Variable Polling Interval

Feature Requests
1

There is a lot of value Polling at 1 minute interval it’s great to find issues with Micro Bursts or for example when bandwidth peaks for 2 minutes and remains low for the remaining 5 minutes

However having every device to poll every device at 1 minute is not optimal

Modern Network devices and Servers can easily handle 1 minute interval, however PDUs, SeriaConsoles, UPS and other devices are slow to respond and there is no benefit polling them every minute.

I’m migrating from Observium and I originally wrote the code to have 1 minute polling into Observium, At this stage Observium cannot keep up with the features of LibreNMS.

Anyone open to take a bounty to add variable polling that is compatible with the standard LibreNMS pollers ?

Devices could be separated in classes and the default could be 5 minutes while special devices would go for more aggressive 1 minute.

2

If you are mostly interested in up/down alerts, perhaps this will be of interest:

I’ve seen crazy schems with multiple pollers (each doing different intervals). But those are prone to issues since RRD files hard-code the heartbeat (interval).

3

For alerting we use ICINGA, switch notifications and SNMP TRAPS.
The faster polling interval is really to catch bursts of traffic that fade when polling at 5 minutes.

4

LibreNMS supports alerting for traps, and they are not bound to the polling interval.

5

Not concerned about alerting, just NEED 1 minute resolution on usage graphs.