You will need to run Nagios Core 4.0 or a newer server and have at least two hosts, one of which is only reachable via the other. The host that allows communication with the other is the parent host. You should be reasonably confident that a loss of connectivity to the parent host necessarily implies that the child host will become unreachable from the monitoring server.

Access to the web interface of Nagios Core would also be useful as making this change will change the appearance of the network map; this is discussed in the "Using the network map" recipe in this chapter.

Our example will use a Nagios Core monitoring server, olympus.example.net, which will monitor the following three hosts:

The hosts are connected as in the following diagram:

Note that the Nagios Core server olympus.example.net is only able to communicate with the corsica.example.net webserver if the calpe.example.net router works correctly. If calpe.example.net were to enter a DOWN state, we would see corsica.example.net enter a DOWN state too. Take a look at the following screenshot:

This is a little misleading as we don't actually know whether corsica.example.net is down. It might be, but with the router between the hosts not working correctly, Nagios Core has no way of knowing. A more informative and accurate status for the host would be UNREACHABLE; this is what the configuration we're about to add will arrange.

We can configure a parent-child relationship for our two hosts like so:

With this done, if the parent host enters a DOWN state and the child host can't be contacted, then the child host will enter an UNREACHABLE state rather than also being flagged as DOWN:

The child host's contacts will also receive UNREACHABLE notifications instead of DOWN notifications for the child host, provided the u flag is included in notification_options for the host and host_notification_options for the contacts. Refer to the Specifying which states to be notified about recipe in Chapter 4, Configuring Notifications, for details on this.

This is a simple application of Nagios Core's reachability logic. When the check to calpe.example.net fails for the first time, Nagios Core notes that it is a parent host for one child host: corsica.example.net. If during checks for the child host it finds that it cannot communicate with it, then instead of the DOWN state, it flags an UNREACHABLE state, firing a different notification event.

The primary advantages to this are two fold:

With a little planning and some knowledge of the network, all we need to do is add a few parents directives to host definitions to build a simple network structure and Nagios Core behaves much more intelligently as a result. This is one of the easiest ways to refine the notification behavior of Nagios Core; it can't be recommended enough!

Note that a child host can itself be a parent to other hosts in turn, allowing a nesting network structure. Perhaps in another situation, we might find the corsica.example.net server is two routers away from the monitoring server, as shown in the following diagram:

In this case, not only is corsica.example.net the child host of calpe.example.net, but calpe.example.net is itself the child host of janus.example.net. We could specify this relationship in exactly the same way. Take a look at the following code:

define host {
    use        linux-router
    host_name  calpe.example.net
    alias      calpe
    address    192.0.2.129
    parents    janus.example.net
}

It's also possible to set multiple parents for a host if there are two possible paths to the same machine, as follows:

define host {
    use        linux-server
    host_name  corsica.example.net
    alias      corsica
    address    192.0.2.71
    parents    calpe.example.net,janus.example.net
}

With this configuration, corsica.example.net would only be deemed UNREACHABLE if both of its parent hosts were down. This kind of configuration is useful to account for redundant paths in a network—for example, spanning tree technologies or BGP/OSPF failover.

After you set up a good basic structure for your network using the parents directive, definitely check out the Using the network map recipe in this chapter to get some automatic visual feedback about your network's structure as generated through your new configuration.