| [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 |
|---|
| [369] | 6 | # 2018/11/03 Gabriel Moreau - ajax |
|---|
| [240] | 7 | # |
|---|
| 8 | # velvice.cgi |
|---|
| [380] | 9 | # Copyright (C) 2014-2019, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France |
|---|
| [240] | 10 | # |
|---|
| 11 | # Need NagiosStatus http://exchange.nagios.org/directory/Addons/APIs/Perl/NagiosStatus-2Epm/details |
|---|
| 12 | # Possible command http://old.nagios.org/developerinfo/externalcommands/commandlist.php |
|---|
| 13 | # |
|---|
| [250] | 14 | # apt-get install perl-modules libnagios-object-perl libhtml-parser-perl liburi-encode-perl libcolor-calc-perl libyaml-syck-perl |
|---|
| [322] | 15 | # apt-get install libdatetime-event-recurrence-perl libdatetime-set-perl |
|---|
| [240] | 16 | |
|---|
| 17 | use strict; |
|---|
| 18 | use warnings; |
|---|
| [382] | 19 | use version; our $VERSION = version->declare('0.10.8'); |
|---|
| [240] | 20 | |
|---|
| 21 | use CGI; |
|---|
| 22 | use HTML::Entities (); |
|---|
| 23 | use Nagios::StatusLog; |
|---|
| 24 | use URI::Encode qw(uri_encode uri_decode); |
|---|
| 25 | use Color::Calc (); |
|---|
| 26 | use YAML::Syck; |
|---|
| 27 | |
|---|
| [257] | 28 | my $query = CGI->new(); |
|---|
| 29 | my $cgi_check = uri_decode($query->param('check')); |
|---|
| 30 | my $cgi_script_name = $query->script_name(); |
|---|
| [260] | 31 | my $cgi_path = $cgi_script_name =~ s{/[^/]+\.cgi$}{}r; |
|---|
| [369] | 32 | my $cgi_only; |
|---|
| 33 | $cgi_only++ if uri_decode($query->param('only')) eq 'body'; |
|---|
| [263] | 34 | undef $query; |
|---|
| [256] | 35 | |
|---|
| [321] | 36 | my %STATUS_DB = ( |
|---|
| 37 | CRITICAL => {id => 3, color => '#F88888'}, |
|---|
| 38 | WARNING => {id => 2, color => '#FFFF00'}, |
|---|
| 39 | PENDING => {id => 1, color => '#E0E0E0'}, |
|---|
| 40 | ); |
|---|
| 41 | |
|---|
| [240] | 42 | my $config = {}; |
|---|
| 43 | $config = YAML::Syck::LoadFile('/etc/nagios3/velvice.yml') if -e '/etc/nagios3/velvice.yml'; |
|---|
| [256] | 44 | $config->{'nagios-server'} ||= {}; |
|---|
| 45 | $config->{'nagios-server'}{'status-file'} ||= '/var/cache/nagios3/status.dat'; |
|---|
| 46 | $config->{'nagios-server'}{'nagios-cmd'} ||= '/var/lib/nagios3/rw/nagios.cmd'; |
|---|
| [260] | 47 | $config->{'nagios-server'}{'portal-url'} ||= $cgi_path =~ s{/cgi-bin/}{/}r . '/'; |
|---|
| [258] | 48 | $config->{'nagios-server'}{'status-cgi'} ||= "$cgi_path/status.cgi"; |
|---|
| [263] | 49 | $config->{'nagios-server'}{'stylesheets'} ||= $config->{'nagios-server'}{'portal-url'} =~ s{/?$}{/stylesheets}r; |
|---|
| [328] | 50 | $config->{'nagios-server'}{'image'} ||= $config->{'nagios-server'}{'portal-url'} =~ s{/?$}{/images}r; |
|---|
| [256] | 51 | $config->{'host-mapping'} ||= {}; |
|---|
| 52 | $config->{'color-downtime'} ||= {}; |
|---|
| [261] | 53 | $config->{'color-downtime'}{'day-min'} ||= 3; |
|---|
| 54 | $config->{'color-downtime'}{'day-max'} ||= 50; |
|---|
| 55 | $config->{'color-downtime'}{'factor'} ||= 0.7; |
|---|
| [256] | 56 | $config->{'remote-action'} ||= {}; |
|---|
| [322] | 57 | $config->{'refresh'} ||= 0; |
|---|
| [240] | 58 | |
|---|
| 59 | sub hostmapping { |
|---|
| 60 | my $host = shift; |
|---|
| 61 | |
|---|
| [256] | 62 | return exists $config->{'host-mapping'}{$host} ? $config->{'host-mapping'}{$host} : $host; |
|---|
| [240] | 63 | } |
|---|
| 64 | |
|---|
| 65 | sub downtime { |
|---|
| 66 | my ($time_change) = @_; |
|---|
| 67 | |
|---|
| 68 | my $now = time; |
|---|
| 69 | return sprintf '%.1f', ($now - $time_change) / (60 * 3600); |
|---|
| 70 | } |
|---|
| 71 | |
|---|
| 72 | sub alertcolor { |
|---|
| [320] | 73 | my ($status, $downtime) = @_; |
|---|
| [240] | 74 | |
|---|
| [320] | 75 | my $color = '#0000FF'; |
|---|
| 76 | $color = $STATUS_DB{$status}->{'color'} if exists $STATUS_DB{$status}; |
|---|
| 77 | |
|---|
| [261] | 78 | $downtime = $downtime - $config->{'color-downtime'}{'day-min'}; # same color first days |
