NAME

Sys::Statistics::Linux - Front-end module to collect system statistics

SYNOPSIS

    use Sys::Statistics::Linux;

    my $lxs = Sys::Statistics::Linux->new(
        sysinfo   => 1,
        cpustats  => 1,
        procstats => 1,
        memstats  => 1,
        pgswstats => 1,
        netstats  => 1,
        sockstats => 1,
        diskstats => 1,
        diskusage => 1,
        loadavg   => 1,
        filestats => 1,
        processes => 1,
    );

    sleep 1;
    my $stat = $lxs->get;

DESCRIPTION

Sys::Statistics::Linux is a front-end module and gather different linux system information like processor workload, memory usage, network and disk statistics and a lot more. Refer the documentation of the distribution modules to get more information about all possible statistics.

MOTIVATION

My motivation is very simple... every linux administrator knows the well-known tool sar of sysstat. It helps me a lot of time to search for system bottlenecks and to solve problems, but it's hard to parse the output if you want to store the statistics into a database. So I thought to develope Sys::Statistics::Linux. It's not a replacement but it should make it simpler to you to write your own system monitor.

If Sys::Statistics::Linux doesn't provide statistics that are strongly needed then let me know it.

TECHNICAL NOTE

This distribution collects statistics by the virtual /proc filesystem (procfs) and is developed on the default vanilla kernel. It is tested on x86 hardware with the distributions RHEL, Fedora, Debian, Ubuntu, Asianux, Slackware, Mandriva and openSuSE (SLES on zSeries as well but a long time ago) on kernel versions 2.4 and/or 2.6. It's possible that it doesn't run on all linux distributions if some procfs features are deactivated or too much modified. As example the linux kernel 2.4 can compiled with the option CONFIG_BLK_STATS what turn on or off block statistics for devices.

Don't give up if some of the modules doesn't run on your hardware! Tell me what's wrong and I will try to solve it! You just have to make the first move and to send me a mail. :-)

VIRTUAL MACHINES

Note that if you try to install or run Sys::Statistics::Linux under virtual machines on guest systems that some statistics are not available, such as SockStats, PgSwStats and DiskStats. The reason is that not all /proc data are passed to the guests.

If the installation fails then try to force the installation with

    cpan> force install Sys::Statistics::Linux

and notice which tests fails, because this statistics maybe not available on the virtual machine - sorry.

DELTAS

The statistics for CpuStats, ProcStats, PgSwStats, NetStats, DiskStats and Processes are deltas, for this reason it's necessary to initialize the statistics before the data can be prepared by get(). These statistics can be initialized with the methods new(), set() and init(). For any option that is set to 1, the statistics will be initialized by the call of new() or set(). The call of init() re-initialize all statistics that are set to 1 or 2. By the call of get() the initial statistics will be updated automatically. Please refer the section "METHODS" to get more information about the usage of new(), set(), init() and get().

Another exigence is to sleep for a while - at least for one second - before the call of get() if you want to get useful statistics. The statistics for SysInfo, MemStats, SockStats, DiskUsage, LoadAVG and FileStats are no deltas. If you need only one of these information you don't need to sleep before the call of get().

The method get() prepares all requested statistics and returns the statistics as a Sys::Statistics::Linux::Compilation object. The inital statistics will be updated.

MANUAL PROC(5)

The Linux Programmer's Manual

    http://www.kernel.org/doc/man-pages/online/pages/man5/proc.5.html

If you have questions or don't understand the sense of some statistics then take a look into this awesome documentation.

OPTIONS

All options are identical with the package names of the distribution in lowercase. To activate the gathering of statistics you have to set the options by the call of new() or set(). In addition you can deactivate statistics with set().

