| 1 | # -*- perl -*- | ||
| 2 | |||
| 3 | # Copyright (c) 2009 AdCopy | ||
| 4 | # Author: Jeff Weisberg | ||
| 5 | # Created: 2009-Apr-02 11:16 (EDT) | ||
| 6 | # Function: track status of peers | ||
| 7 | # | ||
| 8 | # $Id: Status.pm,v 1.38 2010/10/30 19:46:16 jaw Exp $ | ||
| 9 | |||
| 10 | package AC::Yenta::Status; | ||
| 11 | use AC::Yenta::Kibitz::Status; | ||
| 12 | use AC::Yenta::Debug 'status'; | ||
| 13 | use AC::Yenta::Config; | ||
| 14 | use AC::Yenta::MySelf; | ||
| 15 | use AC::Misc; | ||
| 16 | use Sys::Hostname; | ||
| 17 | use JSON; | ||
| 18 | use Socket; | ||
| 19 | require 'AC/protobuf/yenta_status.pl'; | ||
| 20 | use strict; | ||
| 21 | |||
| 22 | my $KEEPDOWN = 1800; # keep data about down servers for how long? | ||
| 23 | my $KEEPLOST = 600; # keep data about servers we have not heard about for how long? | ||
| 24 | my $SAVEMAX = 1800; # do not save if older than | ||
| 25 | |||
| 26 | my $PORT; | ||
| 27 | |||
| 28 | our $DATA = bless { | ||
| 29 | allpeer => {}, # yenta_status | ||
| 30 | sceptical => {}, | ||
| 31 | mappeer => {}, # {map} => { id => id } | ||
| 32 | peermap => {}, # {id} => @map | ||
| 33 | datacenter => {}, # {dc} => { id => id } | ||
| 34 | peertype => {}, # {ss} => { id => id } | ||
| 35 | }; | ||
| 36 | |||
| 37 | sub init { | ||
| 38 | my $port = shift; | ||
| 39 | |||
| 40 | $PORT = $port; | ||
| 41 | |||
| 42 | AC::DC::Sched->new( | ||
| 43 | info => 'kibitz status', | ||
| 44 | freq => (conf_value('time_status_kibitz') || 5), | ||
| 45 | func => \&periodic, | ||
| 46 | ); | ||
| 47 | AC::DC::Sched->new( | ||
| 48 | info => 'save status', | ||
| 49 | freq => (conf_value('time_status_save') || 15), | ||
| 50 | func => \&save_status, | ||
| 51 | ); | ||
| 52 | } | ||
| 53 | |||
| 54 | # start up a client every so often | ||
| 55 | sub periodic { | ||
| 56 | |||
| 57 | # clean up down or lost peers | ||
| 58 | for my $id ( keys %{$DATA->{allpeer}} ){ | ||
| 59 | my $p = $DATA->{allpeer}{$id}; | ||
| 60 | next unless $p; | ||
| 61 | |||
| 62 | next if $p->{status} == 200 && $p->{timestamp} > $^T - $KEEPLOST; | ||
| 63 | _maybe_remove( $id ); | ||
| 64 | } | ||
| 65 | |||
| 66 | # randomly pick a peer | ||
| 67 | my($id, $ip, $port) = _random_peer(); | ||
| 68 | return unless $id; | ||
| 69 | |||
| 70 | # start a client | ||
| 71 | debug("starting status kibitz client to $id"); | ||
| 72 | |||
| 73 | my $c = AC::Yenta::Kibitz::Status::Client->new( $ip, $port, | ||
| 74 | info => "status client: $id", | ||
| 75 | status_peer => $id, | ||
| 76 | ); | ||
| 77 | __PACKAGE__->isdown($id) unless $c; | ||
| 78 | |||
| 79 | $c->start(); | ||
| 80 | } | ||
| 81 | |||
| 82 | sub _random_peer { | ||
| 83 | |||
| 84 | # sceptical | ||
| 85 | my @peer = values %{$DATA->{sceptical}}; | ||
| 86 | |||
| 87 | # known peer | ||
| 88 | unless( @peer ){ | ||
| 89 | @peer = map { $DATA->{allpeer}{$_} } keys %{$DATA->{peertype}{yenta}}; | ||
| 90 | } | ||
| 91 | |||
| 92 | if( @peer ){ | ||
| 93 | my $p = $peer[ rand(@peer) ]; | ||
| 94 | debug("using peer $p->{server_id}"); | ||
| 95 | return ($p->{server_id}, $p->{ip}, undef) | ||
| 96 | if int rand(@peer+1); # sometimes use the seed | ||
| 97 | } | ||
| 98 | |||
| 99 | # seed peer | ||
| 100 | my $seed = conf_value('seedpeer'); | ||
| 101 | my $p = $seed->[ rand(@$seed) ]; | ||
| 102 | my ($ip, $port) = split /:/, $p; | ||
| 103 | $port ||= my_port(); | ||
| 104 | |||
| 105 | # don't talk to self. any of my addrs. | ||
| 106 | my $ipinfo = my_network_info(); | ||
| 107 | for my $i (@$ipinfo){ | ||
| 108 | return if $ip eq $i->{ipa} && $port == $PORT; | ||
| 109 | } | ||