|---|
| 79 | $downtime = $config->{'color-downtime'}{'day-max'} if $downtime > $config->{'color-downtime'}{'day-max'}; # max 50 days for color |
|---|
| [240] | 80 | $downtime = 0 if $downtime < 0; |
|---|
| 81 | |
|---|
| [261] | 82 | my $factor = ($downtime * $config->{'color-downtime'}{'factor'}) / $config->{'color-downtime'}{'day-max'}; |
|---|
| [240] | 83 | return Color::Calc::color_light_html($color, $factor); |
|---|
| 84 | } |
|---|
| 85 | |
|---|
| [284] | 86 | sub nosbreak { |
|---|
| 87 | my ($str) = @_; |
|---|
| 88 | |
|---|
| [285] | 89 | return $str =~ s/\s/\ /gr; |
|---|
| [284] | 90 | } |
|---|
| 91 | |
|---|
| [322] | 92 | my $log = Nagios::StatusLog->new( |
|---|
| 93 | Filename => $config->{'nagios-server'}{'status-file'}, |
|---|
| 94 | Version => 3.0 |
|---|
| 95 | ); |
|---|
| 96 | |
|---|
| 97 | # refresh configuration |
|---|
| 98 | if (exists $config->{'refreshments'}) { |
|---|
| 99 | require DateTime::Event::Recurrence; |
|---|
| 100 | require DateTime::SpanSet; |
|---|
| 101 | |
|---|
| 102 | my @refreshments; |
|---|
| 103 | SET: |
|---|
| 104 | for my $set (@{$config->{'refreshments'}}) { |
|---|
| 105 | my $start = DateTime::Event::Recurrence->weekly(days => $set->{'days'}, hours => $set->{'start'}); |
|---|
| 106 | my $end = DateTime::Event::Recurrence->weekly(days => $set->{'days'}, hours => $set->{'end'}); |
|---|
| 107 | my $spanset = DateTime::SpanSet->from_sets(start_set => $start, end_set => $end); |
|---|
| 108 | push @refreshments, {refresh => $set->{'refresh'}, spanset => $spanset}; |
|---|
| 109 | } |
|---|
| 110 | |
|---|
| [324] | 111 | my $now = DateTime->now(time_zone => 'local'); |
|---|
| [322] | 112 | SET: |
|---|
| 113 | for my $set (@refreshments) { |
|---|
| 114 | next SET if not $set->{'spanset'}->contains($now); |
|---|
| 115 | |
|---|
| 116 | $config->{'refresh'} = $set->{'refresh'}; |
|---|
| 117 | last SET; |
|---|
| 118 | } |
|---|
| 119 | } |
|---|
| 120 | |
|---|
| [240] | 121 | my %hostdown; |
|---|
| 122 | my @serviceproblems; |
|---|
| 123 | my %hostcount; |
|---|
| 124 | my @futurecheck; |
|---|
| 125 | HOST: |
|---|
| 126 | for my $host (sort $log->list_hosts()) { |
|---|
| 127 | my $host_stat = $log->host($host); |
|---|
| 128 | |
|---|
| 129 | if ($host_stat->status eq 'DOWN') {TESTIF:{ |
|---|
| 130 | for my $srv ($log->list_services_on_host($host)) { |
|---|
| 131 | last TESTIF if $log->service($host, $srv)->status eq 'OK' or $log->service($host, $srv)->status eq 'PENDING'; |
|---|
| 132 | } |
|---|
| 133 | |
|---|
| 134 | $hostdown{$host} = $host_stat; |
|---|
| 135 | next HOST; |
|---|
| 136 | }} |
|---|
| 137 | |
|---|
| [319] | 138 | SRV: |
|---|
| [240] | 139 | for my $srv ($log->list_services_on_host($host)) { |
|---|
| [319] | 140 | my $status = $log->service($host, $srv)->status; |
|---|
| 141 | |
|---|
| 142 | next SRV if $status eq 'OK'; |
|---|
| 143 | |
|---|
| 144 | push @serviceproblems, $log->service($host, $srv); |
|---|
| 145 | |
|---|
| 146 | my $downtime = downtime($log->service($host, $srv)->last_state_change); |
|---|
| [320] | 147 | my $color = alertcolor($status, $downtime); |
|---|
| [319] | 148 | |
|---|
| 149 | my $status_id = 0; |
|---|
| [320] | 150 | $status_id = $STATUS_DB{$status}->{'id'} if exists $STATUS_DB{$status}; |
|---|
| [319] | 151 | |
|---|
| 152 | #$hostcount{$host}++; |
|---|
| 153 | $hostcount{$host} ||= {count => 0, color => $color, status_id => $status_id, downtime => $downtime}; |
|---|
| 154 | $hostcount{$host}->{'count'}++; |
|---|
| [380] | 155 | if (($status_id > $hostcount{$host}->{'status_id'}) |
|---|
| 156 | or (($status_id == $hostcount{$host}->{'status_id'}) and ($downtime < $hostcount{$host}->{'downtime'}))) { |
|---|
| [319] | 157 | $hostcount{$host}->{'downtime'} = $downtime; |
|---|
| 158 | $hostcount{$host}->{'status_id'} = $status_id; |
|---|
| 159 | $hostcount{$host}->{'color'} = $color; |
|---|
| [240] | 160 | } |
|---|
| 161 | } |
|---|
| 162 | } |
|---|
| 163 | |