The options must be set with one of the following values:

    0 - deactivate statistics
    1 - activate and init statistics
    2 - activate statistics but don't init

In addition it's possible to pass a hash reference with options.

    my $lxs = Sys::Statistics::Linux->new(
        processes => {
            init => 1,
            pids => [ 1, 2, 3 ]
        },
        netstats => {
            init => 1,
            initfile => $file,
        },
    );

Option initfile is useful if you want to store initial statistics on the filesystem.

    my $lxs = Sys::Statistics::Linux->new(
        cpustats => {
            init     => 1,
            initfile => '/tmp/cpustats.yml',
        },
        diskstats => {
            init     => 1,
            initfile => '/tmp/diskstats.yml',
        },
        netstats => {
            init     => 1,
            initfile => '/tmp/netstats.yml',
        },
        pgswstats => {
            init     => 1,
            initfile => '/tmp/pgswstats.yml',
        },
        procstats => {
            init     => 1,
            initfile => '/tmp/procstats.yml',
        },
    );

Example:

    #!/usr/bin/perl
    use strict;
    use warnings;
    use Sys::Statistics::Linux;

    my $lxs = Sys::Statistics::Linux->new(
        pgswstats => {
            init => 1,
            initfile => '/tmp/pgswstats.yml'
        }
    );

    $lxs->get(); # without to sleep

The initial statistics are stored to the temporary file:

    #> cat /tmp/pgswstats.yml
    --- 
    pgfault: 397040955
    pgmajfault: 4611
    pgpgin: 21531693
    pgpgout: 49511043
    pswpin: 8
    pswpout: 272
    time: 1236783534.9328

Every time you call the script the initial statistics are loaded/stored from/to the file. This could be helpful if you doesn't run it as daemon and if you want to calculate the average load of your system since the last call. Do you understand? I hope so :)

To get more information about the statistics refer the different modules of the distribution.

    sysinfo     -  Collect system information              with Sys::Statistics::Linux::SysInfo.
    cpustats    -  Collect cpu statistics                  with Sys::Statistics::Linux::CpuStats.
    procstats   -  Collect process statistics              with Sys::Statistics::Linux::ProcStats.
    memstats    -  Collect memory statistics               with Sys::Statistics::Linux::MemStats.
    pgswstats   -  Collect paging and swapping statistics  with Sys::Statistics::Linux::PgSwStats.
    netstats    -  Collect net statistics                  with Sys::Statistics::Linux::NetStats.
    sockstats   -  Collect socket statistics               with Sys::Statistics::Linux::SockStats.
    diskstats   -  Collect disk statistics                 with Sys::Statistics::Linux::DiskStats.
    diskusage   -  Collect the disk usage                  with Sys::Statistics::Linux::DiskUsage.
    loadavg     -  Collect the load average                with Sys::Statistics::Linux::LoadAVG.
    filestats   -  Collect inode statistics                with Sys::Statistics::Linux::FileStats.
    processes   -  Collect process statistics              with Sys::Statistics::Linux::Processes.

METHODS

new()

Call new() to create a new Sys::Statistics::Linux object. You can call new() with options. This options would be passed to the method set().

Without options

    my $lxs = Sys::Statistics::Linux->new();

Or with options

    my $lxs = Sys::Statistics::Linux->new( cpustats => 1 );

Would do nothing

    my $lxs = Sys::Statistics::Linux->new( cpustats => 0 );

It's possible to call new() with a hash reference of options.

    my %options = (
        cpustats => 1,
        memstats => 1
    );

    my $lxs = Sys::Statistics::Linux->new(\%options);

set()

Call set() to activate or deactivate options.

The following example would call new() and initialize Sys::Statistics::Linux::CpuStats and delete the object of Sys::Statistics::Linux::SysInfo.

    $lxs->set(
        processes =>  0, # deactivate this statistic
        pgswstats =>  1, # activate the statistic and calls new() and init() if necessary
        netstats  =>  2, # activate the statistic and call new() if necessary but not init()
    );

It's possible to call set() with a hash reference of options.

    my %options = (
        cpustats => 2,
        memstats => 2
    );

    $lxs->set(\%options);

get()

Call get() to get the collected statistics. get() returns a Sys::Statistics::Linux::Compilation object.

    my $lxs  = Sys::Statistics::Linux->new(\%options);
    sleep(1);
    my $stat = $lxs->get();

Or you can pass the time to sleep with the call of get().

    my $stat = $lxs->get($time_to_sleep);

Now the statistcs are available with

    $stat->cpustats

    # or

    $stat->{cpustats}

Take a look to the documentation of Sys::Statistics::Linux::Compilation for more information.

init()

The call of init() initiate all activated statistics that are necessary for deltas. That could be helpful if your script runs in a endless loop with a high sleep interval. Don't forget that if you call get() that the statistics are deltas since the last time they were initiated.

