I just want to say the NTP Pool group has been very educational. I started running time servers years ago on Raspberry PI’s as a robotics project for my son using simple USB GPS as source. That little hobby became hard core by setting up redundant GPS receivers’ external antenna with minimal multipathing and having PPS signals for accuracy coming from both a Ublox and Adafruit GPS modules.
I joined the NTP Pool program once my servers stabilized and constantly watched and reported on the accuracy of my services along with the traffic shaping by AT&T of UDP 123 during congestion events. Since joining I am now serving up both IPv4 and IPv6 and now have a logging dashboard for both network and for accuracy that I keep up on a monitor here at home.
I have learned so much from this community both here and other sources. I am now servicing about 7 to 8k reqests every 30 minutes.
My Time Servers are on Raspberry PI 4s. I have complied NTPsec and GPSd for these directly and didn’t rely on the versions in the repository.
On the connection data I have my firewall logging “inbound rules” which are pushed into a Graylog server that I use for central syslog logging. I then use the Graylog dashboard to filter those into the graphs and map.
As for collecting system data and some application data I use Telegraf and its NTP plugin uses the ntpmon utility to collect the data. I have configured Telegraf to store these items into an InfluxDB2 time series database for operating system and NTP stats. The dashboard for the NTP stats and operating system data is Grafana.
Lots of moving parts and took some patience to get it to where it is today. It was a good learning experience since at my day job we have a motto if you can’t measure it, it never happened when it comes to operational reddiness.
None of the software I used for this solution required me to pay for it. I use the community or opensource versions. But I do donate to many open-source projects. For example, I do donate to the pfBlockerNG dev team for all the hard work they do to close the front door to services I host from bad actors. I do donate to QUAD 9 for their DNS servers that provide filtered DNS. I do donate to the PiHole group because they do the work of Angles to allow people to filter what they don’t want on their networks.
That’s a great story, NTP is a good topic for kids as it touches on so many subjects.
Graphs are a necessity imo, can’t manage/tune what you can’t measure etc. But then not having anything to compare against makes tuning difficult, so I thank you for yours, and here’s my graph for offset.
If you want to experiment a little, it’d be interesting to me if adjusting the “netspeed” on manage.ntppool.org gives you a proportional change in query rates, and how quickly it changes and changes back.
My current netspeed is set to 768k for both IPv4 and IPv6 hosts. Let me up that to the next value 1 mbps and do some measurement tracking this weekend. I’ll record date/time of the up change and watch the volume and then change it back and follow the volume. I am thinking Saturday Noon Central time to up it and then Sunday Noon Central to put it back.
@ask I changed the netspeed parameters over the weekend and I guess there isn’t much of a difference in the arrival rates with the change from 768 kbps to 1 mbps. I set the speed back on Monday. Here is the chart for past 7 days:
You might try cranking your IPv6 rate up to the maximum. Only clients which specify 2.*.pool.ntp.org get any IPv6 responses, and many of those are not IPv6-enabled. The upshot is the IPv6 traffic is a tiny fraction of the IPv4 traffic, so I doubt you will have any problems going all the way. I’m hoping IPv6 will be more widely distributed by the pool before long.
After a couple days with boosted IPv6 my volume hasn’t changed but the ratio of IPv6 and IPv4 is now about 50/50. I still blame AT&T for throttling NTP.