|---|
| 164 | my $now = time; |
|---|
| 165 | my ($sec, $min, $hour, $mday, $mon, $year, $wday, $yday, $isdst) = localtime $now; |
|---|
| 166 | $year += 1900; |
|---|
| 167 | $mon++; |
|---|
| [377] | 168 | # my $date_sec = sprintf '%04i-%02i-%02i %02i:%02i:%02i', $year, $mon, $mday, $hour, $min, $sec; |
|---|
| [376] | 169 | my $date_min = nosbreak(sprintf '%04i-%02i-%02i %02i:%02i', $year, $mon, $mday, $hour, $min); |
|---|
| [240] | 170 | |
|---|
| [369] | 171 | my $htmlpage; |
|---|
| 172 | |
|---|
| 173 | $htmlpage .= <<"ENDH" if not $cgi_only; |
|---|
| [240] | 174 | Content-Type: text/html |
|---|
| 175 | |
|---|
| 176 | <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> |
|---|
| 177 | <html lang="en"> |
|---|
| 178 | <head> |
|---|
| 179 | <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> |
|---|
| [323] | 180 | ENDH |
|---|
| [369] | 181 | |
|---|
| 182 | $htmlpage .= <<"ENDH" if $cgi_only; |
|---|
| 183 | Content-Type: text/xml |
|---|
| 184 | |
|---|
| 185 | <?xml version="1.0" encoding="utf-8"?> |
|---|
| 186 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Frameset//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-frameset.dtd"> |
|---|
| 187 | ENDH |
|---|
| 188 | |
|---|
| 189 | #$htmlpage .= " <meta http-equiv=\"Refresh\" content=\"$config->{'refresh'}\">" if $config->{'refresh'} > 59; # minimum 1 min |
|---|
| 190 | $htmlpage .= <<"ENDH" if not $cgi_only; |
|---|
| [240] | 191 | <title>Nagios Velvice</title> |
|---|
| [282] | 192 | <link rel="stylesheet" type="text/css" href="$config->{'nagios-server'}{'stylesheets'}/velvice.css"> |
|---|
| 193 | <link rel="shortcut icon" type="image/ico" href="$config->{'nagios-server'}{'image'}/favicon.ico"> |
|---|
| [369] | 194 | <script type="text/javascript"> |
|---|
| 195 | var refresh_sec = 900; // 15 min = 900 s |
|---|
| [370] | 196 | var refresh_interval; |
|---|
| [369] | 197 | function refresh() { |
|---|
| 198 | var req = new XMLHttpRequest(); |
|---|
| [377] | 199 | var datenow = new Date(); |
|---|
| [369] | 200 | console.log("Grabbing Value"); |
|---|
| 201 | req.open("GET", '$cgi_script_name?only=body', true); // Grabs whatever you've written in this file |
|---|
| 202 | req.onload = function () { |
|---|
| 203 | if (req.status == 200) { |
|---|
| [372] | 204 | document.getElementById('master-body').innerHTML = req.responseXML.getElementById('master-body').innerHTML; |
|---|
| [378] | 205 | console.log("Update Page at", datenow.toLocaleTimeString()); |
|---|
| [371] | 206 | |
|---|
| [370] | 207 | refresh_next = req.responseXML.getElementById('master-body').getAttribute("refresh"); |
|---|
| 208 | if (refresh_next < 60 && refresh_next != 0) { |
|---|
| 209 | refresh_next = 60; |
|---|
| 210 | } |
|---|
| 211 | if (refresh_next != refresh_sec) { |
|---|
| 212 | refresh_sec = refresh_next; |
|---|
| 213 | clearInterval(refresh_interval); |
|---|
| 214 | refresh_interval = setInterval(refresh, refresh_sec * 1000); |
|---|
| [378] | 215 | console.log("Update at", datenow.toLocaleTimeString(), "Refresh Interval:", refresh_sec); |
|---|
| [371] | 216 | document.getElementById('master-body').setAttribute("refresh", refresh_sec); |
|---|
| [370] | 217 | } |
|---|
| [369] | 218 | } |
|---|
| 219 | } |
|---|
| 220 | req.send(null); |
|---|
| 221 | } |
|---|
| 222 | |
|---|
| 223 | function init() { // This is the function the browser first runs when it's loaded. |
|---|
| [370] | 224 | refresh_sec = document.getElementById('master-body').getAttribute("refresh"); |
|---|
| 225 | if (refresh_sec < 60 && refresh_sec != 0) { |
|---|
| 226 | refresh_sec = 60; |
|---|
| 227 | } |
|---|
| [373] | 228 | // Set the refresh() function to run every 900 seconds. 1 second would be 1000 |
|---|
| 229 | refresh_interval = setInterval(refresh, refresh_sec * 1000); |
|---|
| [378] | 230 | console.log("Start Refresh Interval:", refresh_sec); |
|---|
| [369] | 231 | } |
|---|
| 232 | </script> |
|---|
| [240] | 233 | </head> |
|---|
