Note to Debian users: if you have installed torrus using Debian packages, all the needed dependencies outlined below have been satisfied and installation tasks carried out by the Debian package management system. The information below is provided for information purposes only and may not apply to the Debian torrus installations. Please see the README.Debian file in /usr/share/doc/torrus-common for more details.
Torrus should easily run on any UNIX-like operation system, as long as the required components are provided.
If the OS mounts /var/run directory to a temporary filesystem, piddir
variable must point to a path on a persistent storage. Alternatively
piddir can point to /var/run itself. The default path is
/var/run/torrus, and temporary filesystems do not preserve subdirectories
after a system reboot.
Note to VmWare users: as recommended by Gord Philpott, Linux kernel for
guest machine needs to be compiled with the following feature enabled:
Character devices ---> Enhanced Real Time Clock Support.
For other guest operating systems (e.g. FreeBSD), in order to let the
Torrus collector and monitor run properly, you need to add the following
lines to your torrus-siteconfig.pl:
$Torrus::Scheduler::maxSleepTime = 15; $Torrus::Scheduler::ignoreClockSkew = 1;
# Debian package is maintained by Jurij Smakov and is available at # <http://pkg-torrus.alioth.debian.org>.
Perl version 5.8.1 or higher is required. Version 5.8.0 is not
recommended because of memory access bugs and unpredictable behaviour.
Perl interpreter should be in PATH environment variable when
running ./configure, otherwise use
./configure PERL=/path/to/perl
Torrus also supports Perl multithreading in Device Discovery engine and
in the Collector. Minimum requirements are as follows: Perl version 5.8.8
or higher, threads module version 1.41 or higher, threads::shared module
version 1.03 or higher. Older versions had bugs that were leading to
severe memory leaks and instability. After installing Perl 5.8.8,
you need to upgrade the threads and threads::shared modules manually
from CPAN. If any of the requirements is not met, the installer will
automatically disable the threads support in Torrus.
Round Robin Database Tool is available at:
<http://ee-staff.ethz.ch/~oetiker/webtools/rrdtool/>.
Both stable version 1.0.x and the development version 1.1.x
are supported.
As of this writing, rrdtool version 1.1.x is still in its development
stage, but it is quite ready for using in production environments.
There's a small glitch in legend text alignment, and it will
hopefully be fixed.
libxml2 is the XML parser library from Gnome
project <http://www.gnome.org>,
available as a standalone package.
Versions 2.4.23 and above were tested.
There are two distinct packages with slightly conflicting names:
The Berkeley DB from Sleepycat Software <http://www.sleepycat.com>,
and BerkeleyDB Perl module available from CPAN. We will always refer
the latter as BerkeleyDB Perl module.
The Berkeley DB database engine is used for all data storage.
Versions 4.2.52 and above are recommended.
Older versions of Berkeley DB had problems with database integrity
in concurrent usage. By default, it installs
into /usr/local/BerkeleyDB.4.2/.
Unfortunately, the BerkeleyDB Perl module cannot find the include and
library files in that directory. Thus either configure Berkeley DB
with the option --prefix=/usr/local, or you need to edit
config.in in BerkeleyDB Perl module distribution.
Torrus requires an HTTP server. It is compatible with the following combinations:
FastCGI is supported in Torrus version 1.0.9 or higher. It is the preferred way of using Torrus, more stable than mod_perl.
Note: Starting from Torrus version 1.0.9, libapreq2 is no longer
required.
See Torrus Web Interface Reference for details on HTTP server configuration.
The Perl modules required are listed below. They are all available at CPAN <http://www.cpan.org>. Some of them require other modules to be installed. You can install all required modules with the following command, executed from within the distribution directory:
perl -I `pwd`/setup_tools -MCPAN -e 'install Bundle::Torrus'
This is a Perl interface for libxml2, providing DOM standards compliant
interface. It is recommended that you use version 1.54_3 or higher.
Perl interface to the database software.
Version 0.19 or higher is required.
This is a powerful set of tools for text template processing. Torrus uses it for HTML files output. See also the project homepage: <http://www.template-toolkit.org>
Daemon invocation routines.
