Can anyone interpret these loopstats? - NTP
This is a discussion on Can anyone interpret these loopstats? - NTP ; Folks,
I've been monitoring my Windows Vista system using the Meinberg
distribution NTP, and it seems to have a peculiar behaviour of alternate
good and bad timekeeping:
http://www.david-taylor.myby.co.uk/mrtg/gemini_ntp.html
Can anyone who understands loopstats and peerstats better than I do
explain ...
-
Can anyone interpret these loopstats?
Folks,
I've been monitoring my Windows Vista system using the Meinberg
distribution NTP, and it seems to have a peculiar behaviour of alternate
good and bad timekeeping:
http://www.david-taylor.myby.co.uk/mrtg/gemini_ntp.html
Can anyone who understands loopstats and peerstats better than I do
explain what's going on. The files are here:
http://www.david-taylor.myby.co.uk/g...ista-stats.zip
In the event log NTP is showing errors: "Unable to initialise the .rnd
file", but I don't think that's key to the operation of NTP itself. The
NTP service shows as running, and the W32time service as stopped.
Thanks,
Puzzled from Edinburgh (a.k.a. David)
-
Re: Can anyone interpret these loopstats?
David J Taylor wrote:
> Folks,
>
> I've been monitoring my Windows Vista system using the Meinberg
> distribution NTP, and it seems to have a peculiar behaviour of alternate
> good and bad timekeeping:
>
> http://www.david-taylor.myby.co.uk/mrtg/gemini_ntp.html
>
> Can anyone who understands loopstats and peerstats better than I do
> explain what's going on. The files are here:
>
> http://www.david-taylor.myby.co.uk/g...ista-stats.zip
>
> In the event log NTP is showing errors: "Unable to initialise the .rnd
> file", but I don't think that's key to the operation of NTP itself. The
> NTP service shows as running, and the W32time service as stopped.
>
> Thanks,
> Puzzled from Edinburgh (a.k.a. David)
>
>
How about your ntp.conf? What does your ntpq -p banner look like?
I would make a small bet that the network between you and your upstream
servers is a real mess but looking only at your graphs doesn't tell me
much more than that.
I'm rather puzzled by the timing of the relatively flat period on your
daily graph. It seems that you are getting the best performance between
about 11 AM and 9:30 PM. I would expect the network to be busiest
during that period and consequently I would expect to see the worst
performance during that period.
What happened on Friday of week 45? It looks as if that was about the
time everything turned to worms!
-
Re: Can anyone interpret these loopstats?
David J Taylor wrote:
> Folks,
(...)
> In the event log NTP is showing errors: "Unable to initialise the .rnd
> file", but I don't think that's key to the operation of NTP itself. The
> NTP service shows as running, and the W32time service as stopped.
Hello Devid--
You could find a GNU/linux or freebsd system with openssl installed:
# openssl rand 1024 > x
....then move the file "x" to someplace your "other os" ntp can find it
with the name "dot rnd" (not sure if you can specify another name).
Not sure about the data plot question. They don't look that bad to me,
actually.
//Steve
-
Re: Can anyone interpret these loopstats?
"rtxo" wrote in message
news:13kecijgavaarf8@corp.supernews.com...
> David J Taylor wrote:
>> In the event log NTP is showing errors: "Unable to initialise the .rnd
>> file", but I don't think that's key to the operation of NTP itself.
>> The NTP service shows as running, and the W32time service as stopped.
> You could find a GNU/linux or freebsd system with openssl installed:
>
> # openssl rand 1024 > x
>
> ...then move the file "x" to someplace your "other os" ntp can find it
> with the name "dot rnd" (not sure if you can specify another name).
There's an environment variable 'RANDFILE' on my system with a likely-
looking path. OTOH, it may be for PuTTY.
Groetjes,
Maarten Wiltink
-
Re: Can anyone interpret these loopstats?
Richard B. Gilbert wrote:
> David J Taylor wrote:
>> Folks,
>>
>> I've been monitoring my Windows Vista system using the Meinberg
>> distribution NTP, and it seems to have a peculiar behaviour of
>> alternate good and bad timekeeping:
>>
>> http://www.david-taylor.myby.co.uk/mrtg/gemini_ntp.html
>>
>> Can anyone who understands loopstats and peerstats better than I do
>> explain what's going on. The files are here:
>>
>> http://www.david-taylor.myby.co.uk/g...ista-stats.zip
>>
>> In the event log NTP is showing errors: "Unable to initialise the
>> .rnd file", but I don't think that's key to the operation of NTP
>> itself. The NTP service shows as running, and the W32time service
>> as stopped. Thanks,
>> Puzzled from Edinburgh (a.k.a. David)
>>
>>
>
> How about your ntp.conf? What does your ntpq -p banner look like?
