Re: wrong NTP servers list is used

[Lubomir Rintel <lkundrak v3 sk>]

Hi Mike,

On Wed, 2014-10-15 at 23:36 +0000, Mike wrote:
> I needed to hack /etc/default/ntpdate on a Debian system to contain:
>
>  NTPDATE_USE_NTP_CONF=no
>
> so that it would use the DHCP-supplied ntp-servers when I "ifup eth0".
>
>
> When I try to use NetworkManager 0.9.4.0 to bring up the interface instead,
> it uses a different NTP servers list.  I know this from the /var/log/syslog
> output, where I find a lot of
>
>  ntpdate[...]: sendto(...): Operation not permitted
>
> and corresponding netfilter log lines showing the NTP communication blocked
> (as desired), with a final
>
>  ntpdate[...]: no server suitable for synchronization found
>
> The servers listed number in the dozens and appear similar to those chosen
> by the "ifup" path when NTPDATE_USE_NTP_CONF=yes, i.e.,
>
>  NTPSERVERS="0.debian.pool.ntp.org 1.debian.pool.ntp.org 2.debian.pool.ntp.org 3.debian.pool.ntp.org"
>
> seems to be getting used.
>
>
> This is odd, since I thought /etc/default/ntpdate is only used from the
> /usr/sbin/ntpdate-debian script, whose logic would choose either the
> DHCP-acquired ntp-servers (via /var/lib/ntpdate/default.dhcp), or nothing.
>
> So, where is NetworkManager getting these NTP servers from?

NetworkManager does not do anything with NTP.

It's probably the DHCP client shipped with your distribution that
invokes ntpdate upon getting the DHCP reply. Look for "script" clause in
you dhclient.conf or into dhclient.d, if present.

Lubo