On 2010-11-09, David Woolley <[email protected]> wrote:
> unruh wrote:
>
>> If I recall correctly his problem was that he had 4 outside sources plus
>> the local time. Then something disconnected him from those outside
>> sources for a long time ( hours) After that was fixed, the system never
>> again went back to those outside sources-- gave zero reachability even
>> after they were reconnected, as if ntpd gave up on them after it could
>> not reach them for a while. 
>> 
>
> It had fallen back to a 1024 second poll interval, so it might be as 
> simple as he failed to wait until a successful poll was made.  He was 
> possibly expecting it to find the servers within tens of seconds.

That is possible. Only he can tell us how long he waited.


>
> One could possibly argue that a response from another server, after
> the
> local clock became selected, ought to force the poll interval down.

I think that ntp degrades the poll interval if there is not response, to
prevent the situation where ntpd suddenly floods a server that has gone
down. I think it used to decrease the poll interval which resulted in a
server coming up being bombarded by thousands of ntp requests/sec,
because
everyone had decreased their poll interval.

>
> I don't know if ntpd handles the poll interval differently when there is 
> a local clock selected as against when there is no source, but it should 
> probably not.

_______________________________________________
questions mailing list
[email protected]
http://lists.ntp.org/listinfo/questions

Reply via email to