xen-multi

#! /usr/bin/perl

# -*- perl -*-

=head1 NAME

xen_multi - Munin multigraph plugin to monitor Xen domains activity

=head1 APPLICABLE SYSTEMS

This plugin should work on any system running a Xen hypervisor and where xentop
is installed. It also needs Munin 1.4.0 or higher, since it uses AREASTACK
(available from 1.3.3) and multigraph (available from 1.4.0).

=head1 CONFIGURATION

xentop requires superuser privileges, so you need to include in your
configuration:

 [xen-multi]
 user root
 env.xenskip "<space separated module list>"

Modules: cput, cpup, mem, disk, net

Then restart munin-node and you're done.

=head1 INTERPRETATION

This plugin produces four different graphs: CPU usage, memory usage, disk IOs
and network traffic.

In each graph, all Xen domains (including dom0) have their data stacked, giving
an overall amount of resources used.

NOTE: xentop always reports 0 for dom0's disk IOs and network traffic, but
both graphs show its entry all the same, so each domain can keep its own color
along the different graphs.

=head2 CPU time

This graph shows a percentage of the CPU time used by each domain.

=head2 CPU usage

This graph shows a percentage of the CPU percentage used by each domain.

=head2 Memory usage

This graph shows the amount of memory (in bytes) used by each domain.

=head2 Disk IOs

This graph shows the number of disk read and write operations for each domain.

=head2 Network traffic

This graph shows the amount of bits received and transmitted for each domain.

=head1 ACKNOWLEDGEMENTS

Michael Renner for the C<diskstats> plugin which I borrowed some code from.

=head1 VERSION

0.90

=head1 MAGIC MARKERS

  #%# family=auto
  #%# capabilities=autoconf

=head1 AUTHOR

Raphael HALIMI <raphael.halimi@gmail.com>
Modified by Krisztian IVANCSO <github-ivan@ivancso.net>

=head1 LICENSE

GPLv2

=cut

use strict;
use Munin::Plugin;

my $XEN_SKIP = $ENV{xenskip} || "";

# autoconf - quite simple
if ($ARGV[0] eq "autoconf") {
  if (`which xentop`) {
    print "yes\n";
  } else {
    print "no (xentop not found)\n";
  }
  exit 0;
}


#
# Common steps for both "config" and a normal run
#

#
# trim_label
#
# Trims a given label to it's non-wrapping size
# $type = 'pos' for normal graphs and 'neg' for mirror graphs
# pos: nnn.nnU_ times 4
# neg: nnn.nnU/nnn.nnU_ times 4

sub trim_label {
  my ($type, $label) = @_; my $data_characters;
  my ($graph_width, $graph_border_width, $padding_characters, $pixels_per_character) = (497,3,5,6);
  if ($type eq 'pos') {$data_characters = 32;} elsif ($type eq 'neg') {$data_characters = 64;} else {return $label;}

  my $available_characters = abs(($graph_width+$graph_border_width)/$pixels_per_character)-$padding_characters-$data_characters;

  # If the label is longer than the available space, we write as many
  # characters as we can on both left and right, and two dots in the middle
  if ( $available_characters < length $label ) {
    my $center = ($available_characters-2)/2;
    $label = substr($label,0,($center)) . '..' . substr($label,-$center);
  }

  return $label;
}

# Global variables
my (%domains,$domain,@domainlist,$munindomain,$cpusecs,$cpupercent,$memk,$nettxk,$netrxk,$vbdrd,$vbdwr);

# There is a bug in xentop, its output is too high in rare cases.
# https://lists.xenproject.org/archives/html/xen-users/2019-04/msg00020.html
# Workaround to process 3 iterations after the first one (4 iterations) and
# choose the lowest cpusecs value line.

open (XENTOP,"xentop -b -f -i4 |") or die "Could not execute xentop, $!";

# Now we build a hash of hashes to store information
while (<XENTOP>) {
  # Some cleaning first
  s/^\s+//; chomp;
  # Skip the headers
  next if /^NAME/;

  # We define only what we need
  ($domain,undef,$cpusecs,$cpupercent,$memk,undef,undef,undef,undef,undef,$nettxk,$netrxk,undef,undef,$vbdrd,$vbdwr,undef,undef,undef) = split(/\s+/);

  # We have to store the sanitised domain name in a separate variable
  $domains{$domain}{'munindomain'} = clean_fieldname($domain);

  # We need the remaining data only for a normal run
  if ($ARGV[0] eq "") {
    # the cnt key counts the iterations
    # - skip the first one because xentop gives 0% cpu for the first iteration ( cnt > 1 )
    # - process if not processed yet ( !defined ) or if current cpusecs less than any earlier
    # normally the cpusecs is monotonous per domain as it grows so a smaller value means that the previous one was wrong

    $domains{$domain}{'cnt'}++;
    if ( $domains{$domain}{'cnt'} > 1 && ( !defined $domains{$domain}{'cpusecs'} || $cpusecs < $domains{$domain}{'cpusecs'} ) ) {
      $domains{$domain}{'cpusecs'} = $cpusecs;
      $domains{$domain}{'cpupercent'} = $cpupercent;
      $domains{$domain}{'mem'} = $memk;
      $domains{$domain}{'nettx'} = $nettxk;
      $domains{$domain}{'netrx'} = $netrxk;
      $domains{$domain}{'vbdrd'} = $vbdrd;
      $domains{$domain}{'vbdwr'} = $vbdwr;
    }
  }
}

@domainlist = sort(keys(%domains));

#
# config - quite simple, too
#

if ($ARGV[0] eq "config") {
  if ($XEN_SKIP !~ /cput/) {
		print "multigraph xen_cpu_time\n";
		print "graph_title Xen domains CPU time\n";
		print "graph_args --base 1000 -l 0\n";
		print "graph_vlabel cpu seconds\n";
		print "graph_scale no\n";
		print "graph_category virtualization\n";
		print "graph_info This graph shows CPU time for each Xen domain.\n";
		for $domain (@domainlist) {
			print "$domains{$domain}{'munindomain'}_cpu_time.label ".trim_label('pos',$domain)."\n";
			print "$domains{$domain}{'munindomain'}_cpu_time.type DERIVE\n";
			print "$domains{$domain}{'munindomain'}_cpu_time.cdef $domains{$domain}{'munindomain'}_cpu_time,100,*\n";
			print "$domains{$domain}{'munindomain'}_cpu_time.min 0\n";
			print "$domains{$domain}{'munindomain'}_cpu_time.draw AREASTACK\n";
		}
  }

  if ($XEN_SKIP !~ /cpup/) {
		print "\nmultigraph xen_cpu\n";
		print "graph_title Xen domains CPU utilization\n";
		print "graph_args --base 1000 -l 0 --upper-limit 100\n";
		print "graph_vlabel %\n";
		print "graph_scale no\n";
		print "graph_category virtualization\n";
		print "graph_info This graph shows CPU utilization for each Xen domain.\n";
		for $domain (@domainlist) {
			print "$domains{$domain}{'munindomain'}_cpu.label ".trim_label('pos',$domain)."\n";
			print "$domains{$domain}{'munindomain'}_cpu.draw AREASTACK\n"
		}
  }

  if ($XEN_SKIP !~ /mem/) {
		print "\nmultigraph xen_mem\n";
		print "graph_title Xen domains memory usage\n";
		print "graph_args --base 1024 -l 0\n";
		print "graph_vlabel bytes\n";
		print "graph_category virtualization\n";
		print "graph_info This graph shows memory usage for each Xen domain.\n";
		for $domain (@domainlist) {
			print "$domains{$domain}{'munindomain'}_mem.label ".trim_label('pos',$domain)."\n";
			print "$domains{$domain}{'munindomain'}_mem.cdef $domains{$domain}{'munindomain'}_mem,1024,*\n";
			print "$domains{$domain}{'munindomain'}_mem.draw AREASTACK\n";
		}
  }

  if ($XEN_SKIP !~ /net/) {
		print "\nmultigraph xen_net\n";
		print "graph_title Xen domains network traffic\n";
		print "graph_args --base 1000\n";
		print "graph_vlabel bits per \${graph_period} in (-) / out (+)\n";
		print "graph_category virtualization\n";
		print "graph_info This graph shows network traffic for each Xen domain.\n";
		for $domain (@domainlist) {
			print "$domains{$domain}{'munindomain'}_netrx.label none\n";
			print "$domains{$domain}{'munindomain'}_netrx.cdef $domains{$domain}{'munindomain'}_netrx,8192,*\n";
			print "$domains{$domain}{'munindomain'}_netrx.type COUNTER\n";
			print "$domains{$domain}{'munindomain'}_netrx.graph no\n";
			print "$domains{$domain}{'munindomain'}_nettx.label ".trim_label('neg',$domain)."\n";
			print "$domains{$domain}{'munindomain'}_nettx.cdef $domains{$domain}{'munindomain'}_nettx,8192,*\n";
			print "$domains{$domain}{'munindomain'}_nettx.type COUNTER\n";
			print "$domains{$domain}{'munindomain'}_nettx.draw AREASTACK\n";
			print "$domains{$domain}{'munindomain'}_nettx.negative $domains{$domain}{'munindomain'}_netrx\n";
		}
  }

  if ($XEN_SKIP !~ /disk/) {
		print "\nmultigraph xen_disk\n";
		print "graph_title Xen domains disk IOs\n";
		print "graph_args --base 1000\n";
		print "graph_vlabel IOs per \${graph_period} read (-) / write (+)\n";
		print "graph_category virtualization\n";
		print "graph_info This graph shows disk IOs for each Xen domain.\n";
		for $domain (@domainlist) {
			print "$domains{$domain}{'munindomain'}_vbdrd.label none\n";
			print "$domains{$domain}{'munindomain'}_vbdrd.type COUNTER\n";
			print "$domains{$domain}{'munindomain'}_vbdrd.graph no\n";
			print "$domains{$domain}{'munindomain'}_vbdwr.label ".trim_label('neg',$domain)."\n";
			print "$domains{$domain}{'munindomain'}_vbdwr.type COUNTER\n";
			print "$domains{$domain}{'munindomain'}_vbdwr.draw AREASTACK\n";
			print "$domains{$domain}{'munindomain'}_vbdwr.negative $domains{$domain}{'munindomain'}_vbdrd\n";
		}
  }

  exit 0;
}


#
# Normal run
#

if ($XEN_SKIP !~ /cput/) {
	print "multigraph xen_cpu_time\n";
	for $domain (@domainlist) {
		print "$domains{$domain}{'munindomain'}_cpu_time.value $domains{$domain}{'cpusecs'}\n";
	}
}

if ($XEN_SKIP !~ /cpup/) {
	print "\nmultigraph xen_cpu\n";
	for $domain (@domainlist) {
		print "$domains{$domain}{'munindomain'}_cpu.value $domains{$domain}{'cpupercent'}\n";
	}
}

if ($XEN_SKIP !~ /mem/) {
	print "\nmultigraph xen_mem\n";
	for $domain (@domainlist) {
		print "$domains{$domain}{'munindomain'}_mem.value $domains{$domain}{'mem'}\n";
	}
}

if ($XEN_SKIP !~ /net/) {
	print "\nmultigraph xen_net\n";
	for $domain (@domainlist) {
		print "$domains{$domain}{'munindomain'}_nettx.value $domains{$domain}{'nettx'}\n";
		print "$domains{$domain}{'munindomain'}_netrx.value $domains{$domain}{'netrx'}\n";
	}
}

if ($XEN_SKIP !~ /disk/) {
	print "\nmultigraph xen_disk\n";
	for $domain (@domainlist) {
		print "$domains{$domain}{'munindomain'}_vbdrd.value $domains{$domain}{'vbdrd'}\n";
		print "$domains{$domain}{'munindomain'}_vbdwr.value $domains{$domain}{'vbdwr'}\n";
	}
}