You can't do much with nVentory if it doesn't have data about your nodes (computers, network gear, storage devices, console and KVM servers, etc.)

Computers

The easiest way to register your computer nodes with nVentory is to install the client package on them and run either the Ruby or Perl client in registration mode (nv --register). That will gather up a bunch of info about the computer and register it with nVentory. In fact, if you build client packages using the Makefile and packaging config files included in the distribution then when you install the package on a machine it will insert a cron job to regularly run the client in registration mode.

An alternative route, if perhaps you don't control some or any of the computers you need to inventory and thus can't install the client, would be to scan them via SNMP (presuming an SNMP daemon has been installed and configured). The distribution doesn't yet include an example script for this.

CPU Utilization Graphs in the Web UI

Very basic CPU utilization information is gathered from each computer when it runs the nVentory client. The client depends on sar data for this information, so first you need to ensure that sar is installed and running properly on each computer. The client packages have appropriate dependencies to ensure sar is installed. When the nVentory client registers it will send this averaged CPU information to the nVentory server. This individual computer data can be aggregated to calculate average CPU utilization for each node group (daily) as well as globally. The global CPU utilization will appear on the homepage dashboard as a graph, node group averages appear on the node group pages. In order to populate these CPU utilization graphs, you'll need to download the following shell script, edit the PATH variable and set it in cron to run once a day at the time of your choosing:

https://nventory.svn.sourceforge.net/svnroot/nventory/trunk/server/script/process_metrics.sh

Note that this is not intended as a replacement for performance monitoring software like cacti. The client reports data every three hours by default, and that is averaged to daily figures. The purpose is to allow you to find areas of your environment that are over or under-provisioned, and to track your progress in improving the overall utilization of your environment. Projects like virtualization, server consolidation, etc. should tend to improve overall utilization over time.

Network Gear

For most network gear your best bet is to scan with SNMP.

Cisco

In the perl client directory you will find a cisco_scanner script. It reads its configuration from /etc/cisco_scanner.conf. There are two variables to set: 'seeds' and 'community'. With seeds you give the script a handful of Cisco nodes in your environment, the script will walk the CDP (Cisco Discovery Protocol) tree to find the rest of your Cisco gear automatically. community is your SNMP read community string. As the cisco_scanner script finds new gear it will add appropriate hardware profiles to the database. You will need to go in and set an appropriate number of "outlets" (i.e. switch ports) and the "Network" outlet type in each hardware profile. You can set appropriate rack size and color as well so that the device displays properly in the visualization charts. If the hardware model is a chassis switch (6509, 6513, etc.) then set the number of outlets to 0 and cisco_scanner will dynamically create/remove outlets for each individual piece of gear corresponding to the number of ports it finds when scanning the device.