In article <[EMAIL PROTECTED]>, Christophe Greisberger <[EMAIL PROTECTED]> wrote:
> I have an exact clock and set the stratum to 16. > Is there any ntp.conf tweak to force the stratum to a lower value? The best solution is to use a clock for which ntpd has a clock driver so that ntpd knows that the clock is working and can have confidence that, for example, the time isn't simply being stepped every few minutes, but is being properly disciplined in both phase and frequency. (You many want to consult your clock supplier; if they don't have an ntpd driver, it is likely that they are not seriously in the market.) As clock drivers are generally better supported on Unix, a good idea would be to use a cast off machine (can be a 486DX) as the machine with the clock and run FreeBSD on it. (Such a machine will be much better as a time server than any NT box.) Failing that, you can configure a local clock driver. The disadvantage with this is it doesn't know the true time quality, e.g. time since last clock sample and maximum drift rate, so it will report falsely optimistic time quality parameters. It will also not know if the clock has failed. The workaround is to fudge its stratum to a relatively high value. The default is around 3, but normally people are advised to use 10, which makes it very difficult for someone with access to a known valid source to use it by mistake. As this is a more legitimate use of local clock than the increasingly common one of using it to distribute an arbitrary function of time from an unsynchronised machine, one might tolerate a lower stratum, but the lack of fault reporting is still an issue. > Or even a parameter of ntpd so that it accepts high stratum clocks? There is never going to be a parameter in any official version that disables this part of the time quality checks. Stratum 16 means unsynchronised and a server reporting it will normally also send other indications of this state, e.g. leap indicator = 3. _______________________________________________ questions mailing list [email protected] https://lists.ntp.isc.org/mailman/listinfo/questions
