IPA_CONV(8)IPA_CONV(8)NAMEipa_conv-- IPA accounting/database module for converting statistics
DESCRIPTIONipa_conv is an IPA accounting/database module with the following fea‐
tures:
- The module is designed for converting statistics;
- Statistics is converted according to custom plans, several plans
can be used at the same time;
- Each plan can have converting rules for months, days, week days
and time ranges;
- Statistics of one static rule can be converted and sent to
another static rule.
The name of the accounting system and the name of the database is conv.
DESIGN OVERVIEWipa_conv consists of one accounting module and one database module, but
these two modules are implemented in one object file. This file must
be loaded two times in the ipa.conf(5) configuration file, one time
using the ac_mod parameter and second time using the db_mod parameter.
Actually only a single copy of an object file is brought into the
address space of ipa(8) and the module checks this at the time of ini‐
tialization.
An arbitrary static rule can use this module as a database (in this
case this rule is called ``source rule'') and/or as an accounting sys‐
tem (in this case this rule is called ``destination rule''). Each
source rule is the source of statistics for some one destination rule.
Each source rule has a converting plan for converting statistics for
corresponding destination rule. One destination rule can receive con‐
verted statistics from several source rules.
When the database part of the module gets statistics for the source
rule it converts this statistics according to the converting plan and
sends converted statistics to the destination rule. The destination
rule receives converted statistics from the accounting part of the mod‐
ule. In other words, the module works like a pipe between two rules,
it converts statistics from one rule and sends converted statistics to
another rule.
A converting plan consists of time intervals with information how to
convert original statistics. This information contains two numbers,
the first number X is the amount of original statistics and the second
number Y is the amount of corresponding converted statistics: full X
units of the original statistics are converted to Y units (X > 0 and
Y >= 0). By default the module scales X and Y values to the minimum
possible values, but this feature can be disabled.
There are several formats of time intervals (given below). The data‐
base part of the module always chooses the most specific time interval.
The most specific time interval is chosen when the database part of the
module gets updated statistics for some source rule. If the value of
statistics was s1 at t1 time and the value of statistics is s2 at t2
time for some source rule and there were not database updates between
t1 and t2 of this rule, then the time interval is chosen for t2 time
and the difference (s2 - s1) is converted and sent to the destination
rule (this difference can be negative). Too long time intervals
between database updates can cause inaccurate converting of statistics.
Usually a source rule should be written above the corresponding desti‐
nation rule in the configuration file and it is better to tell ipa(8)
to not change internally their order (use the keep_rules_order parame‐
ter). If a source rule is processed first and the corresponding desti‐
nation rule is processed second at the same time, then the destination
rule will receive converted statistics from a source rule immediately.
Otherwise it will receive converted statistics during next database
update time.
This module works with increasing statistics and decreasing statistics.
But since a database module receives only non-negative statistics, then
this module cannot handle correctly situation when statistics of some
rule became negative (in this case the module's database part receives
zero statistics from ipa(8)). If this statistics becomes positive
again, then the module will work as expected.
CONFIGURATION FILE FORMAT
One part of the configuration for the module is integrated into
ipa.conf(5), other parts of the configuration are kept in files with
converting plans. The configuration prefix of this module is ``conv''.
The conv:scale parameter specifies whether the module should scale val‐
ues in all loaded converting plans:
conv:scale = <boolean>;
This parameter should not be placed in any section and should be used
before conv:load parameters. The default value of this parameter is
``yes'' and the module scales values in all plans.
The conv:load parameter loads a file with converting plan and gives a
local name to loaded plan:
conv:load = <plan-name> "/path/file";
This parameter should not be placed in any section and it should be
used before conv:plan parameters. The given file is loaded and parsed
when the module parses this parameter.
The conv:dest parameter specifies the destination rule name, the
conv:plan parameter specifies the converting plan for this rule:
conv:dest = <rule-name>;
conv:plan = <plan-name>;
These two parameters can be placed in the rule section. The module
checks whether the given destination rule exists at the end of configu‐
ration file parsing. If a destination rule is inactive, then it does
not received converted statistics.
The conv:quiet parameter allows to make the module less or more ver‐
bose:
conv:quiet = <value>;
This parameter should not be placed in any section. The default value
of this parameter is 3. If its value has 0x1 bit off, then the module
sends a warning message when ipa(8) deinitializes some destination rule
and this rule still have some converted statistics. If its value has
0x2 bit off, then the module sends a warning message when ipa(8) deini‐
tializes some source rule and this rule still have some not converted
original statistics.
The conv:debug parameter helps to debug the module:
conv:debug = <level>;
This parameter should not be placed in any section. Only two values
are allowed for the debug level: 0 and 1. The default value of this
parameter is 0 and debugging is disabled.
CONVERTING PLAN FORMAT
A file with converting plan consists of lines. Each line should have
not more than 255 characters. Each line can have a shell-like comment
that starts with the '#' character. Empty lines or lines with only
comments are ignored.
A converting plan for one time interval should be written in one line
containing three fields. The first field is a time interval, the sec‐
ond field is the amount of original statistics and the third field is
the amount of corresponding converted statistics. These fields are
separated by space or tab characters.
A time interval can be given in different formats, but time intervals
of one format should not overlap. Here is the list of time interval
formats (time intervals types are checked exactly in the same order
when the most specific time interval is being searched):
o year.mon.day/h1:m1-h2:m2 specifies time interval for year, month
and day.
o year.mon.day specifies year, month and day.
o mon.day/h1:m1-h2:m2 specifies time interval for month and day.
o mon.day specifies month and day.
o mon specifies a month.
o X/h1:m1-h2:m2 specifies time interval for a week day. Valid
values for X are: `S' Sunday, `M' Monday, `T' Tuesday, `W'
Wednesday, `H' Thursday, `F' Friday, `A' Saturday.
o X specifies a week day
o h1:m1-h2:m2 specifies time interval.
o `*' specifies the default settings and must exist in any plan.
A month can be given as the first three characters of English language
month name.
A day starts at 00:00 and ends at 24:00.
Example:
There is one rule r1 that gets statistics from some accounting system.
It is necessary to get only 50% of that statistics at night from 00:00
till 07:00 and statistics should not be changed at other time.
Let's create a file with converting plan /etc/ipa/plan/night:
# Interval Original Converted
* 1 1 # 100%
00:00-07:00 2 1 # 50% (night)
Let's create ipa.conf, rule r2 will get converted statistics from rule
r1:
/* ... */
ac_mod "ipa_conv.so";
db_mod "ipa_conv.so";
conv:load = night "/etc/ipa/plan/night";
rule r1 {
/* ... */
db_list = conv;
conv:dest = r2;
conv:plan = night;
}
rule r2 {
/* ... */
ac_list = conv;
}
Rule r1 can use several databases, rule r2 can use databases and
another accounting systems. The accounting system used by rule r1 is
not shown in this example.
AUTHOR
Andrey Simonenko <simon@comsys.ntu-kpi.kiev.ua>
BUGS
If you find any, please send email me.
July 22, 2009 IPA_CONV(8)
