On 2007-08-14, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote: > On Aug 14, 11:25 am, Steve Kostecke <[EMAIL PROTECTED]> wrote:
<snip: minimal multicast / orphan howto> > Thanks for the quick reply. > > With regard to your reply, just to be sure I am understanding the > right thing and based on my readings, the ntp configuration that > I want to test/use eventually would use the config parameters > "manycastclient" and "manycastserver" instead of "multicastclient" and > "broadcast" : multicast/broadcast and manycast are entirely different modes. In multicast/broadcast mode, the client listens on the multicast/broadcast address. Multicast/broadcast servers send out an NTP packet every 64 seconds. When the client first hears a server a temporary unicast association is established between that client and server. This temporary association is used to calculate the broadcast delay between the server and the client. Once the broadcast delay is known the unicast association is discarded and the client resumes listening on the multicast/broadcast address and now utilizes the time stamps in the NTP packets. In manycast mode, the client sends out server discovery probes. Once the client locates a server a unicast association is established between the client and server. > # I am already on a secure network and I want the smallest IP packets > # possible > disable auth This makes a negligible difference in packet size and leaves these systems open to a rogue time server. > # Each computer connected to a local NEMA GPS > server 127.127.20.1 iburst prefer minpoll 4 maxpoll 10 1. iburst has no effect on ref-clocks 2. prefer is meaningless when you only have one ref-clock 3. 'minpoll 4 maxpoll 10' is almost the default for ntpd. In my experience 'minpoll 4 maxpoll 4' makes ntpd follow the GPS more closely. > tos orphan 7 > manycastclient 239.1.1.1 > manycastserver 239.1.1.1 Fine. Any multicast or broadcast address will do. > driftfile /etc/ntp.drift It is not considered a good practice to allow daemons to write to /etc /var/something/ntp.drift is a safer choice > If all the GPS devices loose their lines of sight with satellites, > then NTP will automatically enter the orphan mode on each computer, > which means that one of the computer will be automatically "elected" > as the "time master" Based on random values generated by ntpd at start-up. > and that all the other computers will be synchronized on it (of course > then the time in the system will drift with respect to the UCT time) If you have one system which you know has a more stable clock set that system's orphan stratum to 6 (i.e. one less than everyone else's) so that it will be prefered in the event that your systems enter orphan mode. -- Steve Kostecke <[EMAIL PROTECTED]> NTP Public Services Project - http://support.ntp.org/ _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
