[ntp:questions] how to change the runtime priority - AND crazy psycho GPS

Dave Hart hart at ntp.org
Sun Feb 5 07:32:55 UTC 2012


On Sun, Feb 5, 2012 at 01:54, Ron Frazier (NTP)
<timekeepingntplist at c3energy.com> wrote:
> I thought there was a way to override the application's built in priority
> selection, other than using task manager.

You may be referring to a command-line option to raise the priority.
That option makes a difference for ntpd on most platforms, but on
Windows it always raises to realtime (or high if lacking privilege).

>  I may have to try the method you
> mentioned if my testing indicates it's necessary.  Here's why I think it may
> be a problem.  Take a look at the following loopstats file:
>
> http://dl.dropbox.com/u/9879631/loopstats.20120129p08-3-real-priority
>
> Note that is has a number of discrepancies and jaggedy areas and
> discontinuities along the time axis..  It also has a maximum excursion of
> around 30 ms offset.
>
> Now look at this file:
>
> http://dl.dropbox.com/u/9879631/loopstats.20120203p08-3-above-priority
>
> This is representative of what I've been getting from my USB GlobalSat
> BU-353 when I have it set on Above Normal priority after switching it with
> the task manager away from real time priority.  Note that the scales are
> different if you chart it.  There are no significant discontinuities and
> maximum excursions are in the 15 ms range.  For a USB only device with no
> PPS, I think this is a pretty good chart.  It seems to me, this is
> performing better.  I'll admit I have to do more testing to confirm any
> patterns.

Please post again when you feel you have enough data for a valid comparison.

> Actually, I've diverted my attention to troubleshooting a worse problem.
>  Once I figure that out, I plan to get back to the priority issue.  After
> about 3 days of rock solid operation similar to the above chart, my GPS
> suddenly goes phycho on me.  Offsets go wild and get up into the 100 ms
> range.  See the next file:
>
> http://dl.dropbox.com/u/9879631/loopstats.20120203p08-1-above-crazy-gps
>
> As you can see, this continued for about 12 hours until I got tired of it.
>  I stopped the ntpd service.  Unplugged the GPS.  Waited 30 seconds.  Then
> plugged it in again and restarted the ntpd service.  Now I'm back to a
> little over a full day of rock solid operation again.  I'm waiting to see if
> it freaks out 2 days from now.  If anyone can help me determine what's up
> with that, I'd be very grateful.  I have verified that the TWO sets of
> firmware in the device (one for GlobalSat and one for the SIRF III chipset)
> are up to date as well as the USB - serial driver in my PC.  I suspect a
> design flaw in the GPS firmware, perhaps a memory leak, or maybe a similar
> flaw in the Prolific USB - serial driver.  However, I'm just speculating.

That's a bizarre loopstats alright.  In two 8 second intervals you go
from < 5 msec offset to 80 msec and back to less than 5 msec,
suggesting a wayward system clock.  I wish I had more feedback -- this
is unfamiliar territory for me, though A Carver's recent post
regarding ntpd on netbsd/sparc had similarly wild loopstats, I don't
really have any clue what might be causing either problem.

> Here's my ntp.conf file:
>
> http://dl.dropbox.com/u/9879631/ntp.conf
>
> By the way, can you do attachments on this NTP Questions list.  I think I
> read somewhere that you cannot, which is why I'm using dropbox.  However, I
> may have to delete these files after a while..

Attachments and HTML are stripped by the mailing list.  If space
and/or clutter are a concern, you might consider using a pastebin
website.  Those often permit indefinite-lifetime posts (good until the
site goes away or has trouble).

Cheers,
Dave Hart


More information about the questions mailing list