Jan Ceuleers wrote: > Richard B. Gilbert wrote: > >> Sorry, it's the "orthogonal" part that's bothering me. My dictionary >> says "pertaining to or composed of right angles". It's frequently >> used as a buzz word but seems to be without content in the context of >> NTP. > > It also means "independent" or "uncorrelated". For example see > http://en.wikipedia.org/wiki/Orthogonal#Computer_science > > So Uwe's point is correct: designing an NTP hierarchy to be a failsafe > system (up to a point) does not preclude it from also reporting failures > even if they are not (yet) service-affecting. In fact, (and Uwe also > made that point in his RAID analogy) _not_ reporting failures gives the > administrator a false sense of security. > > So +1: ntpd should report failures to syslog. > > The question is what sorts of things it should be reporting. Things that > I can think of: > > - synchronisation not achieved within the expected period after startup; > - stratum higher than expected > - smaller than expected number of servers reachable > - the set of reachable servers consists of exactly two servers of equal > stratum (which is the worst case) >
Okay, but . . . . Somebody has to be checking syslog fairly frequently. You'd better believe that there are machines out there that could catch fire without anyone noticing. Some of them may be serving time and even keeping time well. _______________________________________________ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions