1#!/usr/local/bin/perl
2# -*- perl -*-
3#
4# proc_ - Munin plugin to for Process information
5# Copyright (C) 2009  Redpill Linpro AS
6# Copyright (C) 2010  Trygve Vea
7#
8# Author: Kristian Lyngstøl <kristian@redpill-linpro.com>
9# Author: Trygve Vea <tv@redpill-linpro.com>
10#
11# This program is free software; you can redistribute it and/or modify
12# it under the terms of the GNU General Public License as published by
13# the Free Software Foundation; either version 2 of the License, or
14# (at your option) any later version.
15#
16# This program is distributed in the hope that it will be useful,
17# but WITHOUT ANY WARRANTY; without even the implied warranty of
18# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19# GNU General Public License for more details.
20#
21# You should have received a copy of the GNU General Public License along
22# with this program; if not, write to the Free Software Foundation, Inc.,
23# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24
25=head1 NAME
26
27proc_ - Munin plugin to monitor various aspects of named processes
28
29=head1 APPLICABLE SYSTEMS
30
31Processes running under Linux
32
33=head1 CONFIGURATION
34
35The plugin needs to be able to parse the /proc-filesystem.
36
37The configuration section shows the defaults
38  [proc_*]
39     env.procname init
40     env.category Process Info
41
42env.procname defines the processname as seen inside the parenthesis of the
43second column in /proc/pid/stat. If you don't get the data you expect, you
44can check if the value is what you expect here.
45
46env.category is used to override the default category the plugin will show
47up in.
48
49=head1 INTERPRETATION
50
51Each graph uses data from the proc filesystem.
52
53=head1 MAGIC MARKERS
54
55 #%# family=auto
56 #%# capabilities=autoconf suggest
57
58=head1 VERSION
59
60 $Id$
61
62=head1 BUGS
63
64The CPU usage graph will be misleading in an event where you have multiple
65processes monitored, but less then all of them is restarted (or exits). This
66is due to the nature of counters, and I need to track state of individual
67processes to do this in a reliable way. It's on my TODO.
68
69=head1 PATCHES-TO
70
71Dunno.
72
73=head1 AUTHOR
74
75Kristian Lyngstøl <kristian@redpill-linpro.com>
76Trygve Vea <tv@redpill-linpro.com>
77
78=head1 THANKS
79
80Thanks to Kristian Lyngstøl, I stole most of the code in this plugin from his
81varnish_-plugin, which is a really nice outline of how a wildcardplugin should
82look like.
83
84=head1 LICENSE
85
86GPLv2
87
88=cut
89
90use strict;
91
92# Set to 1 to enable output when a variable is defined in a graph but
93# omitted because it doesn't exist in varnishstat.
94my $DEBUG = 0;
95
96# Set to 1 to ignore 'DEBUG' and suggest all available aspects.
97my $FULL_SUGGEST = 0;
98
99# You should set the env-var "procname" to filter processes of their name.
100# This will default to "init" unless you specify anything else.
101my $procname = exists $ENV{'procname'} ? $ENV{'procname'} : "init";
102
103# You can set the env-var "procargs" to filter processes of their running
104# arguments.
105my $args = exists $ENV{'procargs'} ? $ENV{'procargs'} : undef;
106
107# You can set the env-var "category" to override the default category.
108my $category = exists $ENV{'category'} ? $ENV{'category'} : "Process info";
109
110my %procstats = ();
111my $self;
112
113
114# Parameters that can be defined on top level of a graph. Config will print
115# them as "graph_$foo $value\n"
116my @graph_parameters = ('title','total','order','scale','vlabel','args');
117
118# Parameters that can be defined on a value-to-value basis and will be
119# blindly passed to config. Printed as "$fieldname.$param $value\n".
120my @field_parameters = ('graph', 'min', 'max', 'draw', 'cdef', 'warning',
121                        'colour', 'info', 'type');
122# Data structure that defines all possible graphs (aspects) and how they
123# are to be plotted. Every top-level entry is a graph/aspect. Each top-level graph
124# MUST have title set and 'values'.
125#
126# Graphs with 'DEBUG' set to anything is omitted from 'suggest'.
127#
128# 'rpn' on values allows easy access to graphs consisting of multiple
129# values from procstats. (Reverse polish notation). The RPN
130# implementation only accepts +-*/ and procstats-values.
131#
132# Any value left undefined will be left up to Munin to define/ignore/yell
133# about.
134#
135# See munin documentation or rrdgraph/rrdtool for more information.
136my %ASPECTS = (
137    'cpu' => {
138        'title' => "CPU Usage: $procname",
139        'order' => 'stime utime',
140        'args' => '-l 0',
141        'values' => {
142            'utime' => {
143                'type' => 'COUNTER',
144                'label' => 'User time',
145                'draw' => 'STACK'
146            },
147            'stime' => {
148                'type' => 'COUNTER',
149                'label' => 'System time',
150                'draw' => 'AREA'
151            }
152        }
153    },
154    'ctxt_switches' => {
155        'title' => "Context switches: $procname",
156        'values' => {
157            'voluntary_ctxt_switches' => {
158                'type' => 'COUNTER',
159                'label' => 'Voluntary Context Switches'
160            },
161            'nonvoluntary_ctxt_switches' => {
162                'type' => 'COUNTER',
163                'label' => 'Nonvoluntary Context Switches'
164            }
165        }
166    },
167    'threads' => {
168        'title' => "Thread count: $procname",
169        'values' => {
170            'threads' => {
171                'type' => 'GAUGE',
172                'label' => 'Number of threads'
173            }
174        }
175    },
176    'processes' => {
177        'title' => "Process count: $procname",
178        'values' => {
179            'processes' => {
180                'type' => 'GAUGE',
181                'label' => 'Number of processes'
182            }
183        }
184    },
185    'memory' => {
186        'title' => "Memory usage: $procname",
187        'vlabel' => 'bytes',
188        'order' => 'VmStk VmExe VmLib VmData VmRSS VmSize',
189        'args' => '-l 0',
190        'values' => {
191            'VmSize' => {
192                'type' => 'GAUGE',
193                'label' => 'Virtual Memory Size'
194            },
195            'VmRSS' => {
196                'type' => 'GAUGE',
197                'label' => 'Resident set size',
198            },
199            'VmData' => {
200                'type' => 'GAUGE',
201                'label' => 'Data size',
202            },
203            'VmStk' => {
204                'type' => 'GAUGE',
205                'label' => 'Stack size',
206            },
207            'VmExe' => {
208                'type' => 'GAUGE',
209                'label' => 'Segments size',
210            },
211            'VmLib' => {
212                'type' => 'GAUGE',
213                'label' => 'Shared library size',
214            }
215        }
216    }
217);
218
219# Populate %procstats with values.
220sub populate_stats
221{
222    foreach my $line(`grep -h \\\($procname\\\) /proc/*/stat`) {
223        if ($line =~ /^(\d+) \((.*)\) (.) \-?\d+ \-?\d+ \-?\d+ \-?\d+ \-?\d+ \d+ \d+ \d+ \d+ \d+ (\d+) (\d+) \d+ \d+ \d+ \d+ (\d+) \-?\d+ \d+ (\d+) (\d+)/) {
224            $procstats{"utime"} += $4;
225            $procstats{"stime"} += $5;
226            $procstats{"threads"} += $6;
227            $procstats{"vsize"} += $7;
228            $procstats{"rss"} += $8;
229            $procstats{"processes"} += 1;
230            foreach my $line(`cat /proc/$1/status`){
231                if ($line =~ /^Vm(.*):\s+(\d+) kB$/){
232                    $procstats{"Vm$1"} += ($2*1024);
233                }
234                if ($line =~ /^(.*)_ctxt_switches:\s+(\d+)$/){
235                    $procstats{"$1_ctxt_switches"} += $2;
236                }
237            }
238        }
239    }
240}
241
242# Bail-function.
243sub usage
244{
245    if (@_ && "@_" ne "") {
246        print STDERR "@_" . "\n\n";
247    }
248    print STDERR "Known arguments: suggest, config, autoconf.\n";
249    print STDERR "Run with suggest to get a list of known aspects.\n";
250    exit 1;
251}
252
253# Print 'yes' and exit true if it's reasonable to use this plugin.
254# Otherwise exit with false and a human-readable reason.
255sub autoconf
256{
257    print "no (Probably a yes, read about the plugin!)\n";
258    exit 0;
259}
260
261# Suggest relevant aspects/values of $self.
262# 'DEBUG'-graphs are excluded.
263sub suggest
264{
265    foreach my $key (keys %ASPECTS) {
266        if (defined($ASPECTS{$key}{'DEBUG'}) && $FULL_SUGGEST != 1) {
267            next;
268        }
269        print "$key\n";
270    }
271}
272
273# Print the value of a two-dimensional hash if it exist.
274# Returns false if non-existent.
275#
276# Output is formatted for plugins if arg4 is blank, otherwise arg4 is used
277# as the title/name of the field (ie: arg4=graph_title).
