What NTP Servers are you running

Yes I do.

But I can’t ‘dictate’ time…the VPS are locked to the host-system.

I use Cinfu.com and they gave me locations where there is no limit, no firewall, no restrictions.

Not all VPS-ISP’s give this or will help you in the pool.

I run 1 of my monitors on a VPS and it works great.

The costs are ok, but the smallest package isn’t the best.

Typical the costs are about 100€/$ a year.


Here are mine

All of them are stratum 2, running Chrony and are VPSes rented from HostHatch. The Norwegian servers are set to 3Gbit net speed, while the one in Singapore is 3Mbit. Quite a lot more pressure over in Asia.

Also, just for fun, I have a python script on all of them responding on TCP port 13 (daytime protocol), and using GeoIP to determine the requester’s local timezone. If anyone wanna try: telnet badeand.net 13

1 Like

It’s correct :slight_smile:

bas@workstation:~$ telnet badeand.net 13
Trying 2a0d:5600:30:3e::bade...
Connected to badeand.net.
Escape character is '^]'.
Sun Feb 25 19:31:55 CET 2024
Connection closed by foreign host.

1 Like

@Bas Are your VPSs unlimited data or do they come with a limited monthly bundle?
I’m just trying to work out what a reasonable bundle would be.

They are unlimited, however you have to ask the the locations that don’t have an active firewall that triggers on ‘attacks’ and tell them you run an NTP-server.
As not all VPS-locations are the same. Ask them upfront and tell them what you want to do.

They do whitelist you if needed.


Hi @Kets_One

How much work can depend on your existing skill set. Coming from a sysadmin/SRE background I find my systems do not provide me with enough to keep me entertained.

Costs - There are a number of VPS providers with systems in the $US5 ~ $US10 range that work quite well. They generally have about 1TB ~ 2TB each available. I am looking at providers in less well served pool areas such as Asia and South America.

How much traffic a particular system handles can depend on where it is in the world. For example I have a $5 VPS in Sydney that is going to use about 62GB of its 1TB allocation this month. I also have a $6 VPS in Tokyo that looks like it will use 1.1TB of its 2TB allocation. Both systems are set to 50Mbit connection speed in the pool. I also have a $24 VPS in Singapore that is set to a 6Mbit connection speed but looks like it will use about 3.6TB this month.

davehart@host> telnet badeand.net 13
Trying 2a0d:5600:30:3e::bade…
Connected to badeand.net.
Escape character is ‘^]’.
Sun Feb 25 17:34:20 CST 2024
Connection closed by foreign host.
davehart@host> ping -S 66.220.13.__ badeand.net
PING badeand.net ( from 66.220.13.__: 56 data bytes
64 bytes from icmp_seq=0 ttl=51 time=155.759 ms
64 bytes from icmp_seq=1 ttl=51 time=155.714 ms

That’s a machine at fmt1.he.net in Freemont, CA, US. ipinfo.io shows the correct geolocation, but apparently your provider thinks it’s two timezones further east in US CST rather than the correct PST.

1 Like

I’m using Maxmind’s GeoLite2 database, so I’m not entirely surprised. If you want, you can check if they got your location right in their paid version: GeoIP web services demo | MaxMind
Another issue I’ve identified is that it will report “CST” for a few places in Asia as well because their time zone is the same as “China Standard Time”, making it ambiguous.

On my manage page, the public profile checkbox is already on and the URL is a seemingly-random jumble of letters and numbers bcy3ip4drx8rk giving the profile URL Network Time Protocol’s pool servers titled from my Organization name. I tried changing the public profile URL suffix to davehart but I get an error message:

500 - Server Error

Ouch! That didn’t work, our server hit a bad gear.

(request id 01HQHARD3VG98JD7VKETADSYX7

Oops, I flubbed that test earlier. I failed to notice the telnet was using IPv6. Over IPv4 it did get the right timezone. Over IPv6 it’s wrong:

davehart@tom> telnet -s 66.220.13.__ badeand.net 13
Trying 2a0d:5600:30:3e::bade…
Connected to badeand.net.
Escape character is ‘^]’.
Sun Feb 25 16:26:12 PST 2024
Connection closed by foreign host.
davehart@tom> ping6 -S 2001:470:1:205::__ badeand.net
PING6(56=40+8+8 bytes) 2001:470:1:205::__ → 2a0d:5600:30:3e::bade
16 bytes from 2a0d:5600:30:3e::bade, icmp_seq=0 hlim=44 time=157.244 ms
16 bytes from 2a0d:5600:30:3e::bade, icmp_seq=1 hlim=44 time=156.993 ms

I tried the MaxMind demo lookup for the IPv6 address and it came up with a location near Tampa, FL, US and an ISP name of Hurricane Electric IPv6 Tunnel Broker, which is off by ~5000km. They do offer free IPv6 tunnels, which I gratefully use with my sadly IPv4-only fiber provider in Maryland, but this host is directly situated in a he.net data center in Freemont, CA, with no IPv6 tunnel.

The he.net tunnel gives me less jitter in general than my IPv4 native, as the tunnel terminates in Ashburn, VA, US, which is also where most of my IPv4 traffic goes, and he.net IPv6 is less jittery than my ISP’s el cheapo Cogent IPv4 transit for most destiniations. Some destinations go via he.net IPv4 as well, my guess is only those which he.net IPv4 reaches but Cogent doesn’t. he.net is known for being default free (“tier 1”) on IPv6 but paying for some transit on IPv4.

LeoNTP v1
Easy, straight-forward, high-capacity

I have one of those (also a Leontp v1) laying around aswell. Currently not operational since it only supports IPv4 and IPv4 traffic tends to saturate my connection more than IPv6 (due to regular peaks).
Also, i noticed some drift when it loses connection with GPS satellites. I suspect it does not have a TXCO, but im not sure.

Ohh that’s not good.

@ask can you take a look ?

1 Like

Private multiple VPS with chrony

Corporate multiple:
Baremetal: ntpd / chrony
VPS: chrony
Hardware NTP/PTP Server

The Meinberg systems are excellent! Usually “appliances” are miserable, but Meinberg’s long-term support and general engineering acumen are excellent. (And thanks for sharing everyone, it’s fun to see).


Thanks for showing the error (and the request ID – the web system keeps tracing information and logs for a few weeks so that made it quick to lookup exactly what happened).

As I emailed you I fixed the bug; it wasn’t checking if a URL path was available or not. It’ll go out next I update the production site.

I also added your primary email address to the old account you had (~9 years ago) that had the old suffix assigned so you can re-shuffle them as appropriate.

Our main setup consists of two Meinbergs (GPS/Galileo/DCF77) and a Rubidium holdover clock that are ‘feeding’ time via PTP to the NTP pool servers, which are basically of the shelf hardware systems running Linux (low latency kernel image) and Chrony, as described here. The nice thing is that we can scale this up horizontally as demand increases.

We also run a global anycast instance (any.time.nl), which consist of 30+ VPS’s from Vultr and a number of smaller unicast VM’s: pool.ntp.org: SIDN's pool servers.

1 Like

Sounds lik a logical setup. Pictures?

It’s all explained here (including some pictures):


But basically the main setup I mentioned is this:

(mind you; you don’t need to have a setup like this to be contributing to the NTP pool; any setup can in principle suffice, including ‘el cheapo’ VM’s)


I have a router doing NAT in front of the LeoNTP so IPv6 is not a problem for me. However, due to the massive traffic for zones in Asia, not all routers can handle that much of NAT.

LeoNTP v1 doesn’t hold time very well without GPS, this is true.