[ntp:questions] Re: two ntp servers

Jed Clear clear at alum.mit.edu
Tue Feb 21 03:03:41 UTC 2006


rancid moth wrote:
> 
> hello john
> 
> no it seems to definitely be reaching them.  i have checked using the
> dmpeers command.  also running the ntpq -p i get the exact same thing as
> what is on the other box.  also note that in my trials i have set a machines
> time from off the other machine and made them perfectly in synch.  when i
> then take that machine and point it to the upstream server it then drifts
> until it is three minutes ahead.  i just dont get it.  odd dont you think?

Could be something misconfigured in your ntp.conf that prevents ntp from
actually setting the local clock, so it continues to drift.  Auth can be
tricky.  Also I recall posts of an option that seems to have changed logical
sense after a certain older ntpd version, but examples might not be updated.

I would start with a very simple ntp.conf:

	server ntp.yourISP.com

and *nothing* else.  If your ISP doesn't offer NTP (you might have to ask their
support desk), go pick an open stratum 2 off the wicki.  Don't use the pool
initially as your two servers could (should) get different servers for the same
name.  If that works, add a bit of your current ntp.conf back in at a time until
it breaks.  From what I've read here, you shouldn't be using minpoll and maxpoll
options at all.

If you find you need to come back for more advice, those more knowledgeable than
myself could probably help more if they knew what version of ntpd you're running
and what's in your ntp.conf.

HTH,

-Jed

> 
> "John Pettitt" <jpp at cloudview.com> wrote in message
> news:43F4FDCA.3070402 at cloudview.com...
> > rancid moth wrote:
> >> hello,
> >>
> >> not being very familiar with the ntp protocol, and currently introducing
> >> myself to it properly (i.e. have always used though never dug deeply into
> >> it), i have established two ntp services on two different Linux machines.
> >> Both ntp servers are pointing to the _same_ upstream server.  There is a
> >> four minute difference in time on the machines.  Lets call them machine A
> >> and B.  initially i established them to the same time, by synching B's
> >> time
> >> from A which itself was synched from the upstream internet ntp server.
> >> If i
> >> then modify B's config to also synch from the same upstream source, B,
> >> given
> >> enough time, will drift 4 minutes ahead of A.  Can someone please point
> >> me
> >> to an introductory explanation as to why/how this can occur.
> >>
> >> cheers
> >> moth
> >>
> >>
> >
> > Basically your setup isn't working for some reason - if you do "ntpq -p"
> > on both boxes you'll
> > probably see that one or both of them are not actually talking to the
> > upstream.    As a general rule
> > you need at least four upstream servers for stability (see
> > www.pool.ntp.org for how to get more
> > servers).  It's probably a good idea to peer the two local servers and
> > then point them at diverse
> > (overlapping is ok) groups of upstream servers as this will give to a very
> > robust setup.
> >
> > If I had to guess I'd say that a firewall somplace is blocking the ntp
> > packets.
> >
> > John




More information about the questions mailing list