faraphel/chrony
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
864 commits 8 branches 75 tags 4.3 MiB
Commit graph

55 commits

Author SHA1 Message Date
Miroslav Lichvar
2c033989b6 Update comments referencing RFC 1305 2014-10-14 16:52:22 +02:00
Miroslav Lichvar
a0d5abef88 sourcestats: remove tracking of skew change 
This is not used since commit 7a6ee1d.
 2014-09-19 10:07:03 +02:00
Miroslav Lichvar
75efa5174c Convert disabled log messages to debug or remove them 2014-06-27 10:17:35 +02:00
Miroslav Lichvar
cb74f3e7ad Update copyright years 2014-06-05 18:06:56 +02:00
Miroslav Lichvar
13d734c8d2 sourcestats: reset SST instance thoroughly when dropping samples 2014-06-02 16:46:53 +02:00
Miroslav Lichvar
be3439fef1 sourcestats: assert dump file is loaded with no accumulated samples 2014-04-29 12:07:38 +02:00
Miroslav Lichvar
0b71504ee9 sourcestats: fix signedness in scanf format 2014-03-21 13:20:24 +01:00
Miroslav Lichvar
1c3aff37de Convert TRACEON LOG messages to DEBUG_LOG 2013-11-27 14:35:41 +01:00
Miroslav Lichvar
922e2fe23b Fix Clang static analyzer warnings about never read values 2013-10-07 17:50:27 +02:00
Miroslav Lichvar
f8d609fee5 Add minimum skew limit to sourcestats 2013-06-19 10:22:49 +02:00
Miroslav Lichvar
6d2fb9f782 Add minsamples and maxsamples directives 
Allow configuration of the maximum and minimum number of samples per
source.
 2013-06-13 16:23:32 +02:00
Miroslav Lichvar
22e5ed44c2 Modify SST_GetSelectionData to return only necessary data 2013-06-12 16:06:33 +02:00
Miroslav Lichvar
ff8e04f9ba Fix fabs use on delay 2013-06-12 15:30:28 +02:00
Miroslav Lichvar
3eff836b2e Set stratum from last sample instead of best 2013-06-03 18:57:54 +02:00
Miroslav Lichvar
2b9fe764d5 Drop unused SST_GetReferenceData() 2013-06-03 16:03:07 +02:00
Miroslav Lichvar
359d444343 Remove unncessary return statements 2013-05-21 15:08:34 +02:00
Miroslav Lichvar
0c738d84af Update copyright years 2012-02-28 13:11:56 +01:00
Miroslav Lichvar
b088b70f82 Check sample ordering on accumulation 
If the newly accumulated sample is not newer than than the last one,
discard the source history and start from scratch. This can happen after
loading an invalid dump or when the system clock was stepped.
 2011-08-31 18:36:10 +02:00
Miroslav Lichvar
e0c9ed44f9 Limit skew used in NTP test4 
With iburst and very jittery sources the source skew can reach very high
values which makes the NTP test4 fail even with relatively small delays.
Limit the skew to 2000 ppm to avoid getting state where a source is unable
to accept more than first three iburst samples.
 2011-08-12 15:37:58 +02:00
Miroslav Lichvar
91279a0f28 Store reference IDs in uint32_t 2011-06-13 15:34:16 +02:00
Miroslav Lichvar
da2c8d9076 Use config.h 2011-05-24 18:07:06 +02:00
Miroslav Lichvar
bb40f4aff4 Modify weight calculation again 
Dividing the weights by variance or unweighted variance seems to have a
significant negative impact on response with normally distributed
network delays.

