Context Navigation

source: trunk/nagios-velvice/velvice.cgi @ 270

Last change on this file since 270 was 270, checked in by g7moreau, 7 years ago
Update CSS style
Property svn:keywords set to `Id`
File size: 16.1 KB

Rev	Line
[240]	1	#!/usr/bin/env perl
	2	#
	3	# 2014/05/15 Gabriel Moreau <Gabriel.Moreau@univ-grenoble-alpes.fr>
	4	# 2017/06/22 Gabriel Moreau - big update
[256]	5	# 2018/06/25 Gabriel Moreau - make velvice generic
[240]	6	#
	7	# velvice.cgi
	8	# Copyright (C) 2014-2018, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France
	9	#
	10	# Need NagiosStatus http://exchange.nagios.org/directory/Addons/APIs/Perl/NagiosStatus-2Epm/details
	11	# Possible command http://old.nagios.org/developerinfo/externalcommands/commandlist.php
	12	#
[250]	13	# apt-get install perl-modules libnagios-object-perl libhtml-parser-perl liburi-encode-perl libcolor-calc-perl libyaml-syck-perl
[240]	14
	15	use strict;
	16	use warnings;
[270]	17	use version; our $VERSION = version->declare('0.7.6');
[240]	18
	19	use CGI;
	20	use HTML::Entities ();
	21	use Nagios::StatusLog;
	22	use URI::Encode qw(uri_encode uri_decode);
	23	use Color::Calc ();
	24	use YAML::Syck;
	25
[257]	26	my $query = CGI->new();
	27	my $cgi_check = uri_decode($query->param('check'));
	28	my $cgi_script_name = $query->script_name();
[260]	29	my $cgi_path = $cgi_script_name =~ s{/[^/]+\.cgi$}{}r;
[263]	30	undef $query;
[256]	31
[240]	32	my $config = {};
	33	$config = YAML::Syck::LoadFile('/etc/nagios3/velvice.yml') if -e '/etc/nagios3/velvice.yml';
[256]	34	$config->{'nagios-server'} \|\|= {};
	35	$config->{'nagios-server'}{'status-file'} \|\|= '/var/cache/nagios3/status.dat';
	36	$config->{'nagios-server'}{'nagios-cmd'} \|\|= '/var/lib/nagios3/rw/nagios.cmd';
[260]	37	$config->{'nagios-server'}{'portal-url'} \|\|= $cgi_path =~ s{/cgi-bin/}{/}r . '/';
[258]	38	$config->{'nagios-server'}{'status-cgi'} \|\|= "$cgi_path/status.cgi";
[263]	39	$config->{'nagios-server'}{'stylesheets'} \|\|= $config->{'nagios-server'}{'portal-url'} =~ s{/?$}{/stylesheets}r;
[256]	40	$config->{'host-mapping'} \|\|= {};
	41	$config->{'color-downtime'} \|\|= {};
[261]	42	$config->{'color-downtime'}{'day-min'} \|\|= 3;
	43	$config->{'color-downtime'}{'day-max'} \|\|= 50;
	44	$config->{'color-downtime'}{'factor'} \|\|= 0.7;
[256]	45	$config->{'remote-action'} \|\|= {};
[240]	46
	47	my $log = Nagios::StatusLog->new(
[256]	48	Filename => $config->{'nagios-server'}{'status-file'},
[240]	49	Version => 3.0
	50	);
	51
	52	sub hostmapping {
	53	my $host = shift;
	54
[256]	55	return exists $config->{'host-mapping'}{$host} ? $config->{'host-mapping'}{$host} : $host;
[240]	56	}
	57
	58	sub downtime {
	59	my ($time_change) = @_;
	60
	61	my $now = time;
	62	return sprintf '%.1f', ($now - $time_change) / (60 * 3600);
	63	}
	64
	65	sub alertcolor {
	66	my ($color, $downtime) = @_;
	67
[261]	68	$downtime = $downtime - $config->{'color-downtime'}{'day-min'}; # same color first days
	69	$downtime = $config->{'color-downtime'}{'day-max'} if $downtime > $config->{'color-downtime'}{'day-max'}; # max 50 days for color
[240]	70	$downtime = 0 if $downtime < 0;
	71
[261]	72	my $factor = ($downtime * $config->{'color-downtime'}{'factor'}) / $config->{'color-downtime'}{'day-max'};
[240]	73	return Color::Calc::color_light_html($color, $factor);
	74	}
	75
	76	my %hostdown;
	77	my @serviceproblems;
