Impact of ARP lookup on local NTP traffic

[Gary E. Miller]

Yo Hal!

On Mon, 30 May 2016 18:13:35 -0700
Hal Murray <hmurray at megapathdsl.net> wrote:

> gem at rellim.com said:
> > Yes, but then that would only work for some version of Linux.
> > Changing minpoll fixes all OS, and keeps the change local to avoid
> > unintended consequences.   
> 
> Only if they have a timeout in the right range.

Oh, it is in the right range.  Linux by default has a gc_timeout of
60 seconds.  Anytime after that Linux may delete the ARP entry.

The default minpoll is 64 seconds, default maxpoll is 1024 seconds, so
the window is small, but real.

ddrown showed me the graphs that prove it.

> You probably want to change maxpoll rather than minpoll.

Yup, I see you are correct.  I wonder what happens if I leave
minpoll at the default 6 (64 S) and set maxpoll to 5 (32 S).

On interesting side effect of my restarting ntpd while watching the
stats.  I get huge jumps in the offset when I restart and that
totally confuses the PLL for hours.

Maybe ntpd is reseting the kernel PLL to center on restart, instead
of leaving it at its current setting?

ntpd really needs a way to reload and not restart.

I'm running ddrown stats generator and the effect is huge.

> >> I haven't found a simple way to print out the timer from
> >> the command line.  
> > Which timer?  The poll times are in 'ntpq -p'   
> 
> I was thinking of the time remaining in an arp slot.

Oh, interesting question.  I don't know how to do that.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem at rellim.com  Tel:+1 541 382 8588
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.ntpsec.org/pipermail/devel/attachments/20160530/841d4811/attachment.bin>