Auto discovery feels like magic for ten minutes: your whole network appears, with graphs. Then the honest work starts - because discovery finds devices, not understanding.
Paessler ships auto discovery as the fast start: give PRTG an IP range and it detects the devices in it and creates starter sensors automatically. For a first day that is genuinely valuable. But discovery has a personality - it finds what answers, assumes what it cannot see, and creates sensors with the enthusiasm of someone who will never have to read the alerts. Here is what to trust, what to verify, and what it will never find for you.
Anything awake that answers ping, speaks SNMP with the community string PRTG knows, or accepts the WMI credentials you gave it. Routers, managed switches, printers and servers usually appear with sensible starter sensors, and discovery groups a lot of it correctly. Run it during working hours, when everything is switched on - a discovery at midnight builds a network of only the things that never sleep.
Treat discovery output as a rough draft. Rename every device to the name your team uses, not the name the scan guessed. Delete starter sensors nobody will ever read - discovery loves creating disk sensors for recovery partitions. Fix credentials per device group so WMI sensors turn green. Then walk the office list: which known device is missing from the tree? Add it by hand. The day the tree matches the building, discovery has done its real job.
PRTG can rerun discovery on a schedule, which sounds like free maintenance. In practice, schedule it rarely and review what it adds, or your sensor budget quietly fills with every new phone and guest laptop that wandered onto the subnet. A monthly rediscovery with a human deleting the noise beats a weekly one nobody audits.
Every sensor in this guide is created live in the video - installation, discovery, CPU and RAM, bandwidth and firewall status.
Watch Full VideoAlmost always credentials. Without working Windows credentials in the device settings, PRTG cannot query WMI, so it falls back to ping. Add an account with local admin rights on that server and the CPU, memory and disk sensors become available.
A scan sends pings and SNMP/WMI queries across the range - light traffic for a small office. Very old or fragile devices occasionally dislike aggressive scans, so run the first discovery while you are present and watching.
Only if the probe can route to it and you include the range. For separate sites, PRTG's model is a remote probe installed at that site reporting back to your core server - discovery then runs locally there.