Brown, Ken F. wrote: > Please bear with me...I've taken over the time services so am just > getting a handle on this... > > (in other words...newbie alert :) >
Once upon a time ALL of us were newbies! Most of us usually remember that. > We have 4 time servers - 3 with a GPS-style acquisition devices that > communicate via the com port - and one with just NTP running on it...all > are running on Windows 2000 (apparently bitten really, really bad with > wildly drifting time...long, long ago) > > The NTP version is [email protected] (from Meinberg) > > The GPS devices are all working (and connecting with hyperterm when > watching the device) all 3 servers are getting time from the devices > (initially - basically, I had to do the manual leap second adjustment > which is what caused me to dig into this). > > After NTP is restarted, and it may be a couple of hours or more...the > servers start picking each other as a 'better' time source. For example > (from the Meinberg GUI monitor): > time2 (refid=GPS) Stratum=1 > time3 (refid=time2) Stratum=2 > time1 (refid=time3) Stratum=3 > time4 (refid=time2) Stratum=2 This should not be! If time3, time1, and time4 are all equipped with GPS receivers, all should be stratum 1. From what I see above, only time2 is equipped with and/or is actually using a GPS reference. > > They all have good 'reach' values (377). > > I've seen the stratum's get higher and higher - to where the 4 time > servers actually end up with stratum numbers in the 8,9,10,11 ranges. > > I found this by using the Meinberg monitoring software from my > workstation (with NTP installed - same NTP version) with NTP.CONF > containing 'server' statements to all 4 of those time servers...I check > it a couple of times a day and they are constantly changing "who" has > the best time (which is normal, from what I've been reading) but the > refid is changing, also. > > I guess the question is: Is that normal behavior (for peering)? Or > should I just reconfigure so that rather than 'peer' statements they all > have 'server' statements pointing to each other (and not 'peer' > statements). > > I hope that makes sense... > > Thanks! > > > > > > The NTP.CONF file for the 3 time servers (with GPS devices) have lines > similar to below (different com ports involved): > > server 127.127.1.0 #allow synchronization with local > clock > fudge 127.127.1.0 stratum 12 refid LOCAL > > # connected to COM1 - use 127.127.29.1 > # connected to COM2 - use 127.127.29.2 > server 127.127.29.2 prefer You don't need the following fudge statement! > fudge 127.127.29.2 stratum 0 > <snip> _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