| [369] | 234 | ENDH |
|---|
| 235 | |
|---|
| 236 | $htmlpage .= <<"ENDH"; |
|---|
| 237 | <body id="master-body" onload="init()" refresh="$config->{'refresh'}"> |
|---|
| [274] | 238 | <div class="header"> |
|---|
| 239 | <h1> |
|---|
| 240 | <ul> |
|---|
| 241 | <li>Nagios Velvice Alert Panel : <a href="$config->{'nagios-server'}{'portal-url'}">Core Server</a></li> |
|---|
| [374] | 242 | <li><small><a id="refresh" href="$cgi_script_name">$date_min</a></small></li> |
|---|
| [274] | 243 | </ul> |
|---|
| 244 | </h1> |
|---|
| 245 | </div> |
|---|
| [240] | 246 | ENDH |
|---|
| 247 | |
|---|
| [254] | 248 | my %service_name = (); |
|---|
| 249 | my %service_status = (); |
|---|
| [240] | 250 | for my $srv (@serviceproblems) { |
|---|
| 251 | $service_name{$srv->service_description}++; |
|---|
| [254] | 252 | $service_status{$srv->status}++; |
|---|
| [240] | 253 | } |
|---|
| 254 | |
|---|
| 255 | if (scalar @serviceproblems == 0) { |
|---|
| 256 | $htmlpage .= "<p>No alert to recheck.</p>\n"; |
|---|
| 257 | } |
|---|
| 258 | else { |
|---|
| 259 | |
|---|
| 260 | $htmlpage .= "<p>Alert to recheck - Level:\n"; |
|---|
| 261 | $htmlpage .= join ",\n", |
|---|
| [283] | 262 | " <span class='button'><a href='$cgi_script_name?check=all'>ALL</a><small>" . scalar(@serviceproblems) . '</small></span>', |
|---|
| 263 | map(" <span class='button'><a href='$cgi_script_name?check=" . lc(uri_encode($_)) . "'>$_</a><small>$service_status{$_}</small></span>", |
|---|
| 264 | sort keys %service_status); |
|---|
| [240] | 265 | $htmlpage .= ".\n"; |
|---|
| 266 | $htmlpage .= " <br />\n"; |
|---|
| 267 | $htmlpage .= " Service:\n"; |
|---|
| [279] | 268 | $htmlpage .= join ",\n", |
|---|
| [284] | 269 | map(" <span class='button'><a href='$cgi_script_name?check=" . lc(uri_encode($_)) . "'>" . nosbreak($_) . "</a><small>$service_name{$_}</small></span>", |
|---|
| [283] | 270 | sort keys %service_name); |
|---|
| [240] | 271 | $htmlpage .= ".\n"; |
|---|
| 272 | $htmlpage .= "</p>\n"; |
|---|
| 273 | |
|---|
| 274 | my $nagios_cmd; |
|---|
| [256] | 275 | open $nagios_cmd, '>>', $config->{'nagios-server'}{'nagios-cmd'} or die "Can't open file filename: $!"; |
|---|
| [240] | 276 | |
|---|
| [254] | 277 | my %remote_sshdown = (); |
|---|
| 278 | my %remote_db = (); |
|---|
| 279 | my $remote_flag; |
|---|
| [245] | 280 | |
|---|
| [240] | 281 | my $current_host = ''; |
|---|
| 282 | $htmlpage .= "<table border=\"1\">\n"; |
|---|
| [254] | 283 | SERVICE_PROBLEMS: |
|---|
| [240] | 284 | for my $srv (@serviceproblems) { |
|---|
| 285 | my $hostname = $srv->host_name; |
|---|
| 286 | my $service = $srv->service_description; |
|---|
| [245] | 287 | my $status = $srv->status; |
|---|
| [240] | 288 | my $downtime = downtime($srv->last_state_change); |
|---|
| [257] | 289 | my $output = HTML::Entities::encode($srv->plugin_output) =~ s/^[A-Z_\s]+?[:-]//r; |
|---|
| [240] | 290 | |
|---|
| [320] | 291 | my $color = alertcolor($status, $downtime); |
|---|
| [286] | 292 | my $stylecolor = "style='background:$color;'"; |
|---|
| 293 | $htmlpage .= " <tr>\n"; |
|---|
| [240] | 294 | if ($hostname ne $current_host) { |
|---|
| [319] | 295 | $current_host = $hostname; |
|---|
| 296 | my $rowspan = $hostcount{$hostname}->{'count'}; |
|---|
| 297 | my $rowcolor = "style='background:" . $hostcount{$hostname}->{'color'} . ";'"; |
|---|
| 298 | $htmlpage .= " <td $rowcolor rowspan='$rowspan'>" |
|---|
| [279] | 299 | . "<a href=\"$cgi_script_name?check=" . uri_encode($hostname) . '">↯</a></td>' . "\n"; |
|---|
| [319] | 300 | $htmlpage .= " <td $rowcolor class='hoop' rowspan='$rowspan'>" |
|---|
| [279] | 301 | . "<a href=\"$config->{'nagios-server'}{'status-cgi'}?host=" . uri_encode($hostname) . "\">$hostname</a></td>\n"; |
|---|
| [240] | 302 | } |
|---|
| [245] | 303 | |
|---|
| 304 | my $bold; |
|---|
| [254] | 305 | ACTION_STYLE: |
|---|
| [256] | 306 | for my $act_name (keys %{$config->{'remote-action'}}) { |
|---|
