I’ve been working on making most of the NTP Pool server components run under Kubernetes to make it easier to manage and continue to extend (for a particular flavor of “easier”, of course). The beta system is running on https://web.beta.grundclock.com/ (separate NTP Pool database, but same login system and login accounts as the regular site).
I have a bunch of upcoming changes to improve the monitoring system – better availability of long-term monitoring data and better support for multiple monitoring servers.
Whoops. I made the beta site “TLS-only” but didn’t fix the monitoring client to load the CA-roots to validate the certificate. I will fix it Monday evening.
The deployment of the monitor code didn’t like that the site forced TLS.
The easy fix for that meant some changes that could impact the production site, too, so instead I rewrote the monitor in Go. This will allow it to more easily also do traceroutes and do more checks per hour, so hopefully it’ll be worth a few days of no monitoring on the beta site.
It’s running on the same server as the production monitor, so I’m interested in any difference in the monitoring results between the monitoring on the production site and the beta site.
Hey guys, I think it would be helpful to have more NTP servers from around the world on the beta site to test the monitoring changes. For example, there are only 5 IPv4 and 5 IPv6 servers (presumably the same servers) in the entire Asia beta zone. So please add your servers to the beta pool to give Ask some more testing data. I have all my 4 NTP servers (all dual-stack) there already.
The previous login system provided a unique username, so it used to be that. The new login system just uses email addresses, and I don’t want to use those for obvious reasons.
There really should be a way to edit the username or better to edit a “what to be called” field on the site (name, organization name, pseudonym, …) and then use something random looking for the URLs. It’s somewhere on the todo.
http://www.pool.ntp.org/scores/173.255.246.13 and https://web.beta.grundclock.com/scores/173.255.246.13 look a bit different, even after taking into account that the time scales are different. Here are some screen captures for posterity (production and beta, respectively):
This server is hosted at Linode and they claim that the reason for the dropped packets is that I sometimes exceed 30000 outbound packets per second on that server. I’m not quite sure if that’s true, but I’m investigating further. My other NTP servers are doing fine.
Hi Ask, I just extended the deletion date of my server (89.101.218.6 - Ireland) to April 2018 - I’m curious what provider the site in Luxembourg is on. I’m still with Virgin Media business fibre and seeing a consistent score of -5 with this LU site. The US scores are horrible and have been for about 1 year now. Never did get to the bottom of the monitoring issue.
What was the IP? (Send it privately if more appropriate). The only errors I see in the logs were for an IP that timed out on the NTP request (and it looks like that bubbles up to the web UI weirdly).
