On Thu, Aug 1, 2013 at 3:58 AM, James Patterson
<[email protected]> wrote:
>
> Hello,
>
> I have two questions.
>
> 1. I have to use some ntpd servers which list themselves as time
> sources. They list themselves as stratum 16 and with a jitter or 16000,
> so I hope this means they are ignored - but does it?
> i.e. we point at two ntp servers 1.1.1.1 and 2.2.2.2 and when I run ntpq
> -p IP I see that both 1.1.1.1 and 2.2.2.2 are listed with a maximum
> stratum and a huge jitter.
> Is this a good idea, and will it cause a problem? Can you point me at
> some documentation?

Do you mean the undisciplined local clock, fudged to a high stratum?
(As described on http://doc.ntp.org/4.1.1/driver1.htm)?

It's not that uncommon to see this used in the wild, though I tend to
dislike it personally. Where it does make a lot of sense is in
environments where you are serving time to a private LAN and it is more
important that the clocks all agree with each other than that they
align with reality. (Consider a cluster of database servers that don't
agree on the time. Nothing good will come of that.)

I tend to think that is only appropriate in private LANs, though. For
public-facing servers, I'm of the opinion that if you don't have the
correct time, you should not serve guesses. Let the client deal with
it as a routine server failure, not by offering inaccurate time, even
if at a higher stratum. (But that's just my unqualified opinion.)

More to the point, http://www.pool.ntp.org/join/configuration.html
recommends against using a local clock as a source.


> 2. To determine the health of the ntp service on a server we parse the
> output of ntpq -pn and look for large offsets. This doesn't make a lot
> of sense to me since ntp will ignore crazy servers. What's the best way
> of asking ntpd if it is receiving time from a good time source?

I find "ntpq -crv" far more useful here. Look for "sync_ntp" in the
top, and then, in the key/value pairs, you can sanity-check stratum,
root dispersion, and, more directly, offset. I have Munin set up to
plot some of this stuff on graphs.

That said, I do that to satisfy my curiosity. My server in the pool
does not have any active monitoring/alerting (a la Nagios). I just
let the pool's monitoring remote me from the pool if things go awry.

Kind regards,
Matt
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to