I finally got my NTP setup working, and after letting it test a while, I set it to 25 Mbps in the server settings. I quickly saw a lot of traffic, and apparently the majority in the first couple hours was from Tesla—including so many packets chrony’s rate limiting started cutting them off:
But now a question for those who’ve been in the pool longer: is my rate limiting configuration for chrony correct? I didn’t want to set it too open, though I know chrony does a better job limiting NTP-based attacks than some older ntpd setups…
At the time of looking the servers score was 15.5 and slowly climbing so early days. One thing I have noticed is there is no such thing as one size fits all. A setting that works in one part of the world would not work that well in another. As an initial setting it looks good but it would pay to keep an eye on things over the next few weeks to see if there are any issues. The US has quite a few servers in the pool so being one of many gives some protection from load spikes.
I would also be interested to hear how your hosting goes over the next few months. It seems you are with `OrgTechName: Contabo Tech`. Some people seem to have no problems with them while others have a different experience. I think the hosting company and its network connections can have a large influence over the quality of the time being served.
No, as you set the polling rate to 2^1 seconds between, meaning people can poll every 2 seconds.
This can have a big impact on your system(s).
Most people set interval to 4~6 to make polling only allowed 2^4 seconds, meaning once every 16 seconds is allowed.
Setting the burst to 16 is also not a good idea as they can poll 16 times real fast, better is 6~8 burst at a time.
This bursting is for making an ntp-client sync faster with the correct time.
You won’t see errors, as abusive clients will be dropped. In the log-file you can find them.
Most clients behave, but there are a few bad out there and can bring your system down if you allow them to keep polling at a high rate.
Beware, the load can go very high when you allow fast polling by default.
I actually purchased the VPS through Oxtrys (never tried them before) since it said they have a US datacenter in St. Louis — I have that VPS set up with a wireguard tunnel to my studio Time Pi directly (didn’t want to expose my studio’s fiber directly to the world).
Ping times are < 3ms between the STL datacenter and here, but external requests go through Contabo’s Chicago network, unfortunately. So 15+ ms round trip for some requests because of that
But so far I haven’t hit any rate limiting or weird situations. I’ll have to see how it goes over the next few months.
I’d rather be generous with the polling; I had considered raising that limit more, but since this is on shared infrastructure I wanted to leave a sane limit.
Since I’ve done a number of tests in the past on direct NTP servers, where it might be useful to poll once per second for a minute or two, I’m okay with letting others do that on this server. If I see abuse, I’ll raise the limits of course. But for now it seems the limits I have in place are stopping a lot of that already:
