Newb quetsion here concerning overall best practices when adding network devices

I have a hard time grasping how much I should really be throwing at LibreNMS. For example, if I have 40 access switches…what is the best thing to be monitoring? I feel like uplinks and system stats are decent enough for the most part, and if I had trouble with a particular access port I could come back later and add it in to watch it. Is that normal?

I’m very new at this, and wanted to quickly get a Cisco switch into LibreNMS to see what I could see. I just setup a RO community string, pointed it to my LibreNMS server, and enabled traps on the switch. It started giving me data, but I notice it is every port…I’m just sort of lost as I have never done anything with NMS before.

Any guidance you all could share is appreciated!

Just add the devices. On all my access switches I monitor all ports including the access ports. Granted I have the silly port up/down rule off as its not needed. And it just causes noise on the alerts. You want to keep your alerts kept neat and only alert what is needed. Also only send alerts on transports that high importance the other just set to mute if not that important.

In all, I have 600 device and 16787 ports with no issues. Polling time is 178 seconds with all in one server.

I recommend you run the performance doc as you start out and make sure to setup RRDcaching.
https://docs.librenms.org/#Support/Performance/

Also, I recommend watching this video on network alerts… it’s from SW but they have some good points on alerting.

1 Like

Also, traps are not used with LibreNMS just FYI.
Its mostly just SNMP.

Thank you. I’ve also often wondered what type of load network monitoring puts on the network…it sounds like it must be negligible if you are able to do that many with no issues?

I assume you have dedicated hardware doing this? Is it pretty beefy?

Network Monitoring, in general, doesn’t put a load on the network itself its really small data from SNMP.
The big load is on the Network Monitoring servers. Yes, I am using one beefy physical server. Others use VM but just depends on your environment, number devices, ports, sensors etc…

HP Proliantdl380gen8
CPU: 2 Sockets, 24 Cores
Memory: 32GB
Disks: Raid 10, SAS Drives 250 GB