ntpd "STEP storm" with small offset

[Ward WIllats]

Hi Denys:

> On Jul 25, 2018, at 5:53 AM, Denys Vlasenko <vda.linux at googlemail.com> wrote:
> 
> This isn't busybox ntpd message.
> Actual step in ntpd messages looks approximately like this with ntpd -ddd:
> 

<snip>

> If you do run busybox's ntpd, your messages may be from whatever script
> is set up to run via ntpd -S SCRIPT option.

Yes, the output is from my -S SCRIPT hotplug script, and I am printing out the environment variables ntpd.c sets before it exec's it -- I then send that text to our server. My dump is an SQL SELECT from the sever-side DB table where they land.

> However, step offsets should be larger than 1 second.

Yes. That is the problem. I assume ntpd.c is working just fine to discipline the clock, but I can't have it firing my script with tiny offsets all the time, as I take a STEP to mean the clock has jumped more than a second and I need to throw out the buffered timestamped data I am collecting.

As I wrote, I can "fix" it by not passing along to the rest of my system any STEP received by the script with an offset < 1 second, but I don't think the script should be firing at all for these ~100ms deltas.

Thanks,

-- Ward