| 307 | my $act_regex = $config->{'remote-action'}{$act_name}{'regex'}; |
|---|
| 308 | $bold++ if $service =~ m/$act_regex/ and $config->{'remote-action'}{$act_name}{'style'} eq 'bold'; |
|---|
| [240] | 309 | } |
|---|
| [287] | 310 | $htmlpage .= $bold ? " <td $stylecolor class='hoop bold'>" : " <td $stylecolor class='hoop'>"; |
|---|
| [246] | 311 | $htmlpage .= "$service</td>\n"; |
|---|
| [245] | 312 | |
|---|
| [287] | 313 | $htmlpage .= " <td $stylecolor class='hoop'>$status</td>\n"; |
|---|
| 314 | $htmlpage .= " <td $stylecolor class='comment'>$output</td>\n"; |
|---|
| 315 | $htmlpage .= " <td $stylecolor class='days'>$downtime days</td>\n"; |
|---|
| [240] | 316 | |
|---|
| [257] | 317 | if (($cgi_check =~ m/all/i) |
|---|
| 318 | or ($cgi_check =~ m/^$service$/i) |
|---|
| 319 | or ($cgi_check =~ m/critical/i and $status eq 'CRITICAL') |
|---|
| 320 | or ($cgi_check =~ m/warning/i and $status eq 'WARNING') |
|---|
| 321 | or ($cgi_check =~ m/pending/i and $status eq 'PENDING') |
|---|
| [279] | 322 | or ($cgi_check eq $hostname and $status =~ m/^(CRITICAL|WARNING|PENDING)$/) |
|---|
| [240] | 323 | ) { |
|---|
| 324 | $now++; |
|---|
| [291] | 325 | my $interval = $srv->next_check() - $srv->last_check() || 300; # 5 * 60 = 300 |
|---|
| [240] | 326 | $interval = 240 if $interval < 240; |
|---|
| 327 | $interval = 3000 if $interval > 3000; |
|---|
| [291] | 328 | my $future = $now + 20 + int(rand($interval - 20)); |
|---|
| [240] | 329 | |
|---|
| [289] | 330 | $htmlpage .= " <td class='checking'>" . ($future - $now) . "</td>\n"; |
|---|
| [286] | 331 | #$htmlpage .= " -- <b>CHECK</b> [$now/" . ($future - $now) . "]"; |
|---|
| [240] | 332 | printf $nagios_cmd "[%lu] SCHEDULE_FORCED_SVC_CHECK;%s;%s;%lu\n", $now, $hostname, $service, $now; |
|---|
| 333 | # delay future command |
|---|
| 334 | push @futurecheck, sprintf "[%lu] SCHEDULE_FORCED_SVC_CHECK;%s;%s;%lu", $future, $hostname, $service, $future; |
|---|
| 335 | } |
|---|
| 336 | |
|---|
| [254] | 337 | ACTION_PUSH_AND_DEPEND: |
|---|
| [256] | 338 | for my $act_name (keys %{$config->{'remote-action'}}) { |
|---|
| 339 | my $act_regex = $config->{'remote-action'}{$act_name}{'regex'}; |
|---|
| 340 | my $act_status = $config->{'remote-action'}{$act_name}{'status'} || 'ALL'; |
|---|
| 341 | my $act_depend = $config->{'remote-action'}{$act_name}{'depend'} || 'SSH'; |
|---|
| [254] | 342 | |
|---|
| 343 | if ($service =~ m/$act_regex/ and ($act_status eq 'ALL' or $status =~ m/$act_status/)) { |
|---|
| 344 | $remote_db{$act_name} ||= []; |
|---|
| 345 | push @{$remote_db{$act_name}}, $hostname; |
|---|
| 346 | $remote_flag++; |
|---|
| [245] | 347 | } |
|---|
| [254] | 348 | |
|---|
| 349 | # check depend service otherwise |
|---|
| 350 | $remote_sshdown{$act_depend} ||= {}; |
|---|
| 351 | $remote_sshdown{$act_depend}->{$hostname}++ if $service =~ m/$act_depend/; |
|---|
| [245] | 352 | } |
|---|
| [240] | 353 | |
|---|
| 354 | $htmlpage .= " </tr>\n"; |
|---|
| 355 | } |
|---|
| 356 | |
|---|
| 357 | $htmlpage .= "</table>\n"; |
|---|
| 358 | close $nagios_cmd; |
|---|
| 359 | |
|---|
| [254] | 360 | # host down |
|---|
| [240] | 361 | if (%hostdown) { |
|---|
| 362 | $htmlpage .= "<br />\n"; |
|---|
| 363 | $htmlpage .= "<table border='1'>\n"; |
|---|
| [254] | 364 | HOST_DOWN: |
|---|
| [240] | 365 | for my $host (sort keys %hostdown) { |
|---|
| 366 | my $host_stat = $hostdown{$host}; |
|---|
| 367 | my $hostname = $host_stat->host_name; |
|---|
| 368 | my $downtime = downtime($host_stat->last_state_change); |
|---|
| [320] | 369 | my $color = alertcolor('CRITICAL', $downtime); |
|---|
| [382] | 370 | my $stylecolor = "style='background:$color;'"; |
|---|
| 371 | $htmlpage .= " <tr>\n"; |
|---|
| 372 | $htmlpage .= " <td $stylecolor>" . "<a href=\"$cgi_script_name?check=" . uri_encode($hostname) . '">↯</a></td>' . "\n"; |
|---|
| 373 | $htmlpage .= " <td $stylecolor class='hoop'><a href=\"$config->{'nagios-server'}{'status-cgi'}?host=" . uri_encode($hostname) . "\">$hostname</a></td>\n"; |
