[ntp:questions] Local clock question with dialup connection
david at djwhome.demon.co.uk
Wed Apr 18 22:17:52 UTC 2007
In article <77CdnSE-3sg_TLjbnZ2dnUVZ_ozinZ2d at megapath.net>,
hal-usenet at ip-64-139-1-69.sjc.megapath.net (Hal Murray) wrote:
> I see the following when ntpd gets started before eth0 has made contact
> with dhcp. ntp.conf has a single server 192.168.1.3 that's on the
Your init.d scripts should not go beyond the DHCP query until you have
a valid IP address. They should not go beyond the route add default
(which might be done by DHCP) until they have a working route onto the
internet. If you run named, that should be started next. Only then can
you run anything, like sendmail or ntpd, that requires name resolution.
(Even firewall setup may need name resolution.)
The reason that you were only getting one out of three servers is that you
weren't waiting for name resolution to come up before starting ntpd.
>From what I remember, this means you will have to do one of:
1) script the setup of the firewall in a custom init.d script;
2) shell out to the console to allow the firewall to be set up manually,
continuing with the startup when that shell is exited;
3) disable all the higher level TCP start ups in the init.d processing
and run them manually, once you have a valid routing and domain name
service. (Do not disable them with the GUI configuration as that will
probably mean that the firewall won't be opened to support them.)
> No association ID's returned
Because all the name resolutions timed out. You could add the servers
using ntpq, of course.
Incidentally, there is no point in supplying a script to start ntpd, as it
need only consist of one line containing the name of the ntpd program!
Anything else is either unusual options, pre-stepping of the clock, or
distribution specific code to extract the configuration from distribution
specific files. Also note that not every Linux distribution uses the
start up script techniques used by Red Hat.
More information about the questions