Torrus Reporting Setup Guide

In version 1.0.4, Torrus introduces a new and important functionality. Now it is able to produce reports that may be used in billing or resource planning. The reports are meant to be text output, telling the bandwidth usage or volume: no graphs so far.

By default, the reporting functionality is not enabled, and the steps required are described below.

Terminology

The basic term in Torrus reporting is the Service ID. It denotes a single datasource that is represented by a single number. For example, if you want to count input and output interface traffic, this would make two different service IDs. Do not assign service IDs to each and every interface in your network, as it would degrade the performance of your Torrus installation significantly. This functionality is designed for use with important datasources, such as your customers' connection links or and ISP's upstream channels.

Each service ID must be unique across the whole Torrus installation, and across the database that stores them (it is possible to use several Torrus installations with the same database).

Devdiscover, the SNMP discovery engine, allows now to assign service IDs to network interfaces of your SNMP-enabled devices. However, this does not prevent you from assigning them to other Torrus datasources, as it's done by simple configuration parameters.

Report, when generated, is presented by a set of numeric values in the SQL database. Torrus provides also tools for rendering these values into HTML output. In the future, other output formats will be implemented. You can also build your own infrastructure that reads the values from the reports database.

The produced reports may, and are primarily developed for using in billing process. As it is stated in the GNU General Public License text, this program is provided "as is" and without warranty of any kind. It is up to the users of Torrus software to rely or not to rely on the generated numeric data, and the Torrus software developers disclaim any responsibility for the data accuracy and usability.

New in version 1.0.7: You can assign the list of trees where the reports should be generated and shown. Warning: after changing the destination tree, the compiler may complain about duplicate service IDs. Then you need to stop all the torrus processes, including Apache, and then remove the file serviceid_params.db from Torrus database directory, then recompile the trees and start the processes.

Install Perl modules

Install the following Perl modules from CPAN:

  DBI
  DBD::mysql or other RDBMS driver
  DBIx::Abstract
  DBIx::Sequence
  Excel::Writer::XLSX

On many platforms, DBI is already pre-installed. You need to make sure that appropriate DBD driver is installed for your database engine. The setup was tested with MySQL, SQL syntax is compatible with Postgres, and in theory it should run also with Oracle and probably Sybase or DB2. No idea about MSSQL - if you're brave enough, let me know if it works :)

Enable the External Storage and specify the SQL connection

In torrus-siteconfig.pl, add the following lines. The first one tells the collector to use the module for storing the collector results in SQL database. Next lines define the database connection. By default, it refers to the MySQL database named torrus on localhost, with username and password set to imiF1oih.

  push(@Torrus::Collector::loadModules, 'Torrus::Collector::ExternalStorage');
  $Torrus::SQL::connections{'Default'}{'dsn'} =
      'DBI:mysql:database=torrus;host=dbhost.example.com';
  $Torrus::SQL::connections{'Default'}{'username'} = 'torrus';
  $Torrus::SQL::connections{'Default'}{'password'} = 'imiF1oih';

In addition to the default connection, you can specify different connections for different data structures: for example, keep 5-minutes samples on a bulky storage server, and store the reports on a high-availability cluster. See the comments in torrus-config.pl for more details.

Create SQL tables

With your RDBMS client, create the following tables. We assume here that the same database is used for all tables. The SQL syntax is verified with MySQL 4.x and Postgres 8.x. It is brought as much as possible to the standard and platform-independent SQL syntax. Please let me know if it causes problems with your RDBMS.

 /* Collector export table. It usually grows at several megabytes
    per month, and is updated every 5 minutes */
 CREATE TABLE srvexport
 (
  srv_date DATE NOT NULL,            /* date and time of the data sample */
  srv_time TIME NOT NULL,  
  serviceid VARCHAR(64) NOT NULL,    /* unique service ID per counter */
  value DOUBLE PRECISION NOT NULL,   /* collected rate or gauge value */
  intvl INTEGER NOT NULL             /* collection interval -
                                        for counter volume calculation */
  );
 CREATE INDEX srvexp_date ON srvexport (srv_date);
 CREATE INDEX srvexp_srvid ON srvexport (serviceid);
 

 /* Tables for (currently monthly only) report contents.
    These are updated usually once per month, and read at the moment of
    rendering the report output (HTML now, PDF or XML or Excel or whatever
    in the future) */

 /* DBIx::Sequence backend, theplatform-independent inplementation
    of sequences */
 CREATE TABLE dbix_sequence_state
 (
  dataset varchar(50) NOT NULL, 
  state_id INTEGER NOT NULL, 
  CONSTRAINT pk_dbix_sequence PRIMARY KEY (dataset, state_id)
  );