	78	my %hostcount;
	79	my @futurecheck;
	80	HOST:
	81	for my $host (sort $log->list_hosts()) {
	82	my $host_stat = $log->host($host);
	83
	84	if ($host_stat->status eq 'DOWN') {TESTIF:{
	85	for my $srv ($log->list_services_on_host($host)) {
	86	last TESTIF if $log->service($host, $srv)->status eq 'OK' or $log->service($host, $srv)->status eq 'PENDING';
	87	}
	88
	89	$hostdown{$host} = $host_stat;
	90	next HOST;
	91	}}
	92
	93	for my $srv ($log->list_services_on_host($host)) {
	94	if ($log->service($host, $srv)->status ne 'OK') {
	95	push @serviceproblems, $log->service($host, $srv);
	96	$hostcount{$host}++;
	97	}
	98	}
	99	}
	100
	101	my $now = time;
	102	my ($sec, $min, $hour, $mday, $mon, $year, $wday, $yday, $isdst) = localtime $now;
	103	$year += 1900;
	104	$mon++;
	105	my $date = sprintf '%04i-%02i-%02i %02i:%02i', $year, $mon, $mday, $hour, $min;
	106
	107	my $htmlpage = <<"ENDH";
	108	Content-Type: text/html
	109
	110	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
	111	<html lang="en">
	112	<head>
	113	<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
	114	<title>Nagios Velvice</title>
[264]	115	<link rel="stylesheet" type="text/css" href="$config->{'nagios-server'}{'stylesheets'}/velvice.css">
[240]	116	</head>
	117	<body>
	118	<h1>
	119	<ul>
[256]	120	<li>Nagios Velvice Alert Panel : <a href="$config->{'nagios-server'}{'portal-url'}">Core Server</a></li>
[257]	121	<li><small>(<a href="$cgi_script_name">UPDATE</a> - $date)</small></li>
[240]	122	</ul>
	123	</h1>
	124	ENDH
	125
[254]	126	my %service_name = ();
	127	my %service_status = ();
[240]	128	for my $srv (@serviceproblems) {
	129	$service_name{$srv->service_description}++;
[254]	130	$service_status{$srv->status}++;
[240]	131	}
	132
	133	if (scalar @serviceproblems == 0) {
	134	$htmlpage .= "<p>No alert to recheck.</p>\n";
	135	}
	136	else {
	137
	138	$htmlpage .= "<p>Alert to recheck - Level:\n";
	139	$htmlpage .= join ",\n",
[257]	140	" <a href='$cgi_script_name?check=all'>ALL</a><small>(" . scalar(@serviceproblems) . ')</small>',
	141	map(" <a href='$cgi_script_name?check=" . lc(uri_encode($_)) . "'>$_</a>($service_status{$_})", sort keys %service_status);
[240]	142	$htmlpage .= ".\n";
	143	$htmlpage .= " <br />\n";
	144	$htmlpage .= " Service:\n";
[257]	145	$htmlpage .= join ",\n", map(" <a href='$cgi_script_name?check=" . lc(uri_encode($_)) . "'>$_</a><small>($service_name{$_})</small>", sort keys %service_name);
[240]	146	$htmlpage .= ".\n";
	147	$htmlpage .= "</p>\n";
	148
	149	my $nagios_cmd;
[256]	150	open $nagios_cmd, '>>', $config->{'nagios-server'}{'nagios-cmd'} or die "Can't open file filename: $!";
[240]	151
[254]	152	my %remote_sshdown = ();
	153	my %remote_db = ();
	154	my $remote_flag;
[245]	155
[240]	156	my $current_host = '';
	157	$htmlpage .= "<table border=\"1\">\n";
[254]	158	SERVICE_PROBLEMS:
[240]	159	for my $srv (@serviceproblems) {
	160	my $hostname = $srv->host_name;
	161	my $service = $srv->service_description;
[245]	162	my $status = $srv->status;
[240]	163	my $downtime = downtime($srv->last_state_change);
[257]	164	my $output = HTML::Entities::encode($srv->plugin_output) =~ s/^[A-Z_\s]+?[:-]//r;
[240]	165
[245]	166	my $color = $status eq 'CRITICAL' ? '#F88888' : '#FFFF00';
[240]	167	$color = alertcolor($color, $downtime);
	168	$htmlpage .= " <tr style='background:$color;'>\n";
	169	if ($hostname ne $current_host) {
	170	$current_host = $hostname;
[256]	171	$htmlpage .= " <td rowspan='$hostcount{$hostname}' style='vertical-align:middle;'><a href=\"$config->{'nagios-server'}{'status-cgi'}?host=$hostname\">$hostname</a></td>\n";
[240]	172	}
[245]	173
	174	my $bold;
[254]	175	ACTION_STYLE:
[256]	176	for my $act_name (keys %{$config->{'remote-action'}}) {
	177	my $act_regex = $config->{'remote-action'}{$act_name}{'regex'};
	178	$bold++ if $service =~ m/$act_regex/ and $config->{'remote-action'}{$act_name}{'style'} eq 'bold';
[240]	179	}
[246]	180	$htmlpage .= $bold ? ' <td class="bold">' : ' <td>';
	181	$htmlpage .= "$service</td>\n";
[245]	182
	183	$htmlpage .= " <td>$status</td>\n";
[240]	184	$htmlpage .= " <td style='max-width:60%;'><small>$output";
	185
[257]	186	if (($cgi_check =~ m/all/i)
	187	or ($cgi_check =~ m/^$service$/i)
	188	or ($cgi_check =~ m/critical/i and $status eq 'CRITICAL')
	189	or ($cgi_check =~ m/warning/i and $status eq 'WARNING')
	190	or ($cgi_check =~ m/pending/i and $status eq 'PENDING')
[240]	191	) {
	192	$now++;
	193	my $interval = $srv->next_check() - $srv->last_check() \|\| 300;
	194	$interval = 240 if $interval < 240;
	195	$interval = 3000 if $interval > 3000;
[261]	196	my $future = $now + 20 + int(rand($interval - 20)); # 5 * 60 = 300
[240]	197
	198	$htmlpage .= " -- <b>CHECK</b> [$now/" . ($future - $now) . "]";
	199	printf $nagios_cmd "[%lu] SCHEDULE_FORCED_SVC_CHECK;%s;%s;%lu\n", $now, $hostname, $service, $now;
	200	# delay future command
	201	push @futurecheck, sprintf "[%lu] SCHEDULE_FORCED_SVC_CHECK;%s;%s;%lu", $future, $hostname, $service, $future;
	202	}
	203
[254]	204	ACTION_PUSH_AND_DEPEND:
[256]	205	for my $act_name (keys %{$config->{'remote-action'}}) {
	206	my $act_regex = $config->{'remote-action'}{$act_name}{'regex'};
	207	my $act_status = $config->{'remote-action'}{$act_name}{'status'} \|\| 'ALL';
	208	my $act_depend = $config->{'remote-action'}{$act_name}{'depend'} \|\| 'SSH';
[254]	209
	210	if ($service =~ m/$act_regex/ and ($act_status eq 'ALL' or $status =~ m/$act_status/)) {
	211	$remote_db{$act_name} \|\|= [];
	212	push @{$remote_db{$act_name}}, $hostname;
	213	$remote_flag++;
[245]	214	}
[254]	215
	216	# check depend service otherwise
	217	$remote_sshdown{$act_depend} \|\|= {};
	218	$remote_sshdown{$act_depend}->{$hostname}++ if $service =~ m/$act_depend/;
[245]	219	}
[240]	220
	221	$htmlpage .= "</small></td>\n";
	222	$htmlpage .= " <td style='text-align:right;'>$downtime days</td>\n";
	223	$htmlpage .= " </tr>\n";
	224	}
	225
	226	$htmlpage .= "</table>\n";
	227	close $nagios_cmd;
	228
[254]	229	# host down
[240]	230	if (%hostdown) {
	231	$htmlpage .= "<br />\n";
	232	$htmlpage .= "<table border='1'>\n";
[254]	233	HOST_DOWN:
[240]	234	for my $host (sort keys %hostdown) {
	235	my $host_stat = $hostdown{$host};
	236	my $hostname = $host_stat->host_name;
	237	my $downtime = downtime($host_stat->last_state_change);
	238	my $color = alertcolor('#F88888', $downtime);
	239	$htmlpage .= " <tr style='background:$color'>\n";
[256]	240	$htmlpage .= " <td><a href=\"$config->{'nagios-server'}{'status-cgi'}?host=$hostname\">$hostname</a></td>\n";
[240]	241	my @host_service;
	242	for my $srv ($log->list_services_on_host($host)) {
	243	push @host_service, $log->service($host, $srv)->service_description;
	244	}
	245	$htmlpage .= " <td><small>" . join(', ', @host_service) . "</small></td>\n";
