[OpenAFS] Problems in stopping the afs-client

Dr A V Le Blanc Dr A V Le Blanc <LeBlanc@mcc.ac.uk>
Sat, 19 Oct 2002 12:13:57 +0100


On Fri, 18 Oct 2002 16:38:30 -0400, Jim Rees <rees@umich.edu> wrote:
>>    It is true that other means may be used for syncronization, 
>>    but then two services (afs and, say, ntpd) should necessarily
>>    be maintained together instead of one (afs) on every host involved.
> 
> I strongly disagree.  If you are running two different time sync mechanisms,
> and they agree, what's the point?  If they disagree, both will get confused
> and nothing will work.

Actually, there's a problem running ntp when you run afsd without
-nosettime: if afsd ever _does_ change the time, it messes up the
ntp algorithm.  I have AFS servers running ntpd, and rarely more
than .003 second off from each other; all clients run ntpd as well.
Occasionally someone starts afsd without -nosettime, and problems
do occur.

The present defaults _sound_ reasonable, but evidently cause problems
in some cases.  The question is, should they be changed?  I think
the answer is not so obvious.

     -- Owen
     LeBlanc@mcc.ac.uk