 CREATE TABLE dbix_sequence_release
 (
  dataset varchar(50) NOT NULL,    
  released_id INTEGER NOT NULL, 
  CONSTRAINT pk_dbi_release PRIMARY KEY (dataset, released_id)
 );


 /* Each report is characterized by name, date and time.
    Monthly reports are automatically assigned 00:00 of the 1st day
    in the month. The report contains fields for every service ID
    defined across all datasource trees. */
 CREATE TABLE reports
 (
  id INTEGER PRIMARY KEY,
  rep_date DATE NOT NULL,             /* Start date of the report */
  rep_time TIME NOT NULL,             /* Start time of the report */
  reportname VARCHAR(64) NOT NULL,    /* Report name, such as MonthlyUsage */
  iscomplete INTEGER NOT NULL,        /* 0 when the report is in progress, */
                                      /* 1 when it is ready */
  UNIQUE(rep_date, rep_time, reportname)
  );
 CREATE INDEX rep_date_idx ON reports (rep_date);


 /* Each report contains fields. For each service ID,
    the report may contain several fields for various statistics.
    Each field contains information about the units of the value it
    contains */
 CREATE TABLE reportfields
 (
  id INTEGER PRIMARY KEY,
  rep_id INTEGER,
  name VARCHAR(64) NOT NULL,       /* name of the field, such as AVG or MAX */
  serviceid VARCHAR(64) NOT NULL,           /* service ID */
  value DOUBLE PRECISION NOT NULL,          /* Numeric value */
  units VARCHAR(64) NOT NULL DEFAULT '',    /* Units, such as bytes or Mbps */
  UNIQUE (rep_id, name, serviceid)
  );

Devdiscover parameters

Currently devdiscover allows you to assign service IDs to network interfaces' input and output traffic counters. Other types of datasources may be implemented in the future.

Torrus XML configuration parameters

You can skip this section if Devdiscover provides all desired functionality. It explains parameters additional to those described in Torrus XML Configuration Guide.

The collector's ExternalStorage module is designed for storing the data to a generic external storage, and the default external storage is the SQL database.

Enable displaying of the reports in the web interface

First, enable the reports displaying in <torrus-siteconfig.pl>:

 $Torrus::Renderer::displayReports = 1;

Second, configure the ACL for your users that are allowed to see the reports in the web interface:

 torrus acl --modgroup=admin --permit=DisplayReports --for=mytree

In this example, members of the group admin will be prompted with the [Reports] shortcut in the web interface's bottom of the page for a given tree. For easier setup, the tree name may be replaced with asterisk (*) to allow this option for all trees.

Generate reports

Report generation is usually a CPU-intensive task. A monthly report calculation for one service ID may take several dozens of seconds of CPU time. This uit's recommended to use the nice command to lower the process priority.

The torrus genreport (or simply torrus report) command-line utility is designed for two different tasks: calculation of a single report for the period chosen, and generation of output HTML for all reports available.

The example below generates the monthly usage report for September 2005:

 nice torrus report --report=MonthlyUsage --date=2005-09-01 --debug

The next example generates HTML output for all reports that are found in the database:

 nice torrus report --genhtml --tree=customers

The following command generates yearly reports in Excel format:

 nice torrus report --genxlsx --tree=customers

It makes sence to set up a monthly cron job and generate the reports on the first day of every month, with the command like follows:

 nice torrus report --report=MonthlyUsage \
   --date=`perl -e 'use Date::Format;    
         print time2str("%Y-%m-%d", time()-1728000)'`

The HTML output is optimized for printing, and is quite easy to navigate. The overview page provides the list of years. Clicking to the year leads you into the list of monthly reports available. Each monthly report consists of a table for each report name. For MonthlyUsage, the data is organized in five columns: the service ID, average monthly rate, 95th percentile of the rates, maximum rate throughout the month, unavailable samples (how many 5-minutes intervals are missed in the collected data), and monthly volume (which is roughly less than the actual volume by the percentage of missed samples). Also clicking a service ID leads to its monthly report throughout the year.

Getting the sum or maximum value from several service IDs

A new utility has been sponsored by nexellent ag (www.nexellent.ch), a managed hosting solution provider near Zurich.

The utility srvderive can be used to generate a new service ID which would contain sum or maximum of values of other service IDs. You would usually run this utlity monthly, just before generating the monthly reports:

  torrus srvderive --verbose --start=20080801 --month \
    --out=JSMITH_SUM_OUT --func=SUM JSMITH01_OUT JSMITH02_OUT

  torrus srvderive --verbose --start=20080801 --month \
    --out=JSMITH_SUM_IN --func=SUM JSMITH01_IN JSMITH02_IN

Copyright (c) 2005-2008 Stanislav Sinyagin <ssinyagin@k-open.com>