	246	$htmlpage .= " <td style='text-align:right;'>$downtime days</td>\n";
	247	$htmlpage .= " </tr>\n";
	248	}
	249	$htmlpage .= "</table>\n";
	250	}
	251
[254]	252	# remote action
	253	if ($remote_flag) {
[240]	254	require Nagios::Object::Config;
	255	my $parser = Nagios::Object::Config->new();
	256	$parser->parse("/var/cache/nagios3/objects.cache");
	257
[254]	258	REMOTE_ACTION:
	259	for my $act_name (keys %remote_db) {
[256]	260	my $act_depend = $config->{'remote-action'}{$act_name}{'depend'} \|\| 'SSH';
[254]	261
	262	my @action = grep !exists $remote_sshdown{$act_depend}->{$_}, @{$remote_db{$act_name}};
[245]	263	if (@action) {
[256]	264	my $srv_title = $config->{'remote-action'}{$act_name}{'title'} \|\| "Action: $act_name";
[247]	265	$htmlpage .= "<h2>$srv_title</h2>\n";
[245]	266	$htmlpage .= "<pre>\n";
[256]	267	my $remote_action = $config->{'remote-action'}{$act_name}{'command'};
	268	$remote_action = $config->{'remote-action'}{$act_name}{'command-one'}
	269	if @action == 1 and exists $config->{'remote-action'}{$act_name}{'command-one'};
[245]	270	my @hosts;
	271	for my $host (@action) {
	272	my $object = $parser->find_object("$host", "Nagios::Host");
	273	push @hosts, hostmapping($object->address =~ s/\..*$//r);
	274	}
	275	my $hosts_list = join ' ', @hosts;
[246]	276	$htmlpage .= ' ' . $remote_action =~ s{\%m}{$hosts_list}r;
[245]	277	$htmlpage .= "</pre>\n";
	278	}
	279	}
[240]	280	}
	281	}
	282
[269]	283	$htmlpage .= <<"ENDH";
	284	<hr clear="all">
	285	<a href="http://servforge.legi.grenoble-inp.fr/projects/soft-trokata/wiki/SoftWare/NagiosVelvice">Velvice</a> version: $VERSION
	286	<ul>
	287	<li>Written by Gabriel Moreau</li>
	288	<li>Licence GNU GPL version 2 or later and Perl equivalent</li>
	289	<li>Copyright (C) 2014-2018, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France</li>
	290	</ul>
[240]	291	</body>
	292	</html>
	293	ENDH
	294
	295	print $htmlpage;
	296
[254]	297	# delayed future check
[240]	298	if (@futurecheck) {
	299	sleep 2;
	300	my $nagios_cmd;
[256]	301	open $nagios_cmd, '>>', $config->{'nagios-server'}{'nagios-cmd'} or die "Can't open file filename: $!";
[240]	302	print $nagios_cmd "$_\n" for @futurecheck;
	303	close $nagios_cmd;
	304	}
	305
	306	__END__
	307
	308
[241]	309	=head1 NAME
	310
	311	velvice.cgi - nagios velvice alert panel
	312
[265]	313	=head1 USAGE
[241]	314
[265]	315	velvice.cgi
	316	velvice.cgi?check=XXX
	317
	318
[241]	319	=head1 DESCRIPTION
	320
	321	Nagios VELVICE is an acronym for "Nagios leVEL serVICE status".
	322
[265]	323	The Nagios web page is sometimes very graphically charged
	324	and does not necessarily contain the information you need at a glance.
	325	For example, it is quite complicated to restart controls on multiple hosts in one click.
[241]	326
[265]	327	For example, a server that is down should take only one line and not one per service...
	328	Similarly, a service that has been down for 5 minutes or since yesterday
	329	has more weight than a service that has fallen for 15 days.
[241]	330
[265]	331	With Velvice Panel, a broken down server takes only one line.
	332	Services that have been falling for a long time gradually lose their color and become pastel colors.
[241]	333
[265]	334	With Velvice Panel, it is possible through a single click
	335	to redo a check of all services that are in the CRITICAL state.
