Any limitations for 1-minute polling of 2000 cisco routers?

Help
1

Hi,
Can anyone tell me if there’s an approximate limit to how many devices can be polled within a 1-minute timeframe?
Is libreNMS’ poller written in php? If so, how does it handle polling more devices than it can achieve within 1 minute? Does it just fire up several processes of that poller?

I’m currently running cacti with its multithreaded spine poller to poll a lot of routers with 1-minute polling but would like to move to libreNMS for some of its other features… but not at the expense of losing a multithreaded poller.

I’ve read the docs and can see the “poller service” but it says it’s highly experimental and may not work. Has anyone had long-term success with this poller? Are you confident enough with it that you’d deploy it in a production environment?
Or is there anyone here who’s running the default cron poller with 2000+ cisco routers and 1-minute polling and can confirm that it’s working fine for them?

Thank you!

2

i saw much greater cpu load on the server when using 1 minute polling just for 250 devices, are you sure all your 2000 devices complete polling within 60sec? i would prefer 1 min for edge devices to get the granularity and 5 min for the rest, but doesn’t look like this is possible to split them up

Remember also 1 min polling will increase the CPU of your Cisco routers, so if you have routers with high CPU already it wouldn’t be a good idea.

it’s multi-threaded and you can specify the number of threads to be used, my librenms got sluggish with 1min polling @ 72 threads. It will also cause higher IO so you want to check your not waiting on your disk with atop.

3

@leegil were you able to get this working? i am also trying to move from Cacti to LibreNms… but worried about the poller. Please let me know if you were able to get this working?