attila@kinali.ch said: > This is a pretty baseless fear. The servers in the ntp pool are constantly > monitored and those that are off by more than 100ms are quickly removed > (within 2-3 hours, IIRC). Of course, if you are already using one of those, > then the removal will not help you. But you are most likely using 3-5 servers > anyways, which means ntp will remove the "rouge" server on its own. It's more complicated than that. It depends on what code you are using and how you configured things. If you are using ntpd and you said in your ntp.conf server <pool> Then it grabs one and sticks with it until you restart ntpd. In the old days, it was common to use server 0.pool server 1.pool server 2.pool server 3.pool That used the pool before the pool code in ntpd was working. I'm pretty sure some distros set things up that way and some systems are probably still using an old config file. The pool code is supposed to drop bad servers and get replacements. I'm not sure of the details on what "bad" covers. It could be not responding at all or it could be time not good-enough. I'll dig into the code if it matters. If you aren't running ntpd (classic or ntpsec) then I don't know what happens. -- These are my opinions. I hate spam.

Hi Since “rogue” servers are rare, bumping up the number of servers fairly quickly gets you to a very high degree of confidence. Is that 5, 7, 9, or 11? It sounds like a wonderful topic for somebody’s thesis or dissertation :) Given that this is a free resource and that the network usage is negligible even with a dozen servers, the only real downside is being tagged as a resource hog. Bob > On Nov 4, 2019, at 7:05 AM, Hal Murray <hmurray@megapathdsl.net> wrote: > > > attila@kinali.ch said: >> This is a pretty baseless fear. The servers in the ntp pool are constantly >> monitored and those that are off by more than 100ms are quickly removed >> (within 2-3 hours, IIRC). Of course, if you are already using one of those, >> then the removal will not help you. But you are most likely using 3-5 servers >> anyways, which means ntp will remove the "rouge" server on its own. > > It's more complicated than that. > > It depends on what code you are using and how you configured things. > > If you are using ntpd and you said in your ntp.conf > server <pool> > Then it grabs one and sticks with it until you restart ntpd. > > In the old days, it was common to use > server 0.pool > server 1.pool > server 2.pool > server 3.pool > That used the pool before the pool code in ntpd was working. I'm pretty sure > some distros set things up that way and some systems are probably still using > an old config file. > > The pool code is supposed to drop bad servers and get replacements. I'm not > sure of the details on what "bad" covers. It could be not responding at all > or it could be time not good-enough. I'll dig into the code if it matters. > > > If you aren't running ntpd (classic or ntpsec) then I don't know what happens. > > > -- > These are my opinions. I hate spam. > > > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there.

Each NTP Pool server's real-time status can be seen at https://www.ntppool.org/scores/*ip_address* I can't answer questions about the monitor's source code <https://github.com/ntppool/monitor > NTP server timestamp errors happen occasionally. Internet delays and losses are unpredictable. My paper <http://leapsecond.com/ntp/NTP_Paper_Sommars_PTTI2017.pdf> discusses such NTP time transfer issues (thanks Tom). A diverse set of servers may be desirable: Using eleven stratum 2 servers with common stratum 1 sources may not meet your requirements. If the bad guys can intercept NTP traffic timestamps can be altered, unless NTP authentication is used. [This rarely happens.] On Mon, Nov 4, 2019 at 2:51 PM Bob kb8tq <kb8tq@n1k.org> wrote: > Hi > > Since “rogue” servers are rare, bumping up the number of servers fairly > quickly > gets you to a very high degree of confidence. Is that 5, 7, 9, or 11? It > sounds like > a wonderful topic for somebody’s thesis or dissertation :) Given that > this is a free > resource and that the network usage is negligible even with a dozen > servers, the only > real downside is being tagged as a resource hog. > > Bob > > > On Nov 4, 2019, at 7:05 AM, Hal Murray <hmurray@megapathdsl.net> wrote: > > > > > > attila@kinali.ch said: > >> This is a pretty baseless fear. The servers in the ntp pool are > constantly > >> monitored and those that are off by more than 100ms are quickly removed > >> (within 2-3 hours, IIRC). Of course, if you are already using one of > those, > >> then the removal will not help you. But you are most likely using 3-5 > servers > >> anyways, which means ntp will remove the "rouge" server on its own. > > > > It's more complicated than that. > > > > It depends on what code you are using and how you configured things. > > > > If you are using ntpd and you said in your ntp.conf > > server <pool> > > Then it grabs one and sticks with it until you restart ntpd. > > > > In the old days, it was common to use > > server 0.pool > > server 1.pool > > server 2.pool > > server 3.pool > > That used the pool before the pool code in ntpd was working. I'm pretty > sure > > some distros set things up that way and some systems are probably still > using > > an old config file. > > > > The pool code is supposed to drop bad servers and get replacements. I'm > not > > sure of the details on what "bad" covers. It could be not responding at > all > > or it could be time not good-enough. I'll dig into the code if it > matters. > > > > > > If you aren't running ntpd (classic or ntpsec) then I don't know what > happens. > > > > > > -- > > These are my opinions. I hate spam. > > > > > > > > > > _______________________________________________ > > time-nuts mailing list -- time-nuts@lists.febo.com > > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > > and follow the instructions there. > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. >