|---|
| [240] | 374 | my @host_service; |
|---|
| 375 | for my $srv ($log->list_services_on_host($host)) { |
|---|
| [382] | 376 | my $service = $log->service($host, $srv)->service_description; |
|---|
| 377 | push @host_service, $service; |
|---|
| 378 | if ($cgi_check eq $hostname) { |
|---|
| 379 | $now++; |
|---|
| 380 | printf $nagios_cmd "[%lu] SCHEDULE_FORCED_SVC_CHECK;%s;%s;%lu\n", $now, $hostname, $service, $now; |
|---|
| 381 | } |
|---|
| [240] | 382 | } |
|---|
| [382] | 383 | $htmlpage .= " <td $stylecolor><small>" . join(', ', @host_service) . "</small></td>\n"; |
|---|
| 384 | $htmlpage .= " <td $stylecolor style='text-align:right;'>$downtime days</td>\n"; |
|---|
| 385 | $htmlpage .= " <td class='checking'></td>\n" if $cgi_check eq $hostname; |
|---|
| [240] | 386 | $htmlpage .= " </tr>\n"; |
|---|
| 387 | } |
|---|
| 388 | $htmlpage .= "</table>\n"; |
|---|
| 389 | } |
|---|
| 390 | |
|---|
| [254] | 391 | # remote action |
|---|
| 392 | if ($remote_flag) { |
|---|
| [240] | 393 | require Nagios::Object::Config; |
|---|
| 394 | my $parser = Nagios::Object::Config->new(); |
|---|
| 395 | $parser->parse("/var/cache/nagios3/objects.cache"); |
|---|
| 396 | |
|---|
| [277] | 397 | $htmlpage .= "<div class='action'>\n"; |
|---|
| [254] | 398 | REMOTE_ACTION: |
|---|
| 399 | for my $act_name (keys %remote_db) { |
|---|
| [256] | 400 | my $act_depend = $config->{'remote-action'}{$act_name}{'depend'} || 'SSH'; |
|---|
| [254] | 401 | |
|---|
| 402 | my @action = grep !exists $remote_sshdown{$act_depend}->{$_}, @{$remote_db{$act_name}}; |
|---|
| [245] | 403 | if (@action) { |
|---|
| [256] | 404 | my $srv_title = $config->{'remote-action'}{$act_name}{'title'} || "Action: $act_name"; |
|---|
| [247] | 405 | $htmlpage .= "<h2>$srv_title</h2>\n"; |
|---|
| [245] | 406 | $htmlpage .= "<pre>\n"; |
|---|
| [256] | 407 | my $remote_action = $config->{'remote-action'}{$act_name}{'command'}; |
|---|
| 408 | $remote_action = $config->{'remote-action'}{$act_name}{'command-one'} |
|---|
| 409 | if @action == 1 and exists $config->{'remote-action'}{$act_name}{'command-one'}; |
|---|
| [245] | 410 | my @hosts; |
|---|
| 411 | for my $host (@action) { |
|---|
| 412 | my $object = $parser->find_object("$host", "Nagios::Host"); |
|---|
| 413 | push @hosts, hostmapping($object->address =~ s/\..*$//r); |
|---|
| 414 | } |
|---|
| 415 | my $hosts_list = join ' ', @hosts; |
|---|
| [246] | 416 | $htmlpage .= ' ' . $remote_action =~ s{\%m}{$hosts_list}r; |
|---|
| [245] | 417 | $htmlpage .= "</pre>\n"; |
|---|
| 418 | } |
|---|
| 419 | } |
|---|
| [277] | 420 | $htmlpage .= "</div>\n"; |
|---|
| [240] | 421 | } |
|---|
| 422 | } |
|---|
| 423 | |
|---|
| [269] | 424 | $htmlpage .= <<"ENDH"; |
|---|
| [369] | 425 | <hr clear="all" /> |
|---|
| [274] | 426 | <div class="footer"> |
|---|
| 427 | <b><a href="http://servforge.legi.grenoble-inp.fr/projects/soft-trokata/wiki/SoftWare/NagiosVelvice">Velvice</a> |
|---|
| 428 | - version: $VERSION</b> |
|---|
| 429 | (<a href="http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/velvice.html">online manual</a>) |
|---|
| 430 | - Written by Gabriel Moreau |
|---|
| 431 | <ul> |
|---|
| 432 | <li>Licence GNU GPL version 2 or later and Perl equivalent</li> |
|---|
| [381] | 433 | <li>Copyright (C) 2014-2019, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France</li> |
|---|
| [274] | 434 | </ul> |
|---|
| 435 | </div> |
|---|
| [240] | 436 | </body> |
|---|
| [369] | 437 | ENDH |
|---|
| 438 | |
|---|
| 439 | $htmlpage .= <<"ENDH" if not $cgi_only; |
|---|
| [240] | 440 | </html> |
|---|
| 441 | ENDH |
|---|
| 442 | |
|---|
| 443 | print $htmlpage; |
|---|
| 444 | |
|---|
| [254] | 445 | # delayed future check |
|---|
| [240] | 446 | if (@futurecheck) { |
|---|
| 447 | sleep 2; |
|---|
| 448 | my $nagios_cmd; |
|---|
| [256] | 449 | open $nagios_cmd, '>>', $config->{'nagios-server'}{'nagios-cmd'} or die "Can't open file filename: $!"; |