| 110 | |||
| 111 | return("seed/$ip:$port", $ip, $port); | ||
| 112 | } | ||
| 113 | |||
| 114 | # server list for save file | ||
| 115 | sub server_list { | ||
| 116 | my $type = shift; | ||
| 117 | |||
| 118 | ($type, my $where) = split m|/|, $type; | ||
| 119 | # where - no longer used | ||
| 120 | $where ||= my_datacenter(); | ||
| 121 | |||
| 122 | my @peer = keys %{ $DATA->{peertype}{$type} }; | ||
| 123 | return unless @peer; | ||
| 124 | |||
| 125 | # nothing too old | ||
| 126 | @peer = grep { $DATA->{allpeer}{$_}{lastup} > $^T - $SAVEMAX } @peer; | ||
| 127 | return unless @peer; | ||
| 128 | |||
| 129 | return map { $DATA->{allpeer}{$_} } @peer; | ||
| 130 | } | ||
| 131 | |||
| 132 | # save a list of peers, in case I crash, and for others to use | ||
| 133 | sub save_status { | ||
| 134 | |||
| 135 | my $save = conf_value('savestatus'); | ||
| 136 | my $here = my_datacenter(); | ||
| 137 | |||
| 138 | # also save locally running services | ||
| 139 | my @mon = AC::Yenta::Monitor::export(); | ||
| 140 | |||
| 141 | for my $s ( @$save ){ | ||
| 142 | my $file = $s->{file}; | ||
| 143 | my $types = $s->{type}; | ||
| 144 | |||
| 145 | my @peer; | ||
| 146 | for my $type (@$types){ | ||
| 147 | push @peer, server_list($type); | ||
| 148 | |||
| 149 | for my $m (@mon){ | ||
| 150 | push @peer, $m if $m->{subsystem} eq $type; | ||
| 151 | } | ||
| 152 | } | ||
| 153 | |||
| 154 | next unless @peer; | ||
| 155 | |||
| 156 | debug("saving peer status file"); | ||
| 157 | unless( open(FILE, ">$file.tmp") ){ | ||
| 158 | problem("cannot open save file '$file.tmp': $!"); | ||
| 159 | return; | ||
| 160 | } | ||
| 161 | |||
| 162 | for my $pd (@peer){ | ||
| 163 | # only save best addr in save file | ||
| 164 | my($ip, $port) = AC::Yenta::IO::TCP::Client->use_addr_port( $pd->{ip} ); | ||
| 165 | |||
| 166 | my $data = { | ||
| 167 | id => $pd->{server_id}, | ||
| 168 | addr => $ip, | ||
| 169 | port => int($port), | ||
| 170 | status => int($pd->{status}), | ||
| 171 | subsystem => $pd->{subsystem}, | ||
| 172 | environment => $pd->{environment}, | ||
| 173 | sort_metric => int($pd->{sort_metric}), | ||
| 174 | is_local => ($here eq $pd->{datacenter} ? 1 : 0), | ||
| 175 | }; | ||
| 176 | if( $pd->{subsystem} eq 'yenta' ){ | ||
| 177 | $data->{map} = $pd->{map}; | ||
| 178 | } | ||
| 179 | |||
| 180 | print FILE encode_json( $data ), "\n"; | ||
| 181 | } | ||
| 182 | |||
| 183 | close FILE; | ||
| 184 | unless( rename("$file.tmp", $file) ){ | ||
| 185 | problem("cannot rename save file '$file': $!"); | ||
| 186 | } | ||
| 187 | |||
| 188 | } | ||
| 189 | } | ||
| 190 | |||
| 191 | ################################################################ | ||
| 192 | # diagnostic reports | ||
| 193 | sub report { | ||
| 194 | |||
| 195 | my $res; | ||
| 196 | |||
| 197 | for my $v (AC::Yenta::Kibitz::Status::_myself(), AC::Yenta::Monitor::export(), values %{$DATA->{allpeer}} ){ | ||
| 198 | my $id = sprintf '%-28s', $v->{server_id}; | ||
| 199 | my $metric = int( $v->{sort_metric} ); | ||
| 200 | $res .= "$id $v->{hostname}\t$v->{datacenter}\t$v->{subsystem}\t$v->{environment}\t$v->{status}\t$metric\n"; | ||
| 201 | } | ||
| 202 | |||
| 203 | return $res; | ||
| 204 | } | ||
| 205 | |||
| 206 | sub report_long { | ||
| 207 | |||
| 208 | my $res; | ||
| 209 | |||
| 210 | # for my $v (AC::Yenta::Kibitz::Status::_myself(), AC::Yenta::Monitor::export(), values %{$DATA->{allpeer}} ){ | ||
| 211 | # $res .= dumper( $v ) . "\n\n"; | ||
| 212 | # } | ||
| 213 | return $res; | ||
| 214 | } | ||
| 215 | ################################################################ | ||