SNMP queries and traps interface. Version 4.0.3 or higher is required.
Note: Torrus versions prior to 1.0.9 are incompatible with Net::SNMP
version 6.0.0. You may need to downgrate the module to 5.2.0.
Escape and unescape unsafe characters
Mod_perl is a Perl runtime environment integrated into Apache HTTP server.
User session tracking helper
Module for parsing the user input for date and time.
JSON data format support
The HTTP browser must be compatible with HTML 4.01 Strict and CSS2 specifications. The tests were made with the following browsers: IE 5.5, Opera 6.12B1/Solaris, Opera 7.03/Win2K, Netscape 7.02.
XML::XUpdate::LibXML by Petr Pajas
This Perl module implements XUpdate specification <http://www.xmldb.org/xupdate/>. See Torrus User Guide for details.
Available at <http://xsh.sourceforge.net>. This is a shell wrapper for a set of utilities for XML extraction, browsing, and editing.
libxslt is the XSLT stranslation library from Gnome
project <http://www.gnome.org>, available as a standalone package.
It includes a binary executable, xsltproc. See Torrus User Guide for
examples of its usage.
Create the group torrus and a user torrus. The user should be a member of
this group.
Add your Apache daemon user to the group torrus.
Other users that will run any of Torrus processes must be included in
this group.
For example, in Solaris (you may need to specify the GID and UID numbers according to your local administration policy):
groupadd torrus useradd -c 'Torrus Daemon' -d /usr/local/torrus -g torrus torrus usermod -G www,torrus www
Further installation process is mostly as usual (watch out piddir setting,
as described above):
./configure make make install
The default directory layout is described below. Should you need to change it,
there is a number of variables and configuration options that you may use
as ./configure arguments. See ./configure --help for details.
Alternatively you can utilize the script ./setup_tools/configure_fhs,
which is designed to provide an FHS compliant setup. The script is equivalent
to executing
./configure \ --prefix=/opt \ --mandir=/opt/share/man \ pkghome=/opt/torrus \ sitedir=/etc/opt/torrus
Do not try to change any paths by supplying them as make variables,
this would most probably break the installation. The only make variable
that is supported is DESTDIR. It may be used for preparing the package for
further distribution. For example, the following command would install
all Torrus files as if /tmp/stage were the root of the filesystem:
make DESTDIR=/tmp/stage install
The presence of prerequisite Perl modules is checked during the execution
of ./configure. You can disable this by giving --enable-pkgonly option.
The installer sets up the site configuration files, but only if they don't already exist.
Plugin modules should be installed separately, after the Torrus software is successfully installed. For each plugin, unpack the plugin distribution archive to some directory, and execute
torrus install_plugin <UNPACKED_PLUGIN_DIR>
A number of directories is set up by default under the path /var,
and they must be writable by all Torrus processes, including the
Apache web server.
You can control these directories' access rights by setting the following environment variables: var_user, var_group, var_mode, like follows:
./configure var_group=wwwrun
Default values for operating systems other than Cygwin are: var_user=torrus, var_group=torrus, var_mode=775. Setgid bit is set by default for these directories.
If available, you may place these directories on a fast media, for example,
a robust disk array. Changing the varprefix variable would be generally
enough, for example:
./configure varprefix=/vol1/torrus_var
/usr/share/
Home directory for Torrus distribution files
/usr/share/torrus/conf_defaults/
torrus-config.pl and other configuration files
/usr/share/torrus/bin/
Command-line executables
@docdir@/
POD and TXT documentation files
/usr/share/doc/torrus-common/examples/
Miscelaneous example files
/usr/share/perl5/
Perl libraries
/usr/share/torrus/plugins/
Plugins configuration
/usr/share/torrus/scripts/
Scripts
/usr/share/torrus/sup/
Supplementary files, DTDs, MIBs, color schemas, web plain files
/usr/share/torrus/templates/
Renderer output templates
/usr/share/torrus/xmlconfig/
Distrubution XML files
/etc/torrus/
Site configurable files
/etc/torrus/conf/
Place for torrus-siteconfig.pl and other siteconfigs
/etc/torrus/conf/discovery/
Devdiscover input files
/etc/torrus/templates/
User-defined Renderer output templates
/etc/torrus/xmlconfig/
User XML configuration files
/usr/share/man
Place for man pages. All articles have the prefix C<torrus_>
/var/log/torrus/
Daemon logfiles
/var/run/torrus
Daemon PID files
/var/cache/torrus
Renderer cache
/var/lib/torrus/db
Berkeley DB home
/var/lib/torrus/session_data/lock
/var/lib/torrus/session_data/store
Web interface session files
/var/lib/torrus/collector_rrd
Recommended directory for RRD files generated by collectors
Note: RRFW used the path /var/snmpcollector as the default path for
storing the RRD files. According to FHS specification, the recommended
path for service data files is recommended to be a subdirectory of /srv.