|---|
| [240] | 450 | print $nagios_cmd "$_\n" for @futurecheck; |
|---|
| 451 | close $nagios_cmd; |
|---|
| 452 | } |
|---|
| 453 | |
|---|
| 454 | __END__ |
|---|
| 455 | |
|---|
| 456 | |
|---|
| [241] | 457 | =head1 NAME |
|---|
| 458 | |
|---|
| 459 | velvice.cgi - nagios velvice alert panel |
|---|
| 460 | |
|---|
| [265] | 461 | =head1 USAGE |
|---|
| [241] | 462 | |
|---|
| [265] | 463 | velvice.cgi |
|---|
| 464 | velvice.cgi?check=XXX |
|---|
| 465 | |
|---|
| 466 | |
|---|
| [241] | 467 | =head1 DESCRIPTION |
|---|
| 468 | |
|---|
| [275] | 469 | =begin html |
|---|
| 470 | |
|---|
| 471 | <img width="700" alt="Nagios Velvice Alert Panel" title="Nagios Velvice Alert Panel" style="float:right" src="velvice.png" /> |
|---|
| 472 | |
|---|
| 473 | =end html |
|---|
| 474 | |
|---|
| [241] | 475 | Nagios VELVICE is an acronym for "Nagios leVEL serVICE status". |
|---|
| 476 | |
|---|
| [265] | 477 | The Nagios web page is sometimes very graphically charged |
|---|
| 478 | and does not necessarily contain the information you need at a glance. |
|---|
| 479 | For example, it is quite complicated to restart controls on multiple hosts in one click. |
|---|
| [241] | 480 | |
|---|
| [265] | 481 | For example, a server that is down should take only one line and not one per service... |
|---|
| 482 | Similarly, a service that has been down for 5 minutes or since yesterday |
|---|
| 483 | has more weight than a service that has fallen for 15 days. |
|---|
| [241] | 484 | |
|---|
| [265] | 485 | With Velvice Panel, a broken down server takes only one line. |
|---|
| 486 | Services that have been falling for a long time gradually lose their color and become pastel colors. |
|---|
| [241] | 487 | |
|---|
| [265] | 488 | With Velvice Panel, it is possible through a single click |
|---|
| 489 | to redo a check of all services that are in the CRITICAL state. |
|---|
| 490 | Similarly, it is possible to restart a check on all SSH services in breakdowns ... |
|---|
| 491 | In order not to clog the Nagios server, checks are shifted by 2 seconds in time. |
|---|
| 492 | |
|---|
| 493 | There is also a link to the web page of the main Nagios server. |
|---|
| 494 | For each computer, you have a direct link to its dedicated web page on this server. |
|---|
| 495 | |
|---|
| 496 | |
|---|
| 497 | =head1 CONFIGURATION FILE SPECIFICATION |
|---|
| 498 | |
|---|
| 499 | The configuration file must be F</etc/nagios3/velvice.yml>. |
|---|
| 500 | This is not a required file. |
|---|
| 501 | The file is in YAML format because this is a human-readable text file style. |
|---|
| 502 | Other formats could have been Plain XML, RDF, JSON... but they are much less readable. |
|---|
| 503 | |
|---|
| 504 | You can find in the software nagios-velvice an example of configuration: |
|---|
| 505 | L<velvice.sample.yml|http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/velvice.sample.yml>. |
|---|
| 506 | This one is in fact the master reference specification! |
|---|
| 507 | |
|---|
| 508 | The main keys C<nagios-server> and C<color-downtime> have good default values. |
|---|
| 509 | No secondary key is required... |
|---|
| [268] | 510 | The Velvice script try hard to replace ~ by the good value automatically. |
|---|
| [265] | 511 | |
|---|
| [268] | 512 | nagios-server: |
|---|
| 513 | status-file: /var/cache/nagios3/status.dat |
|---|
| 514 | nagios-cmd: /var/lib/nagios3/rw/nagios.cmd |
|---|
| 515 | portal-url: ~/nagios3/ |
|---|
| 516 | status-cgi: ~/cgi-bin/nagios3/status.cgi |
|---|
| 517 | stylesheets: ~/nagios3/stylesheets |
|---|
| 518 | |
|---|
| 519 | The background color of the faulty service line display remains stable with a bright color for at least 3 days. |
|---|
| 520 | Then, it decreases and becomes pastel after 53 days with an intensity of 70% (100% is white and 0% is black). |
|---|
| 521 | |
|---|
| 522 | color-downtime: |
|---|
| 523 | day-min: 3 |
|---|
| 524 | day-max: 50 |
|---|
| 525 | factor: 0.7 |
