[ntp:questions] ntp_gettime - reconnect

David Woolley david at ex.djwhome.demon.co.uk.invalid
Fri Oct 19 20:45:40 UTC 2007

In article <1192814098.089428.67380 at t8g2000prg.googlegroups.com>,
lmr <lizvettem at gmail.com> wrote:

> Is xntpd configurable to reconnect if the connection fails?

xntpd is either seriously obsolete or misnamed.

No version of ntpd uses connection oriented protocols, so there are no
connections to fail.  See the very recent thread about using TCP.

"The" implies a single server; that is not a recommended configuration.

> What happens to the OS time when the connection fails?

When ntpd has not received successful replies, or broadcasts for an 
extended period of time, the time will coast at its best available 
estimate of one second per second.  It will also stop serving time to
downstream ntpds.  Over shorter intervals additional adjustments may
still be in progress to reflect recent replies.

> How can an application know if the connection has failed?

There is no connection, so the following is about recognizing 
extended periods with no time synchronisation.

Depends on the OS.  If you have the kernel time discipline, which would
require that xntpd be misnamed, you should have a utility called
adjtimex, although it also goes by other names, which can tell you the
current estimated and worst case error bounds on the time; it's up to
you to decide what is an acceptable error bound.  Otherwise
you can use the ntpd management protocols, either directly or using
ntpq or ntpd (or whatever they are misnamed as) to read the state
of ntpd.

> What does ntp_gettime returns if the connection has failed?

Which OS.  There seems to be one in Slackware 11 libc, but I can't find
any documentation on my machine.  I suspect, on that system, it is a
front to the same system call as used by adjtimex and will tell you the
error estimates.

More information about the questions mailing list