The path /var/lib/torrus/collector_rrd is used as default by torrus genddx
utility, and you can always change it in the command line or in your
DDX files.
The datasources are configured with %Torrus::Global::treeConfig
hash in torrus-siteconfig.pl.
In this hash, the keys give the tree names. The values for each tree name are pointers to hashes, with the following keys and values: xmlfiles points to an array of source XML file names; run points to a hash with the names of the daemons that would be automatically launched for the tree; desription gives a short line describing the tree contents.
Two additional arrays: @Torrus::Global::xmlAlwaysIncludeFirst and
@Torrus::Global::xmlAlwaysIncludeLast give a list of source XML
files that are included in every tree, in the beginning or in the end of
the XML files list. By default, this array consists of two files:
defaults.xml and site-global.xml. The second one is resided in
the user-configurable directory, and you can use it to override any
default settings.
Example:
@Torrus::Global::xmlAlwaysIncludeFirst =
('defaults.xml', 'site-global.xml');
%Torrus::Global::treeConfig = (
'tree_A' => {
'description' => 'The First Tree',
'xmlfiles' => [qw(a1.xml a2.xml a3.xml)],
'run' => { 'collector' => 1, 'monitor' => 1 } },
'tree_B' => {
'description' => 'The Second Tree',
'xmlfiles' => ['b1.xml', 'b2.xml'],
'run' => {} }
);
XML files are read additively within each tree, in the order
as they are listed. The XML compiler searches for the files in several
directories, listed in @Torrus::Global::xmlDirs. By default, this list
conssists of two paths, one for Torrus distribution files, and the other
for user files.
Files generated by devdiscover usually contain include statements
which add the vendor-specific XML files to the compilation.
Below follows a short description of some XML files that come with Torrus distribution. Only site-global.xml is installed in the directory for user-configurable files, all others are placed in the distribution directory.
Default parameters for the root of the datasources tree.
Default view definitions. Note: this file is automatically
overwritten by make install.
Parameters that you want to change or define for your site.
This file will be compiled for every tree after defaults.xml,
and this is the place to override the defaults. The file that is supplied
with the Torrus distribution has some useful parameters that you may simply
uncomment.
Note: this file is never overwritten by make install.
SNMP collector defaults. The file defines several templates used for collecting SNMP data. Do not change this file. You may redefine the needed parameters in your site-specific files and templates.
SNMP collector definitions and templates for various hardware vendors and
products. devdiscover includes some of these files automatically in the
configuration.
SNMP collector definitions and templates for vendor-independent MIBs. Most files are named after corresponding RFC numbers.
In addition, the distribution package contains several example files.
For more details about XML configuration, see Torrus User Guide and Torrus XML Configuration Guide.
In addition to %Torrus::Global::treeConfig, you may wish to set some other parameters in your site configuration file (torrus-siteconfig.pl).
See torrus-config.pl for the complete list of varaibes that you may override in your site config. Among them, most interesting are:
$Torrus::Renderer::companyName
The text that you specify here will appear in the top left corner of all HTML pages.
$Torrus::Renderer::companyURL
The company name text will be clickable with the URL specified in this variable.
Torrus web interface is designed to run under mod_perl environment (<http://perl.apache.org>).
See the Torrus Web Interface Reference document for detailed instructions on Apache configuration.