|---|
| 526 | |
|---|
| [265] | 527 | With key C<host-mapping>, |
|---|
| 528 | it's good to map C<localhost> to the real name of the computer (hostname). |
|---|
| 529 | |
|---|
| [268] | 530 | host-mapping: |
|---|
| 531 | localhost: srv-nagios |
|---|
| 532 | toto: titi |
|---|
| 533 | |
|---|
| [265] | 534 | The only important key is C<remote-action>. |
|---|
| 535 | You can affiliate as many subkeys as you want. |
|---|
| 536 | Let's take an example: |
|---|
| 537 | |
|---|
| 538 | remote-action: |
|---|
| 539 | oom-killer: |
|---|
| 540 | regex: ^OOM Killer |
|---|
| 541 | title: OOM Killer |
|---|
| 542 | command: tssh -c 'sudo rm /var/lib/nagios3/nagios_oom_killer.log' %m |
|---|
| 543 | command-one: ssh %m 'sudo rm /var/lib/nagios3/nagios_oom_killer.log' |
|---|
| 544 | depend: ^SSH |
|---|
| 545 | status: ALL |
|---|
| 546 | style: bold |
|---|
| 547 | |
|---|
| 548 | C<oom-killer> is just a key for your remote action. |
|---|
| 549 | The regex is used to find which service has a problem... |
|---|
| 550 | The title is use in the result web page (not mandatory - otherwise, it will be C<Action: oom-killer>). |
|---|
| 551 | The C<command> is just written on this web page. |
|---|
| 552 | You have the responsibility to copy / cut it on a terminal. |
|---|
| 553 | For security reasons, the nagios server does not have the right to launch the command on the remote host. |
|---|
| 554 | The wildcard C<%m> is replaced by the list of the host (separated by the space). |
|---|
| 555 | Sometime, the command could be different if there is only one computer (just SSH and no parallel SSH). |
|---|
| 556 | If your command is based on SSH, |
|---|
| 557 | you can have an SSH action only if the remote SSH is running. |
|---|
| 558 | So you can make the remote action depend on the SSH service through a regular expression of your choice. |
|---|
| 559 | |
|---|
| 560 | The last two keys. |
|---|
| 561 | The C<status> key is for CRITICAL or WARNING (or ALL). |
|---|
| 562 | The key C<style> is there to mark in bold the service in error on the web page. |
|---|
| 563 | |
|---|
| 564 | =head1 SEE ALSO |
|---|
| 565 | |
|---|
| 566 | yamllint(1), ysh(1), YAML, Nagios::StatusLog, Color::Calc |
|---|
| 567 | |
|---|
| 568 | In Debian GNU/Linux distribution, packages for C<yamllint> and C<ysh> are: |
|---|
| 569 | |
|---|
| 570 | =over |
|---|
| 571 | |
|---|
| 572 | =item * C<yamllint> - Linter for YAML files (Python) |
|---|
| 573 | |
|---|
| 574 | =item * C<libyaml-shell-perl> - YAML test shell (Perl) |
|---|
| 575 | |
|---|
| 576 | =back |
|---|
| 577 | |
|---|
| 578 | |
|---|
| 579 | Own project ressources: |
|---|
| 580 | |
|---|
| 581 | =over |
|---|
| 582 | |
|---|
| [293] | 583 | =item * L<Web Site|http://servforge.legi.grenoble-inp.fr/projects/soft-trokata/wiki/SoftWare/NagiosVelvice> |
|---|
| [265] | 584 | |
|---|
| 585 | =item * L<Online Manual|http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/velvice.html> |
|---|
| 586 | |
|---|
| [293] | 587 | =item * L<SVN Repository|http://servforge.legi.grenoble-inp.fr/svn/soft-trokata/trunk/nagios-velvice> |
|---|
| [265] | 588 | |
|---|
| [293] | 589 | =item * L<Debian Package|http://servforge.legi.grenoble-inp.fr/pub/soft-trokata/nagios-velvice/download/> |
|---|
| 590 | |
|---|
| [265] | 591 | =back |
|---|
| 592 | |
|---|
| 593 | |
|---|
| 594 | =head1 VERSION |
|---|
| 595 | |
|---|
| [266] | 596 | $Id: velvice.cgi 382 2019-01-11 10:06:09Z g7moreau $ |
|---|
| [265] | 597 | |
|---|
| 598 | |
|---|
| 599 | =head1 AUTHOR |
|---|
| 600 | |
|---|
| 601 | Written by Gabriel Moreau <Gabriel.Moreau(A)univ-grenoble-alpes.fr>, LEGI UMR 5519, CNRS, Grenoble - France |
|---|
| 602 | |
|---|
| 603 | |
|---|
| [241] | 604 | =head1 LICENSE AND COPYRIGHT |
|---|
| 605 | |
|---|
| 606 | Licence GNU GPL version 2 or later and Perl equivalent |
|---|
| 607 | |
|---|
| [381] | 608 | Copyright (C) 2014-2019, LEGI UMR 5519 / CNRS UGA G-INP, Grenoble, France |
|---|