>
> I would make a small bet that the network between you and your
> upstream servers is a real mess but looking only at your graphs
> doesn't tell me much more than that.
Richard, thanks for that, but the network is very clean. The upstream
server is a GPS-synced device, and the performance can be seen before the
Vista upgrade at the end of week 45. Very flat. The ntpd.conf is
essentially unaltered, and whnever I check the banner shows the PC as
synced to the GPS stratum one server.
> I'm rather puzzled by the timing of the relatively flat period on your
> daily graph. It seems that you are getting the best performance
> between about 11 AM and 9:30 PM. I would expect the network to be
> busiest during that period and consequently I would expect to see the
> worst performance during that period.
Yes, I don't think it's network related at all. To me, it seems almost
that there are two sync sources at work, and I'm seeing the beat between
them. But, no, the W32time service is stopped. It puzzled me, so I
asked. You can see the network load into that PC here:
http://www.david-taylor.myby.co.uk/mrtg/gemini.html
Normally, very light. The software tasks now running under Vista are
basically the same as those running under XP before, and the same versions
of the software.
> What happened on Friday of week 45? It looks as if that was about the
> time everything turned to worms!
That was when the PC had Vista installed (fresh, not upgrade). I was
hoping that Heiko might look at this, as he reported "good" performance
with Vista much earlier this year.
Cheers,
David
-
Re: Can anyone interpret these loopstats?
rtxo wrote:
[]
> Hello Devid--
>
> You could find a GNU/linux or freebsd system with openssl installed:
>
> # openssl rand 1024 > x
>
> ...then move the file "x" to someplace your "other os" ntp can find it
> with the name "dot rnd" (not sure if you can specify another name).
>
> Not sure about the data plot question. They don't look that bad to me,
> actually.
>
> //Steve
Thanks, Steve. I stole a .rnd file off one of my other systems and copied
it. We'll see if it makes any difference. I've just given the file full
access to everyone, as Windows Vista is very protective about programs
writing in the root directory of a hard disk. The Windows OpenSSL seems
to require C:\.rnd
Cheers,
David
-
Re: Can anyone interpret these loopstats?
Maarten Wiltink wrote:
[]
> There's an environment variable 'RANDFILE' on my system with a likely-
> looking path. OTOH, it may be for PuTTY.
>
> Groetjes,
> Maarten Wiltink
Thanks, Maarten. Nothing like that on Windows, it appears.
Cheers,
David
-
Re: Can anyone interpret these loopstats?
Maarten Wiltink wrote:
> "rtxo" wrote in message
> news:13kecijgavaarf8@corp.supernews.com...
>> David J Taylor wrote:
>
>>> In the event log NTP is showing errors: "Unable to initialise the .rnd
>>> file", but I don't think that's key to the operation of NTP itself.
>>> The NTP service shows as running, and the W32time service as stopped.
>
>> You could find a GNU/linux or freebsd system with openssl installed:
>>
>> # openssl rand 1024 > x
>>
>> ...then move the file "x" to someplace your "other os" ntp can find it
>> with the name "dot rnd" (not sure if you can specify another name).
>
> There's an environment variable 'RANDFILE' on my system with a likely-
> looking path. OTOH, it may be for PuTTY.
>
Yes but remember (if you knew) that on Windows and ntpd runs under it's
own account as a service. You need to get RANDFILE defined for the
account and it's not a account you can log onto. If you define it
globally, everyone gets it. The alternative is to put it in the home
directory of the account which looks something like X:\Documents and
Settings\ntpd\ where X: is the appropriate drive. YMMV.
Danny
> Groetjes,
> Maarten Wiltink
-
Re: Can anyone interpret these loopstats?
David J Taylor wrote:
> Thanks, Steve. I stole a .rnd file off one of my other systems and copied
> it. We'll see if it makes any difference. I've just given the file full
> access to everyone, as Windows Vista is very protective about programs
> writing in the root directory of a hard disk. The Windows OpenSSL seems
> to require C:\.rnd
The locations that it looks for the file are fully described in
html/hints/winnt.html
Danny
-
Re: Can anyone interpret these loopstats?
Danny Mayer wrote:
> David J Taylor wrote:
>> Thanks, Steve. I stole a .rnd file off one of my other systems and
>> copied it. We'll see if it makes any difference. I've just given
>> the file full access to everyone, as Windows Vista is very
>> protective about programs writing in the root directory of a hard
>> disk. The Windows OpenSSL seems to require C:\.rnd
>
> The locations that it looks for the file are fully described in
> html/hints/winnt.html
>
> Danny
Thanks, Danny. On the Vista system I'm looking at, neither RANDFILE nor
HOME are defined (at least for my user account), so C:\.rnd is what is
used. Would having the .RND file or not make any difference to the
performance of NTP?
