[ntp:questions] peerstat update frequency

unruh unruh at invalid.ca
Sat Mar 24 17:37:06 UTC 2012


On 2012-03-24, unruh <unruh at invalid.ca> wrote:
> On 2012-03-24, A C <agcarver+ntp at acarver.net> wrote:
>> On 3/23/2012 20:48, unruh wrote:
>>> On 2012-03-24, A C<agcarver+ntp at acarver.net>  wrote:
>>>> How often should an individual peer write an entry to the peerstat log?
>>>>    Is it supposed to occur every time the peer is polled or only once
>>>> every certain number of polls?
>>> Every time it is polled. (ntpd writes to the file, the peer does not do
>>> so.)
>>
>> Well that is precisely what is not happening.
>
> I think you should do this one letter at a time, rather than one
> sentence at a time. Each time someone replies you can release another
> letter of the full description of the problem. Keeps up the tension. 
>
> Why in the world do you not, when asking the question, give the full
> context for your question.
>
> Here is the end of my peerstats file for one particular server at a time
> when the poll time was 128 sec
>
> 56010 58994.647 128.233.154.245 9414 -0.000049195 0.020862666 0.005473533 0.000024479
> 56010 59057.647 128.233.154.245 9414 -0.000044852 0.020865740 0.005094022 0.000023695
> 56010 59121.647 128.233.154.245 9414 -0.000041340 0.020855152 0.002255272 0.000020397
> 56010 59251.647 128.233.154.245 9414 -0.000041340 0.020855152 0.004163787 0.000024439
> 56010 59381.647 128.233.154.245 9414 -0.000041340 0.020855152 0.005339178 0.000033796
>
> Note that the time between polls at the end  is about 130 sec.
> (The earlier 67 sec was presumably because the poll interval was 6 then,
> not the 7 of later.)
>
>
>

Look at the source code in 4.2.4 and 4.2.6, yes, the peerstats reporting
was pulled into clock_filter, and the peerstats were only only reported
if it was a selected peer measurement in the filter in 4.2.6. In 4.2.4
all were reported, which makes much more sense to me. 

Thus in 4.2.6 only about 20% of the measurements were reported. 
Makes it much harder to use peerstats to debug network problems, etc. 




More information about the questions mailing list