Bug#393509: [pkg-ntp-maintainers] Bug#393509: ntp: does not adjust time

[Kurt Roeckx]

On Tue, Oct 17, 2006 at 02:26:10PM +0200, Christoph Martin wrote:
> Hi Kurt,
> 
> Oct 16 17:50:35 gonzo ntpd[27046]: frequency initialized 0.000 PPM from
> /var/lib/ntp/ntp.drift
> Oct 16 17:50:45 gonzo ntpd[27046]: synchronized to 134.93.227.254, stratum 2
> Oct 16 17:48:59 gonzo ntpd[27046]: time reset -105.185036 s

That would be caused by the -g option, just picking one of the clocks at
startup, so that your more or less correct.

> Well, I am not shure about that because the other host which works has
> the following stats:
> 
>  134.93.226.2    134.93.225.254   5 u   27   64  377    0.262  -3116.1
> 1075.62
>  134.93.226.19   134.93.227.254   3 u   54   64  377    0.330  -1174.2
> 1370.48
> *134.93.227.254  131.188.3.220    2 u   48   64  377    0.313  -1791.6
> 917.537

So, it's synched to 134.93.227.254, but it's 1.8 seconds off?  And there
is a difference of 1.3 seconds between the clocks?

What is also weird is that your interval still is at 64 second?  Did it
have a time reset recently or something?  Or was it recently restarted?
If it's been running for a while without problems it really should be at
an interval of 1024 seconds.

If you ask me, atleast 134.93.226.2 and 134.93.226.19 aren't working
properly either.

> That is true. I have to look into it. But it might be because the two
> servers have the same list of upstream ntp servers but decide to sync to
> different ones. But I should make shure that they have the same time. Do
> you have any idea how to bind two ntp daemons tightly?

Afaik, ntp doesn't really bind to a single ntp server, but to a group of
servers, and selects the one it thinks is the most accurate.

Having them select different servers shouldn't be a problem, as long as
all servers actually give proper time back.

Maybe you should give an overview of how things are all setup?


Kurt