By the way, with the Meinberg distribution, the page is here:
\NTP\doc\HTML\build\hints\winnt.html
Defining that file seems to have made no difference though - sometimes the
timekeeping is excellent, and sometimes awful. It doesn't seem to depend
on what I'm doing on the PC, and it has me really puzzled!
http://www.david-taylor.myby.co.uk/mrtg/gemini_ntp.html
Checking the ntpq -p output shows all servers at 377 reach, 1024s poll,
and my local GPS stratum 1 server as the preferred clock with just 1.5msec
delay.
Cheers,
David
-
Re: Can anyone interpret these loopstats?
"David J Taylor"
wrote in message news:5RI1j.52262$c_1.25697@text.news.blueyonder.co .uk...
> Maarten Wiltink wrote:
>> There's an environment variable 'RANDFILE' on my system with a likely-
>> looking path. OTOH, it may be for PuTTY.
> Thanks, Maarten. Nothing like that on Windows, it appears.
That _was_ on Windows.
In response to Danny's remark about the account the NTP daemon runs
under: yes, you're right and I hadn't spotted that - but there are
per-user and systemwide environment variables and this one was indeed
set systemwide to %systemroot%/system32/drivers/etc/.rnd.
Groetjes,
Maarten Wiltink
-
Re: Can anyone interpret these loopstats?
Maarten Wiltink wrote:
> "David J Taylor"
> wrote in
> message news:5RI1j.52262$c_1.25697@text.news.blueyonder.co .uk...
>> Maarten Wiltink wrote:
>
>>> There's an environment variable 'RANDFILE' on my system with a
>>> likely- looking path. OTOH, it may be for PuTTY.
>
>> Thanks, Maarten. Nothing like that on Windows, it appears.
>
> That _was_ on Windows.
>
> In response to Danny's remark about the account the NTP daemon runs
> under: yes, you're right and I hadn't spotted that - but there are
> per-user and systemwide environment variables and this one was indeed
> set systemwide to %systemroot%/system32/drivers/etc/.rnd.
>
> Groetjes,
> Maarten Wiltink
As I should have written "Nothing like that on my Windows, it appears".
But, yes, it's in the docs somewhere.
It begs the question: should the RANDFILE environment variable be
something which Meinberg's installer defines, if it is not otherwise set?
Cheers,
David
-
Re: Can anyone interpret these loopstats?
On Nov 23, 1:08 pm, "David J Taylor"
bit.nor-this-bit.co.uk> wrote:
> In the event log NTP is showing errors: "Unable to initialise the .rnd
> file", but I don't think that's key to the operation of NTP itself. The
> NTP service shows as running, and the W32time service as stopped.
>
> Thanks,
> Puzzled from Edinburgh (a.k.a. David)
What are the Power Options settings in Vista? Can you try creating a
custom power profile which prevents processor power management (in
Vista, every default power profile - even the "High Performance"
profile - seems to use processor frequency scaling.)
If things work fine with processor frequency scaling turned off, that
would be an interesting.
Besides processor power managmeent, the power management options for
the Indexing service as well as the multimedia power options could
potentially affect ntpd, I suppose. (The former because of lost
interrupts, and I know the Meinberg build uses multimedia timers on
windows).
-
Re: Can anyone interpret these loopstats?
Ryan Malayter wrote:
> On Nov 23, 1:08 pm, "David J Taylor"
> wrote:
>> In the event log NTP is showing errors: "Unable to initialise the
>> .rnd file", but I don't think that's key to the operation of NTP
>> itself. The NTP service shows as running, and the W32time service
>> as stopped.
>>
>> Thanks,
>> Puzzled from Edinburgh (a.k.a. David)
>
> What are the Power Options settings in Vista? Can you try creating a
> custom power profile which prevents processor power management (in
> Vista, every default power profile - even the "High Performance"
> profile - seems to use processor frequency scaling.)
>
> If things work fine with processor frequency scaling turned off, that
> would be an interesting.
>
> Besides processor power managmeent, the power management options for
> the Indexing service as well as the multimedia power options could
> potentially affect ntpd, I suppose. (The former because of lost
> interrupts, and I know the Meinberg build uses multimedia timers on
> windows).
Ryan,
Some interetsing points there - I've just changed to the High Performance
plan and, while there are no BIOS or Processor settings there (even with
the Show Me Everything enabled), the HD turn-off time was 20 minutes by
default. So I've changed that to Never.
Let's see what that does....
Thanks,
David
-
Re: Can anyone interpret these loopstats?
On Nov 26, 4:04 pm, "David J Taylor"
bit.nor-this-bit.co.uk> wrote:
> Ryan,
>
> Some interetsing points there - I've just changed to the High Performance
> plan and, while there are no BIOS or Processor settings there (even with
> the Show Me Everything enabled), the HD turn-off time was 20 minutes by
> default. So I've changed that to Never.
