I think a separate topic to discuss how we could aggregate some data and maybe get a picture of which IPs are persistently abusing pool resources may be useful.
It could be informal crowdsourcing of data, or we could have a separate lightweight monitoring tool that maybe ranks the worst offenders on our servers once a day and uploads the top 100 somewhere, then another program could rank those and see which ones we might be able to assist in using the pool properly.
A lot of times its not malicious, it’s just someone sets up a network with a giant org with thousands of devices channeled through one IP, and they don’t even know there’s a problem because enough requests make it through to keep clocks ‘mostly’ on time.
Ok I will ask the question - Can someone please define “abusive” while thinking to myself that this question alone will keep this thread busy for the next few years with no sign of consensus.
What is abusive to someone running an NTP server behind a home router does not even register to others running a server on different hardware/network equipment. What changes a traffic spike into abuse? Is there some threshold that tips it into abuse?
There was a similar topic the other month - Rate-limiting / banning traffic to server - #31 by gombadi - discussing DDOS attacks using NTP. I asked the question there - how do you know it is a DDOS attack instead of a “normal” traffic spike? I have yet to see an answer.
One of my servers is in India and it sees multiple daily traffic spikes. For example -
Iptables counts the number of NTP packets coming and going from the server with the minute of the count (packets in previous minute), total outgoing packets and average packets per second duriing that minute. 04:30 UTC is 10am Mumbai so the above shows the rate went from 2.7 million to over 7 million then back to 3.4 million. As the saying goes “Thats a traffic spike”
Is it abusive? Annoying possibly but abusive is not what I would call it. If someone had a server behind a home router and received a similar traffic spike (double the packets for a minute) then they would possbily see it as abusive.
As you said @geerlingguy “A lot of times its not malicious…” That was also the theme of the DDOS thread - how do we, server operators, know if it is malicious, abusive or just a lazy sysadmin pointing thousands of clients to one ntp server.
So I am happy to help improve the quality of the pool but first do we need to mark out the boundries between malicious, abusive, poorly setup/designed software/networks and normal traffic patterns so we are all have a common measuring stick?