[ntp:questions] Will ntpd tell me if it encounters problems?

Maarten Wiltink maarten at kittensandcats.net
Wed Apr 11 15:23:15 UTC 2007

"gluino" <gluino at gmail.com> wrote in message
news:1176293456.182001.141420 at y80g2000hsf.googlegroups.com...

> I am using the default ntpd setup in plain redhat box. Default
> settings except for specifying a nearer time server.
> Everything seems fine (ntpq pasted below), and it has been left alone
> for months.
> Said box hardly receives human attention, and is depended on for time
> sync by a couple hundred local machines.
> My question is: how would I know if ntpd has problems. I'm thinking of
> some kind of alert when ntpd is unable to contact any good time
> servers, for example.
> # ntpq -np
>      remote           refid      st t when poll reach   delay
> offset  jitter
>  .INIT.          16 u    - 1024    0    0.000    0.000
> 4000.00
> +    2 u  534 1024  377  313.466
> -19.302   2.021
> +     3 u  505 1024  377  253.106
> -13.209   0.184
>     LOCAL(0)        10 l   60   64  377    0.000
> 0.000   0.001
> *  .ACTS.           1 u  582 1024  337   31.425
> 4.047   1.380
> Assume the above is healthy?

Well... marginal. Healthy now, but it wouldn't take much.

You have four servers but one has never been reachable at all and
two others are indeed very far away. Although the jitter isn't
bad, I wouldn't trust them too much if the fourth server went away.
Stability _would_ suffer.

Generating alerts could be done by some simple checking of ntpq -p
output from a cron job. Check that you have an asterisk and a few
pluses. Check the stratum, reach, delay, offset, and jitter in the
relevant lines. If you don't like what you see, send an email and
have a human look closer.

No matter with Richard and his exclamation marks say, I consider
the local clock justified. Having several hundred clients following
a single wandering server in the event of problems is probably
preferable to all of them wandering aimlessly. Stratum ten is fine.

Maarten Wiltink

More information about the questions mailing list