278sub print_if_exist
279{
280    my %values = %{$_[0]};
281    my $value = $_[1];
282    my $field = $_[2];
283    my $title = "$value.$field";
284    if (defined($_[3])) {
285        $title = $_[3];
286    }
287    if (defined($values{$value}{$field})) {
288        print "$title $values{$value}{$field}\n";
289    } else {
290        return 0;
291    }
292}
293
294# Walk through the relevant aspect and print all top-level configuration
295# values and value-definitions.
296sub get_config
297{
298    my $graph = $_[0];
299
300    # Need to double-check since set_aspect only checks this if there
301    # is no argument (suggest/autoconf doesn't require a valid aspect)
302    if (!defined($ASPECTS{$graph})) {
303        usage "No such aspect";
304    }
305    my %values = %{$ASPECTS{$graph}{'values'}};
306
307    print "graph_category $category\n";
308    foreach my $field (@graph_parameters) {
309        print_if_exist(\%ASPECTS,$graph,$field,"graph_$field");
310    }
311
312    foreach my $value (keys %values) {
313        # Need either RPN definition or a procstats value.
314        if (!defined($procstats{$value}) &&
315            !defined($values{$value}{'rpn'})) {
316                if ($DEBUG) {
317                    print "ERROR: $value not part of procstats.\n"
318                }
319                next;
320        }
321
322        if (!print_if_exist(\%values,$value,'label')) {
323            print "$value.label ".$ASPECTS{$self}{'values'}{$value}{'label'}."\n";
324        }
325        foreach my $field (@field_parameters) {
326            print_if_exist(\%values,$value,$field);
327        }
328    }
329}
330
331# Read and verify the aspect ($self).
332sub set_aspect
333{
334    $self = $0;
335    $self =~ s/^.*proc_//;
336    if (!defined($ASPECTS{$self}) && @ARGV == 0) {
337        usage "No such aspect";
338    }
339}
340
341# Handle arguments (config, autoconf, suggest)
342# Populate stats for config is necessary, but we want to avoid it for
343# autoconf as it would generate a nasty error.
344sub check_args
345{
346    if (@ARGV && $ARGV[0] eq '') {
347        shift @ARGV;
348    }
349    if (@ARGV == 1) {
350        if ($ARGV[0] eq "config") {
351            populate_stats;
352            get_config($self);
353            exit 0;
354        } elsif ($ARGV[0] eq "autoconf") {
355            autoconf($self);
356            exit 0;
357        } elsif ($ARGV[0] eq "suggest") {
358#            suggest;
359            exit 0;
360        }
361        usage "Unknown argument";
362    }
363}
364
365# Braindead RPN: +,-,/,* will pop two items from @stack, and perform
366# the relevant operation on the items. If the item in the array isn't one
367# of the 4 basic math operations, a value from procstats is pushed on to
368# the stack. IE: 'client_req','client_conn','/' will leave the value of
369# "client_req/client_conn" on the stack.
370#
371# If only one item is left on the stack, it is printed. Otherwise, an error
372# message is printed.
373sub rpn
374{
375    my @stack;
376    my $left;
377    my $right;
378    foreach my $item (@{$_[0]}) {
379        if ($item eq "+") {
380            $right = pop(@stack);
381            $left = pop(@stack);
382            push(@stack,$left+$right);
383        } elsif ($item eq "-") {
384            $right = pop(@stack);
385            $left = pop(@stack);
386            push(@stack,$left-$right);
387        } elsif ($item eq "/") {
388            $right = pop(@stack);
389            $left = pop(@stack);
390            push(@stack,$left/$right);
391        } elsif ($item eq "*") {
392            $right = pop(@stack);
393            $left = pop(@stack);
394            push(@stack,$left*$right);
395        } else {
396            push(@stack,int($procstats{$item}));
397        }
398    }
399    if (@stack > 1)
400    {
401        print STDERR "RPN error: Stack has more than one item left.\n";
402        print STDERR "@stack\n";
403        exit 255;
404    }
405    print "@stack";
406    print "\n";
407}
408
409################################
410# Execution starts here        #
411################################
412
413set_aspect;
414check_args;
415populate_stats;
416
417# We only get here if we're supposed to.
418
419# Walks through the relevant values and either prints the procstat, or
420# if the 'rpn' variable is set, calls rpn() to execute ... the rpn.
421foreach my $value (keys %{$ASPECTS{$self}{'values'}}) {
422    if (defined($ASPECTS{$self}{'values'}{$value}{'rpn'})) {
423        print "$value.value ";
424        rpn($ASPECTS{$self}{'values'}{$value}{'rpn'});
425    } else {
426        print "$value.value ";
427        if (!defined($procstats{$value})) {
428            print "0\n";
429            next;
430        }
431        print "$procstats{$value}\n";
432    }
433}
434