The following example would calculate average statistics for 30 minutes:

    # initiate cpustats
    my $lxs = Sys::Statistics::Linux->new( cpustats => 1 );

    while ( 1 ) {
        sleep(1800);
        my $stat = $lxs->get;
    }

If you just want a current snapshot of the system each 30 minutes and not the average then the following example would be better for you:

    # do not initiate cpustats
    my $lxs = Sys::Statistics::Linux->new( cpustats => 2 );

    while ( 1 ) {
        $lxs->init;              # init the statistics
        my $stat = $lxs->get(1); # get the statistics
        sleep(1800);             # sleep until the next run
    }

If you want to write a simple command line utility that prints the current workload to the screen then you can use something like this:

    my @order = qw(user system iowait idle nice irq softirq total);
    printf "%-20s%8s%8s%8s%8s%8s%8s%8s%8s\n", 'time', @order;

    my $lxs = Sys::Statistics::Linux->new( cpustats => 1 );

    while ( 1 ){
        my $cpu  = $lxs->get(1)->cpustats;
        my $time = $lxs->gettime;
        printf "%-20s%8s%8s%8s%8s%8s%8s%8s%8s\n",
            $time, @{$cpu->{cpu}}{@order};
    }

settime()

Call settime() to define a POSIX formatted time stamp, generated with localtime().

    $lxs->settime('%Y/%m/%d %H:%M:%S');

To get more information about the formats take a look at strftime() of POSIX.pm or the manpage strftime(3).

gettime()

gettime() returns a POSIX formatted time stamp, @foo in list and $bar in scalar context. If the time format isn't set then the default format "%Y-%m-%d %H:%M:%S" will be set automatically. You can also set a time format with gettime().

    my $date_time = $lxs->gettime;

Or

    my ($date, $time) = $lxs->gettime();

Or

    my ($date, $time) = $lxs->gettime('%Y/%m/%d %H:%M:%S');

EXAMPLES

A very simple perl script could looks like this:

    use strict;
    use warnings;
    use Sys::Statistics::Linux;

    my $lxs = Sys::Statistics::Linux->new( cpustats => 1 );
    sleep(1);
    my $stat = $lxs->get;
    my $cpu  = $stat->cpustats->{cpu};

    print "Statistics for CpuStats (all)\n";
    print "  user      $cpu->{user}\n";
    print "  nice      $cpu->{nice}\n";
    print "  system    $cpu->{system}\n";
    print "  idle      $cpu->{idle}\n";
    print "  ioWait    $cpu->{iowait}\n";
    print "  total     $cpu->{total}\n";

Set and get a time stamp:

    use strict;
    use warnings;
    use Sys::Statistics::Linux;

    my $lxs = Sys::Statistics::Linux->new();
    $lxs->settime('%Y/%m/%d %H:%M:%S');
    print $lxs->gettime, "\n";

If you want to know how the data structure looks like you can use Data::Dumper to check it:

    use strict;
    use warnings;
    use Sys::Statistics::Linux;
    use Data::Dumper;

    my $lxs = Sys::Statistics::Linux->new( cpustats => 1 );
    sleep(1);
    my $stat = $lxs->get;

    print Dumper($stat);

How to get the top 5 processes with the highest cpu workload:

    use strict;
    use warnings;
    use Sys::Statistics::Linux;

    my $lxs = Sys::Statistics::Linux->new( processes => 1 );
    sleep(1);
    my $stat = $lxs->get;
    my @top5 = $stat->pstop( ttime => 5 );

BACKWARD COMPATIBILITY

The old options and keys - CpuStats, NetStats, etc - are still available but deprecated! It's not possible to access the statistics via Sys::Statistics::Linux::Compilation and it's not possible to call search() and psfind() if you use the old options.

You should use the new options and access the statistics over the accessors

    $stats->cpustats

or directly with

    $stats->{cpustats}

PREREQUISITES

    Carp
    POSIX
    Test::More
    Time::HiRes
    UNIVERSAL

EXPORTS

No exports.

TODOS

   * Are there any wishs from your side? Send me a mail!

REPORTING BUGS

Please report all bugs to <jschulz.cpan(at)bloonix.de>.

AUTHOR

Jonny Schulz <jschulz.cpan(at)bloonix.de>.

COPYRIGHT

Copyright (C) 2006-2008 by Jonny Schulz. All rights reserved.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.