	336	Similarly, it is possible to restart a check on all SSH services in breakdowns ...
	337	In order not to clog the Nagios server, checks are shifted by 2 seconds in time.
	338
	339	There is also a link to the web page of the main Nagios server.
	340	For each computer, you have a direct link to its dedicated web page on this server.
	341
	342
	343	=head1 CONFIGURATION FILE SPECIFICATION
	344
	345	The configuration file must be F</etc/nagios3/velvice.yml>.
	346	This is not a required file.
	347	The file is in YAML format because this is a human-readable text file style.
	348	Other formats could have been Plain XML, RDF, JSON... but they are much less readable.
	349
	350	You can find in the software nagios-velvice an example of configuration:
	351	L<velvice.sample.yml\|http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/velvice.sample.yml>.
	352	This one is in fact the master reference specification!
	353
	354	The main keys C<nagios-server> and C<color-downtime> have good default values.
	355	No secondary key is required...
[268]	356	The Velvice script try hard to replace ~ by the good value automatically.
[265]	357
[268]	358	nagios-server:
	359	status-file: /var/cache/nagios3/status.dat
	360	nagios-cmd: /var/lib/nagios3/rw/nagios.cmd
	361	portal-url: ~/nagios3/
	362	status-cgi: ~/cgi-bin/nagios3/status.cgi
	363	stylesheets: ~/nagios3/stylesheets
	364
	365	The background color of the faulty service line display remains stable with a bright color for at least 3 days.
	366	Then, it decreases and becomes pastel after 53 days with an intensity of 70% (100% is white and 0% is black).
	367
	368	color-downtime:
	369	day-min: 3
	370	day-max: 50
	371	factor: 0.7
	372
[265]	373	With key C<host-mapping>,
	374	it's good to map C<localhost> to the real name of the computer (hostname).
	375
[268]	376	host-mapping:
	377	localhost: srv-nagios
	378	toto: titi
	379
[265]	380	The only important key is C<remote-action>.
	381	You can affiliate as many subkeys as you want.
	382	Let's take an example:
	383
	384	remote-action:
	385	oom-killer:
	386	regex: ^OOM Killer
	387	title: OOM Killer
	388	command: tssh -c 'sudo rm /var/lib/nagios3/nagios_oom_killer.log' %m
	389	command-one: ssh %m 'sudo rm /var/lib/nagios3/nagios_oom_killer.log'
	390	depend: ^SSH
	391	status: ALL
	392	style: bold
	393
	394	C<oom-killer> is just a key for your remote action.
	395	The regex is used to find which service has a problem...
	396	The title is use in the result web page (not mandatory - otherwise, it will be C<Action: oom-killer>).
	397	The C<command> is just written on this web page.
	398	You have the responsibility to copy / cut it on a terminal.
	399	For security reasons, the nagios server does not have the right to launch the command on the remote host.
	400	The wildcard C<%m> is replaced by the list of the host (separated by the space).
	401	Sometime, the command could be different if there is only one computer (just SSH and no parallel SSH).
	402	If your command is based on SSH,
	403	you can have an SSH action only if the remote SSH is running.
	404	So you can make the remote action depend on the SSH service through a regular expression of your choice.
	405
	406	The last two keys.
	407	The C<status> key is for CRITICAL or WARNING (or ALL).
	408	The key C<style> is there to mark in bold the service in error on the web page.
	409
	410	=head1 SEE ALSO
	411
	412	yamllint(1), ysh(1), YAML, Nagios::StatusLog, Color::Calc
	413
	414	In Debian GNU/Linux distribution, packages for C<yamllint> and C<ysh> are:
	415
	416	=over
	417
	418	=item * C<yamllint> - Linter for YAML files (Python)
	419
	420	=item * C<libyaml-shell-perl> - YAML test shell (Perl)
	421
	422	=back
	423
	424
	425	Own project ressources:
	426
	427	=over
	428
	429	=item * L<Web site\|http://servforge.legi.grenoble-inp.fr/projects/soft-trokata/wiki/SoftWare/NagiosVelvice>
	430
	431	=item * L<Online Manual\|http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/velvice.html>
	432
	433	=item * L<SVN repository\|http://servforge.legi.grenoble-inp.fr/svn/soft-trokata/trunk/nagios-velvice>
	434
	435	=back
	436
	437
	438	=head1 VERSION
	439
[266]	440	$Id: velvice.cgi 270 2018-06-26 17:55:02Z g7moreau $
[265]	441
	442
	443	=head1 AUTHOR
	444
	445	Written by Gabriel Moreau <Gabriel.Moreau(A)univ-grenoble-alpes.fr>, LEGI UMR 5519, CNRS, Grenoble - France
	446
	447
[241]	448	=head1 LICENSE AND COPYRIGHT
	449
	450	Licence GNU GPL version 2 or later and Perl equivalent
	451
[265]	452	Copyright (C) 2014-2018, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France

Note: See TracBrowser for help on using the repository browser.

Download in other formats: