Your pollers can not keep up with the work needed. Check the load of the servers in question and increase worker threads / servers as needed. Check the load after increasing these values. You do not want to overload the server. Another possibility is that your devices are too slow to respond.
That depends on your physical server hardware. If you actually have available resources on the physical host, you may increase the amount of vCPUs but do not overallocate resources. Another solution would be to spin up another poller on another host. You can check the amount of poller / discovery threads from the Settings → Poller → Settings. You can try to increase the thread amount a little (like 10 - 20 %) and see how the server handles the load. If you change these settings, you need to restart the poller service. You should at least monitor the load averages a while when doing this - if the server hangs or the load is too high, you need to get faster hardware or spin up another poller server preferably on another physical host.
Hi Gonzo, could you share what values you have in poller page (http://[your_librenms_instance]/poller)
and do you have a pop up concerning unpolled devices from time to time ?
