[ntp:questions] Re: "ntpd" and hwclock

Richard B. Gilbert rgilbert88 at comcast.net
Tue Oct 4 19:16:40 UTC 2005


Ulrich Windl wrote:

>mike <michael.no.spam.cook at wanadoo.fr> writes:
>
>  
>
>>>My question does ntpd at any point changes the hardware clock ? If not then
>>>who and when it is
>>>done ?
>>>
>>>      
>>>
>>   my linux (mandrake) also syncs the hardware clock at shutdown. most
>>others will probably do so I guess. So that last delta into the 11 min update
>>is taken care of aswell.
>>    
>>
>
>IMHO the correct time to update the CMOS is when the system time changed, not
>when the system goes down. For SUSE Linux and the RTC running localtime (not
>UTC), a single-user-boot followed by a reboot has always messed up the system
>time. For mandrake you may be more lucky...
>
>Regards,
>Ulrich
>  
>
I'm inclined to agree!  Setting the CMOS clock, or equivalent in other 
architectures,  only at system shutdown ignores the probability that the 
system will not do a clean shutdown.   The kernel might panic, the power 
might fail, and then you boot up with your clock off by 37 minutes!

All right, I exagerate.  A little!  But if you are going to use the CMOS 
clock to set the system clock when you boot, it makes a great deal of 
sense to keep it up to date.

I think it makes more sense to use ntpd or ntpdate to set the clock when 
you boot.




More information about the questions mailing list