client configuration: it's sufficient 3 servers ?

[Riccardo Castellani]

If I have 3 internal NTP servers, what do you think if my clients have all
these 3 servers in their configuration ?
According to '5.3.2. Why should I have
more than one clock?' in 'http://www.ntp.org/ntpfaq/NTP-s-algo-real.htm#Q-NTP-ALGO


I read 2 servers, it's worst case, but also 3...

[Rob]

It depends on what you expect and how actively you are monitoring
your network.

When you want your clients to always have correct time even if a
server fails, you need more than one server. There are people who
claim that 2 servers is not good because they may have different
time and you will not be able to tell which one is right, but in
a correctly configured ntp server it will normally not happen that
it serves wrong time without noticing it.
(EXCEPT when you put that LOCAL clock (127.127.1.0) in the server
config, that is one of the reasons why you should always remove that!)

So, two servers is perfectly OK when you want to increase reliability
on a local network and you have monitoring in place so that when
one server dies you will start to repair it before the other server
dies too.

Also, in many environments it does not really matter when the clients
are not synchronized for a couple of hours. So when you have only
one server (or only one server configured in the clients), it may
still be OK when you have monitoring and someone repairs the server
or puts a temporary other server on that address within reasonable
time.

However, when you have the 3 internal servers it is good to include
the server lines in all the clients configs. That is the most
reliable situation overall. I only want to indicate that it is
not mandatory in cases where you don't need 100% reliability.

[David Lord]

My four pool servers each have 5 x internet sources
(= 20 different internet sources) and 2 x local peers,
there are also local sources with MSF and GPS.

There is an explanation of the selection algorithm in
the ntpd documentation as to why having only 2 and 3
sources can give problems. That sets 4 as minimum and
5 for safety.

I've had internet outages where half the sources have
become unreachable.

David

[David Woolley]

Rob wrote:

> When you want your clients to always have correct time even if a
> server fails, you need more than one server. There are people who
> claim that 2 servers is not good because they may have different
> time and you will not be able to tell which one is right, but in

I believe I've seen examples where ntpd reacts to this by marking both
as false tickers, because it cannot establish a majority clique.

[David Taylor]

Riccardo, to cover the case where both remaining servers would be reject
were one out of three to fail, I would suggest that you should have four
internal servers minimum. However, if your clients are also allowed to
use the pool directive, seeing external servers where many more servers
will be allocated to each client, then three internal likely would
suffice, depending on what OS and what timekeeping requirements you have.

So I would have all three internal servers, plus a pool directive in the
client configuration file, something like:

server 192.168.0.1 iburst
server 192.168.0.2 iburst
server 192.168.0.3 iburst
pool uk.pool.ntp.org

I hope that each internal server is stratum-1, and has a separate
GPS/PPS reference or other source.
--
Cheers,
David
Web: http://www.satsignal.eu

[Brian Utterback]

On 5/24/2013 5:27 AM, Rob wrote:
> claim that 2 servers is not good because they may have different
> time and you will not be able to tell which one is right, but in
> a correctly configured ntp server it will normally not happen that
> it serves wrong time without noticing it.

You are missing the point that it is not the case that two different
servers may have different time, it is the case that they *will* have
different time. The only question is by how much. Obviously, the closer
together they are, the less the impact will be.

Brian Utterback

[Rob]

These are two servers within an organization. They can be interconnected
with a "peer" line. They should be closely together until one or both
of them get "unsynchronized", which the clients will notice.

Of course, with a LOCAL clock they will remain synchronized to LOCAL
and they can drift away from real time.

[David Lord]

see <http://www.eecis.udel.edu/~mills/ntp/html/select.html>


David

[Richard B. Gilbert]

The internet is far from the best source of time! A lot of time is
spent trying to make sense of all the noise!

A gadget whose name I can't remember will lock onto three earth
satellites and give you time to plus/minus 50 nano seconds. Very few
people need to slice time that fine!

[Brian Utterback]

Peering them is a good idea, but isn't going to help, since they will
each view the other as being of a higher stratum then their upstream
servers.

> _______________________________________________
> questions mailing list
> ques...@lists.ntp.org
> http://lists.ntp.org/listinfo/questions

[David Lord]

A bit like my GPS receiver that mostly gives offset << 0.010 ms
currently 0.001 ms.


David

[Harlan Stenn]

The NTP FAQ pages are massively out of date.

As soon as Network Time Foundation has enough money to pay somebody to
work on the support wiki content and the FAQ stuff it will get done:

http://networktimefoundation.org - be a member!

If we get any volunteers to work on this it will get done that way too,
but so far nobody new has volunteered to work on it and the current
volunteers have a HUGE pile of things to do.

So if you want to help volunteer, let me know.

And if you want to get *paid* to work on this, get NTF more members so
NTF has adequate funds, and if you have the skills and "deliverability"
to do the work NTF will pay you to do the work.
--
Harlan Stenn <st...@ntp.org>
http://networktimefoundation.org - be a member!