>
> Let's see what that does....
>
Hmm... every Vista PC I've encountered has processor power management
settings available, but they are horribly buried in the UI, as is the
new custom in Vista.
Go to Windows Menu->Control Panel->Power Options. Then click on
"change plan details" for any plan. Then click on "Change Advanced
Power Settings" hyperlink, and finally "change power settings that are
currently unavailable" hyperlink to edit items that require
administrative control. That will get you to a tree view of all power
management settings available on the PC. There should be an entry
called "Processor Power Management".
Of course, we only have Vista on newer hardware. If you don't see
those options, can I ask what sort of hardware are you using? Do you
have power management disabled in your PC's BIOS?
Regards,
Ryan
-
Re: Can anyone interpret these loopstats?
Ryan Malayter wrote:
> On Nov 26, 4:04 pm, "David J Taylor"
[]
> Hmm... every Vista PC I've encountered has processor power management
> settings available, but they are horribly buried in the UI, as is the
> new custom in Vista.
>
> Go to Windows Menu->Control Panel->Power Options. Then click on
> "change plan details" for any plan. Then click on "Change Advanced
> Power Settings" hyperlink, and finally "change power settings that are
> currently unavailable" hyperlink to edit items that require
> administrative control. That will get you to a tree view of all power
> management settings available on the PC. There should be an entry
> called "Processor Power Management".
>
> Of course, we only have Vista on newer hardware. If you don't see
> those options, can I ask what sort of hardware are you using? Do you
> have power management disabled in your PC's BIOS?
>
> Regards,
> Ryan
Ryan,
Thanks for your continuing interest. I've gone through that, including
the "unavailable" settings, and there's no "Processor power management" in
the tree. I'll try and remember to check the BIOS next time I need to
reboot, but I didn't set up this BIOS from scratch myself. It's a fairly
recent motherboard: ASUS A8N SLI Deluxe with an AMD 64 X2 4400+ processor.
Cheers,
David
-
Re: Can anyone interpret these loopstats?
Ryan,
Ryan Malayter wrote:
> On Nov 23, 1:08 pm, "David J Taylor"
> bit.nor-this-bit.co.uk> wrote:
>> In the event log NTP is showing errors: "Unable to initialise the .rnd
>> file", but I don't think that's key to the operation of NTP itself. The
>> NTP service shows as running, and the W32time service as stopped.
>>
>> Thanks,
>> Puzzled from Edinburgh (a.k.a. David)
>
> What are the Power Options settings in Vista? Can you try creating a
> custom power profile which prevents processor power management (in
> Vista, every default power profile - even the "High Performance"
> profile - seems to use processor frequency scaling.)
>
> If things work fine with processor frequency scaling turned off, that
> would be an interesting.
That's indeed an interesting point.
> Besides processor power managmeent, the power management options for
> the Indexing service as well as the multimedia power options could
> potentially affect ntpd, I suppose. (The former because of lost
> interrupts, and I know the Meinberg build uses multimedia timers on
> windows).
Please note that "the Meinberg build" is based on the official NTP sources.
The only thing the ntpd port for Windows does with the multimedia timers is
to set them optionally to highest resolution while the ntpd service is
running.
This is just to avoid virtual steps in the system time returned by the
Windows API calls, when another program changed the MM timer resolution,
which still happened on Vista when I tried last time.
This MM feature is only enabled if the "-M" flag is given on the command
line for ntpd. This can be enabled or disabled in the GUI setup during
installation, or simply by editing the ntpd's ImagePath registry key. So it
can easily be disabled to see whether the behaviour changes.
Martin
--
Martin Burnicki
Meinberg Funkuhren
Bad Pyrmont
Germany
-
Re: Can anyone interpret these loopstats?
Martin Burnicki wrote:
[]
> Please note that "the Meinberg build" is based on the official NTP
> sources.
>
> The only thing the ntpd port for Windows does with the multimedia
> timers is to set them optionally to highest resolution while the ntpd
> service is running.
>
> This is just to avoid virtual steps in the system time returned by the
> Windows API calls, when another program changed the MM timer
> resolution, which still happened on Vista when I tried last time.
>
> This MM feature is only enabled if the "-M" flag is given on the
> command line for ntpd. This can be enabled or disabled in the GUI
> setup during installation, or simply by editing the ntpd's ImagePath
> registry key. So it can easily be disabled to see whether the
> behaviour changes.
>
> Martin
Turning off the HD shutdown had no effect (the system is quite busy do HD
shotdown may never have been invoked), so I'll try removing the -M switch.
Martin, have you tested with different versions of Vista and found any
differences?
In particular, I'm using Vista Ultimate on my testbed right now, and there
may be differences between the firewall in the different Vista editions
(not yet sure this is true, though).
Cheers,
David