that are either unmanaged or don't report the required information via SNMP.

For that remote network, it's a nice "clean" Cisco-only network with SNMP configured and allowed from my monitoring node. So just discovering one device populated the entire network automagically.

So in a "well engineered" network, the worst case is having to manually discover one device per site - and manually add the topology links (typically WAN links/tunnels) to connect sites.