Quickest way to add nodes with ping-only monitoring?

My Goal
As a first past / proof-of-concept, the only thing I’d like to do is start monitoring ping responses from 1,800 switches.

(Mis)Steps Taken
I started with the Discovery Configuration screen, adding some 150 networks with about 20,000 IP devices spread across them. I’ll sometimes get 100 New Suspect events on a given second, but what I’m finding is that the same IP is repeatedly discovered, “queued for a services scan”, but never getting scanned. Looking elsewhere in the event log, I see Node Added events happening every 120 seconds on the dot (minus unknown down-time, after which I restart OpenNMS to kick-start the service scan process).

In trying to speed up service scanning / node adding I made the following changes:

  1. In provisiond-configuration.xml I have tried various thread settings, currently running 310 Import and Write threads and 600 Scan and Rescan threads. (As I write this I’m thinking maybe I should drastically reduce the Rescan thread count).
  2. In poller-configuration.xml I’ve set status=“off” for everything except ICMP (and I just realized https).
    At no point does top indicate any system load issues.

Questions

  1. What is the best way to add 1,800 devices (IP address, names, maybe a location field)? Just enter the addresses individually under “Create Discovery Scan” and enter names later on? (Is there a way to import a spread shit (or heck, even using SQL?)
  2. What is the fastest way to turn those entries into nodes and start monitoring only ICMP responses for them?

Thanks for your time and attention!

Sincerely,
Greg

After a node has been discovered, I’ve typically found the service provisioning can take a little time afterwards to complete. I would strongly make sure that your nodes get added to a requisition, as it will make managing changes to them easier in the future. You can set the default foreign source to only have ICMP, so the other services won’t attempt to poll. (changes to poller-configuration.xml require a service restart to take effect).

Alternatively, if you have the nodes in a spreadsheet or SQL location, you can setup PRIS to automatically import from the outside source (called Foreign Source) on a regular basis. By doing this instead of using Discovery, only your specific devices will be provisioned instead of polling every IP in the subnets. Benefit of this method is that updates to names/IPs/locations/other fields in your database will sync automatically to OpenNMS.

If you really want to make it quick, create a Requisition and remove all Detectors, just add your switch with the management IP and assign ICMP manually. The quickest way is probably don’t use detectors and just maintain the management IPs and services yourself.