< news2mail | Russ Allbery > Software > INN > INN 2.5 Documentation | overchan > |
(Report Usenet Path: statistics (new inpaths))
ninpaths -p -d dumpfile
ninpaths -r site -u dumpfile [-u dumpfile ...] -v level
This is an efficient and space-saving inpaths reporting program. It works as follows: you feed it the Path: lines via an INN channel feed or some other similar method, and from time to time the program writes all its internal counters accumulated so far to a dump file. Another instance of the program picks up all the dump files, adds them up and formats them into the report. The purpose of the final report is to summarize the frequency of occurrence of sites in the Path: headers of articles.
Some central sites accumulate the Path: data from many news servers running this program or one like it, and then report statistics on the most frequently seen news servers in Usenet article Path: lines. The sendinpaths script can be run daily to mail the accumulated statistics to such a site and remove the old dump files.
You can get a working setup by doing the following:
Create a directory at pathlog/path (replacing pathlog here and in
all steps that follow with the full path to your INN log directory).
Do not change the name of the path
subdirectory because it is used
by sendinpaths.
Set up a channel feed using an entry like:
inpaths!:*:Tc,WP:<pathbin>/ninpaths -p -d <pathlog>/path/inpaths.%d
if your version of INN supports WP
(2.0 and later all do). Replace
<pathbin> with the full path to your INN binaries directory, and
<pathlog> with the full path to your INN log directory.
Note that the naming convention of the generated inpaths dump files should
not be changed. sendinpaths explicitly searches files whose name starts
with inpaths.
in the <pathlog>/path directory.
Run the following command to start logging these statistics:
ctlinnd reload newsfeeds 'inpaths feed setup'
Enter into your news user crontab these two lines:
6 6 * * * <pathbin>/ctlinnd flush inpaths! 10 6 * * * <pathbin>/sendinpaths
(the actual time doesn't matter). This will force ninpaths to generate a dump file once a day. Then, a few minutes later, sendinpaths collects the dumps, makes a report, sends the collected statistics, and deletes the old dumps.
Note that you can manually generate a report without mailing it, and without
deleting processed dump files, with sendinpaths -n
. Another useful
command is sendinpaths -c
so as to receive a copy of the e-mail sent
by sendinpaths and therefore make sure that everything is properly set.
In a couple of days, check that your daily statistics properly appear in <http://top1000.anthologeek.net/>.
Save dumps in dumpfile. Any %d
in dumpfile will be replaced with
the current system time when the dump is made. This option should be used
with -p.
The format of these dump files is described below.
Read Path: lines from standard input.
Generate a report for site. Generally site should be the value of pathhost from inn.conf.
Read data from dumpfile. This option can be repeated to read data from multiple dump files.
Set the verbosity level of the report. Valid values for level are 0
,
1
, and 2
, with 2
being the default.
The format of the generated dump files is:
!!NINP <version> <start-time> <end-time> <nb-sites> <nb-articles> <average-time> <site_0> <count_0> <site_1> <count_1> <site_2> <count_2> ... !!NLREC :<site_a>!<site_b>,<count_ab>:<site_c>!<site_d>,<count_cd> ... !!NEND <nb-relations>
where times are UNIX timestamps. Then, nb-sites records follow. Each record is separated by a space or a new line, and consists of a host name site_n followed by a number of appearances count_n. The number of processed Path: header lines is nb-articles.
Afterwards, nb-relations relations follow. In 3.0.x versions, the
relations are separated by a space or a new line, and their syntax is
site_a!site_b!count_ab
where site_a and site_b
are numbers of the site records starting at 0.
In 3.1.x versions, the relations begin with a colon and are
separated by either nothing or a new line. Their syntax is :site_a!site_b,count_ab
with the same meaning as in previous
versions. The count can be omitted when it is 1
. More than two sites can
be specified in the relation (:site_a!site_b!site_c,count_abc
).
For instance:
!!NINP 3.1.1 1302944821 1302944838 5 2 1302944826 newsgate.cistron.nl 1 news.trigofacile.com 2 news.ecp.fr 2 usenet.stanford.edu 1 bleachbot 1 !!NLREC :3!2:2!1,2:4!0:0!2 !!NLEND 4
where the two processed Path: headers are:
Path: news.trigofacile.com!news.ecp.fr!usenet.stanford.edu!not-for-mail Path: news.trigofacile.com!news.ecp.fr!newsgate.cistron.nl!bleachbot!not-for-mail
If your INN doesn't have the WP
feed flag (1.5 does not, 1.6 and 1.7 do,
2.0 and later all do), use the following newsfeeds entry:
inpaths!:*:Tc,WH:<pathbin>/ginpaths
where ginpaths is the following script:
#!/bin/sh exec egrep '^Path: ' | <pathbin>/ninpaths -p -d <pathlog>/path/inpaths.%d
replacing <pathbin> and <pathlog> as above.
This is a slightly modified version of Olaf Titz's original ninpaths program, which is posted to alt.sources and kept on his WWW archive under <http://sites.inka.de/~bigred/sw/>.
The idea and some implementation details for ninpaths come from the original inpaths program, but most of the code has been rewritten for clarity. This program is in the public domain.
$Id$
newsfeeds(5), sendinpaths(8).
< news2mail | Russ Allbery > Software > INN > INN 2.5 Documentation | overchan > |