Divide by the difference between the mean and minimum distance instead.
It should be stable as there is no loop and the response seems to be a
good compromise between the original minimum distance weighting which
works well with normally distributed delays and the variance weighting
which works well with exponentially distributed delays.
 2011-04-29 13:29:56 +02:00
Miroslav Lichvar
66c7ac4d24 Revert using unweighted variance in weight calculation 
This reverts commit 165e6805ab.
 2011-04-29 13:29:24 +02:00
Miroslav Lichvar
165e6805ab In weight calculation use unweighted variance from last regression 
This fixes a positive feedback where weights could reach inf.
Also change the SD_TO_DIST_RATIO constant to get close to the original
response.
 2011-04-13 18:43:25 +02:00
Miroslav Lichvar
c5587b60b2 Set reference time to last sample instead of time on update 
This is done mainly to fix reported root dispersion to include max clock
error after selecting another source without new sample.
 2011-04-11 17:52:04 +02:00
Miroslav Lichvar
faec23f6bd Don't update empty sourcestats on clock update 2011-04-05 16:32:50 +02:00
Miroslav Lichvar
896dad9224 Fix warnings produced by latest gcc 2011-02-15 18:55:34 +01:00
Miroslav Lichvar
e8c5d15690 Remove CVS headers 2011-01-28 12:56:09 +01:00
Miroslav Lichvar
e63cba05b2 Update copyright 2011-01-27 13:05:26 +01:00
Miroslav Lichvar
98f79404d6 Allow selecting source only when last regression was successful 2010-12-15 17:48:30 +01:00
Miroslav Lichvar
c88801065f Divide regression weights by stddev instead of minimum distance 
This improves accuracy with large but stable network delays.
 2010-12-14 17:00:53 +01:00
Miroslav Lichvar
b977c95be4 Add test for ratio of increase in delay to stddev 
Require that the ratio of the increase in delay from the minimum one in
the stats data register to the standard deviation of the offsets in the
register is less than maxdelaydevratio or the difference between
measured offset and predicted offset is larger than the increase in
delay. In the allowed delay increase is included also skew and maximum
clock frequency error.

maxdelaydevratio is 10.0 by default.
 2010-12-07 16:47:57 +01:00
Miroslav Lichvar
feb8811f37 Cache minimum peer delay in sourcestats 2010-12-03 17:25:50 +01:00
Miroslav Lichvar
0de82a70a6 Keep only absolute values of peer delays in sourcestats 
This saves some fabs() calls.
 2010-12-03 13:17:09 +01:00
Miroslav Lichvar
7a6ee1d729 Base poll adjustment on number of sourcestats samples 
Instead of following skew changes, adjust polling interval so that the
number of measurements used in the regression algorithm remains close to
a target value. It can be configured with a new polltarget option
(6 by default).
 2010-10-13 16:49:28 +02:00
Miroslav Lichvar
061d497df0 Fix crash when reloading history with zero samples 2010-08-25 18:32:40 +02:00
Miroslav Lichvar
e591e3622b Extend runs test 
Double the number of samples that are used in the runs test. E.g. with 64
samples in regression the runs test will be tried over the 64 samples and
up to 64 previous samples. The minimum number of samples in now 4.

This improves the response with low-mid jitters by about 50%.
 2010-08-17 16:40:25 +02:00
Miroslav Lichvar
eeb73b3670 Fix updating of best_single_sample 2010-08-16 16:44:49 +02:00
Miroslav Lichvar
a3d47ffc81 Store sourcestats samples in circular buffer 
The samples now don't have to be moved when pruning the register.
 2010-08-10 16:39:52 +02:00
Miroslav Lichvar
d841c86a6e Cleanup sourcestats code a bit 2010-08-10 15:35:17 +02:00
Miroslav Lichvar
7ba6b617a1 Remove weights from sourcestats record 
Weights are calculated before each regression call, no need to store them.
 2010-08-06 14:39:09 +02:00
Miroslav Lichvar
100f732e20 Remove SST_DoUpdateRegression 
The function is not used anywhere and it requires weights to be stored
sourcestats.
 2010-08-06 14:36:56 +02:00
Miroslav Lichvar
40d82675bd Make use of UTI_AdjustTimeval in slew handlers 2010-05-07 18:52:05 +02:00
Miroslav Lichvar
e78e65ef22 Refactor file logging 2010-04-13 15:16:41 +02:00
Miroslav Lichvar
0f70959d8e Add dispersion notification handlers 2010-02-16 18:51:37 +01:00
Miroslav Lichvar
b32432c232 Move estimated offset and error to sourcestats report 
And print the estimated offset in sourcestats output.
 2009-12-14 10:34:33 +01:00
Miroslav Lichvar
2dd9f3373b Replace integer microseconds in reports with floating-point values 2009-12-13 15:00:10 +01:00
Miroslav Lichvar
5b1a8705cf Remove resid_freq and resid_skew from source report 
They were not printed and they are also in sourcestats report.
 2009-12-13 11:07:45 +01:00
Miroslav Lichvar
0f9892fe7a Fix printing refclocks and IPv6 sources in statistics log 2009-11-30 17:18:28 +01:00
Miroslav Lichvar
8e23110aec Update COPYING and FSF address 2009-10-28 17:53:33 +01:00