In short, typical Apache configuration for mod_perl version 1.0 would look like follows:
Alias /torrus/plain "/usr/share/torrus/sup/webplain"
PerlRequire "/usr/share/torrus/conf_defaults/webmux.pl"
<Location /torrus>
SetHandler perl-script
PerlHandler Torrus::ApacheHandler
</Location>
<Location /torrus/plain/>
SetHandler default-handler
Options None
</Location>
By default, Torrus web interface requires user authentication.
You can disable this by changing $Torrus::CGI::authorizeUsers
to zero in your torrus-siteconfig.pl.
ACLs are controlled by acledit utility. See Torrus Manual pages
for detailed description. Example:
torrus acledit --addgroup=staff --permit=DisplayTree --for='*'
torrus acledit --addgroup=staff --permit=GlobalSearch --for='*'
torrus acledit --adduser=jsmith --password=mysecretpassword \
--cn="John Smith" --addtogroup=staff
torrus acledit --addgroup=admin \
--permit=DisplayTree --permit=DisplayAdmInfo --for='*'
In order to clean old HTTP session data, it is recommended to run cleanup script in a cron job, once per day:
#min hour mday month wday who command 5 3 * * * root /usr/share/torrus/bin/cleanup
The Torrus distriubution provides a startup script which you can install in
the init scripts directory (/etc/init.d on most Unixes or /usr/local/etc/rc.d/
on FreeBSD). The script torrus is created during the installation
process in the subdirectory <init.d> of the distribution directory.
The init script reads some parameters from /usr/share/torrus/conf_defaults/initscript.conf, and /etc/torrus/conf/initscript.siteconf if the latter exists. By default, it makes the monitor daemons sleep for 20 minutes when they are launched simultaneously with collector daemons.
Follow these instructions when upgrading from previous Torrus release.
In the previous distribution directory, look up the config.log file. It contains the configure options that you used in previous installation.
Unpack the new release distribution.
Run ./configure with the same options you used before.
Stop the collector and monitor processes
(usually by /etc/init.d/torrus stop).
Stop Apache process.
Run make install.
Start the collector and monitor processes.
Start Apache process.
Also it is recommended to re-compile your XML configuration. If you prefer not to do this, it's recommended that you clear the Web interface cache both in your browser and in Torrus:
torrus clearcache
As of Torrus version 2.01, plugins inherit a new release numbering scheme: a plugin release version is a 3-component number, where the first two numbers are the Torrus release version that is compatible with this plugin.
In a typical Torrus setup, there are two disk space consuming entities: the RRDtool data storage and the Berkeley DB database.
RRDtool data files (or RRD's) are used to store the values that are gathered by the collector daemons. These are the most intensively updated files, so the disk speed is important here. If possible, it is better to spread the RRD files across several physical disks.
Berkeley DB database is used to keep all the configuratiuon data for your datasource trees, and it also keeps some live status information. It's intensively updated during XML compilation only. During normal Torrus working cycle, there are only few updates, not very critical in time. The database is read quite intensively during collector initialization, but usually it's the CPU speed that causes more delay.
For a typical Torrus installation with standard RRD creation parameters, the rough estimation is as follows: the Berkeley DB database occupies from 10 to 13 KB per collector datasource, and the RRD storage takes approximately 80 to 85 KB per datasource.
For a medium-sized system, few hundred megabytes for the Berkeley DB database and several gigabytes for RRD storage would be sufficient. Larger systems require a thorough design and staging work. See Torrus Scalability Guide for more details.
In large installations, the SNMP collector experiences quite intensive
input traffic bursts. Thus the UDP socket receive buffers should
be adapted to sustain the load. By default, Torrus sets the UDP receiving
buffer size, SO_RCVBUF, to 131071 bytes. This should fit most of
installations. However, it's useful to check the network statistics
if there are any UDP buffer overflows. On most systems, the commands
netstat -s -p udp or netstat -s should show you these counters.
The maximum buffer size is usually limited by a system kernel variable,
and can be increased if needed. See $Torrus::Collector::SNMP::RxBuffer
and its comments in torrus-config.pl for more details.
Copyright (c) 2002-2007 Stanislav Sinyagin <ssinyagin@yahoo.com>