On 2013-11-02, antonio.marchese...@gmail.com <antonio.marchese...@gmail.com> wrote: >> >> > server 127.127.1.0 >> >> > fudge 127.127.1.0 stratum 10 >> >> >> >> Remove the above. Although it seems to be ignoring them at the moment, >> they are nearly always wrong, and are incompatible with tos orphan. > > Thanks. I was suggested those parameters on another forum. The reason was > that if the NTPs are not reachable it locks on itself. >
And that accomplishes what for you? >> >> > tos orphan 4 >> >> >> >> Caution 4 is rather low, but that is not relevant here as you are not in >> orphan mode. > > I can change to something higher to be on the safe side. Or remove it. YOu are not using it. > >> >> It looks like the software clock is outside the usable range. What's >> more worrying, if it sometimes works, is that it varies from time to >> time enough to put it just inside the useful range. > > Ok, then I need to ask you what exactly "frequency error" means. I've been > looking everywhere and still I fail to completely understand its meaning. The system advances the system clock by 1/100the of a second or 1/1000 of a second every x ticks of the cpu clock where x is determined on bootup. ntpd determines if the rate of the ticks are done at 1/100 or a second per 1/100 of a second or at some different rate. If it is a different rate then that is a frequency error. Frequecy errors can arise because the crystal driving the system time heats up, or ages or whatever. > > It's a Supermicro Server board, is it possible it's so unreliable?? Sure. No idea what a Supermicro Server board is. > >> The first bit you can deal with by using the kernel time control utility >> (names vary) to correct the frequency in steps of 100 ppm. However, if >> the frequency is not stable, that will not give you good timekeeping. > > Could you point me somewhere about it? I've never heard of it. What is it? Under linux it is a program called adjtimex. > >> > Could someone please help? I can add a forth server if needed, I also >> > understand that it would be recommended to use the iburst parameter. >> >> Adding a Scottish server, or even a fourth one from elsewhere won't help >> as you already have three servers that agree with each other. > > eheh, I beg your pardon, I'm Italian, I mispelled it! :) > >> As it should have stepped before the error reached 7 seconds, it is also >> possible that they disabled stepping. If the drift file value is way >> out, that might cause some problems in getting lock. It will also tend >> to prevent use of the higher quality kernel time discipline. > > How can I verify if the stepping has been disabled or not? > ntp.drift at the moment is -500.000 Which is way out of spec and cannot be corrected by ntpd. > > Thanks for your help! > Antonio _______________________________________________ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
