I had a similar issue a few months back, a sawtooth graph with my server repeatedly dropping out of the pool as its score went above and below the threshold.
Upon investigation I noticed a significant increase in the amount of NTP traffic hitting the server. I presumed this, coupled with the custom rate limiting options I had configured in ntp.conf was the cause. I removed the rate limiting options from the config, but that did not stop the issue, so may not have been a factor. How I eventually solved this was to drop my bandwidth option in the pool to its lowest setting (which cleared the fault and reduced traffic) and then over time gradually increased the bandwidth setting to a reasonable level (though not as high as it was originally). Regards Austin On 25 June 2015 at 00:14, Arnold Schekkerman <[email protected]> wrote: > On 24-06-15 23:54, [email protected] wrote: > > If you look at the CSV log link from your monitor page, your server's > time is > > excellent but the monitoring station is not getting a response > periodically, > > causing your score to decrease by 5 every time because it failed to > respond. Is > > your server being DDoS'd or using more bandwidth than it's allocated? > Your > > stress testing could be causing this as well. Can you ping your server > > externally to check its reachability continuously and check for > intermittent > > downtime? > > > > In other words, the problem isn't between your server and torix and your > > server's ability to keep accurate time, but reachability between client > and your > > server is very poor. I would also check dmesg for any Ethernet issues, > IP > > conflicts, hardware issues, etc. When in doubt, reboot. > > With the CSV-log you can see at what times the monitor sent a time request > to your > server (or you can extrapolate at what time you can expect the next > request). You > can use tcpdump to see if you receive those time requests and if your > server is > sending a valid response. The monitor uses an IP in the net 207.171.x.x/16 > so you > can set tcpdump to only dump those NTP packages. > > If you don't receive a time request at the times the CSV-log indicates, > then you > have a firewall issue at your server or with an upstream internet service > provider. > > If you get the requests and send valid time responses, then there is most > likely an > issue with an upstream ISP as well. > > If you get the request, but you don't send a valid time response (no > response at > all, a KOD, no-sync (both leap bits set), etc.) then it is something in > your local > configuration. > > I hope this helps you to debug the issue further. > > Arnold > _______________________________________________ > pool mailing list > [email protected] > http://lists.ntp.org/listinfo/pool > _______________________________________________ pool mailing list [email protected] http://lists.ntp.org/listinfo/pool