| 216 | |||
| 217 | sub my_port { $PORT } | ||
| 218 | |||
| 219 | |||
| 220 | sub my_instance_id { | ||
| 221 | my $class = shift; | ||
| 222 | return my_server_id() . sprintf('%04x', $$); | ||
| 223 | } | ||
| 224 | |||
| 225 | sub peer { | ||
| 226 | my $class = shift; | ||
| 227 | my $id = shift; | ||
| 228 | |||
| 229 | return $DATA->{allpeer}{$id}; | ||
| 230 | } | ||
| 231 | |||
| 232 | sub allpeers { | ||
| 233 | my $class = shift; | ||
| 234 | |||
| 235 | return values %{$DATA->{allpeer}}; | ||
| 236 | } | ||
| 237 | |||
| 238 | sub mappeers { | ||
| 239 | my $class = shift; | ||
| 240 | my $map = shift; | ||
| 241 | |||
| 242 | return keys %{ $DATA->{mappeer}{$map} }; | ||
| 243 | } | ||
| 244 | |||
| 245 | sub datacenters { | ||
| 246 | my $class = shift; | ||
| 247 | |||
| 248 | return $DATA->{datacenter}; | ||
| 249 | } | ||
| 250 | ################################################################ | ||
| 251 | |||
| 252 | sub _remove { | ||
| 253 | my $id = shift; | ||
| 254 | |||
| 255 | my $ss = $DATA->{allpeer}{$id}{subsystem}; | ||
| 256 | delete $DATA->{peertype}{$ss}{$id} if $ss; | ||
| 257 | |||
| 258 | my $dc = $DATA->{allpeer}{$id}{datacenter}; | ||
| 259 | delete $DATA->{datacenter}{$dc}{$id} if $dc; | ||
| 260 | |||
| 261 | verbose("deleting peer: $id"); | ||
| 262 | delete $DATA->{allpeer}{$id}; | ||
| 263 | |||
| 264 | # remove map info | ||
| 265 | for my $map ( @{$DATA->{peermap}{$id}} ){ | ||
| 266 | delete $DATA->{mappeer}{$map}{$id}; | ||
| 267 | } | ||
| 268 | delete $DATA->{peermap}{$id}; | ||
| 269 | |||
| 270 | # delete its monitored items | ||
| 271 | for my $p (keys %{$DATA->{allpeer}}){ | ||
| 272 | next unless $DATA->{allpeer}{$p}{via} eq $id; | ||
| 273 | _remove($p); | ||
| 274 | } | ||
| 275 | } | ||
| 276 | |||
| 277 | sub _maybe_remove { | ||
| 278 | my $id = shift; | ||
| 279 | |||
| 280 | my $d = $DATA->{allpeer}{$id}; | ||
| 281 | |||
| 282 | if( ($^T - $d->{lastup} > $KEEPDOWN) || ($^T - $d->{timestamp} > $KEEPLOST) ){ | ||
| 283 | |||
| 284 | _remove($id); | ||
| 285 | } | ||
| 286 | } | ||
| 287 | |||
| 288 | sub isdown { | ||
| 289 | my $class = shift; | ||
| 290 | my $id = shift; | ||
| 291 | |||
| 292 | debug("marking peer '$id' as down"); | ||
| 293 | |||
| 294 | if( ! $DATA->{allpeer}{$id} ){ | ||
| 295 | return unless $DATA->{sceptical}{$id}; | ||
| 296 | $DATA->{allpeer}{$id} = $DATA->{sceptical}{$id}; | ||
| 297 | delete $DATA->{sceptical}{$id}; | ||
| 298 | } | ||
| 299 | |||
| 300 | if( $DATA->{allpeer}{$id} ){ | ||
| 301 | $DATA->{allpeer}{$id}{timestamp} = $^T; | ||
| 302 | $DATA->{allpeer}{$id}{status} = 0; | ||
| 303 | $DATA->{allpeer}{$id}{path} = my_server_id(); | ||
| 304 | } | ||
| 305 | _maybe_remove( $id ); | ||
| 306 | } | ||
| 307 | |||
| 308 | ################################################################ | ||
| 309 | |||
| 310 | sub _env_ok { | ||
| 311 | my $class = shift; | ||
| 312 | my $id = shift; | ||
| 313 | my $up = shift; | ||
| 314 | |||
| 315 | # if( $up->{environment} ne conf_value('environment') ){ | ||
| 316 | # verbose("ignoring update from $id - wrong env: $up->{environment}"); | ||
| 317 | # return; | ||
| 318 | # } | ||
| 319 | return 1; | ||
| 320 | } | ||
| 321 | |||
| 322 | sub update_sceptical { | ||
| 323 | my $class = shift; | ||
| 324 | my $id = shift; # ->server_id | ||
| 325 | my $up = shift; | ||
| 326 | |||
| 327 | return unless $class->_env_ok($id, $up); | ||
| 328 | |||
| 329 | if( $DATA->{allpeer}{$id} ){ | ||
| 330 | # already known | ||
| 331 | delete $DATA->{sceptical}{$id}; | ||
| 332 | return; | ||
| 333 | } | ||
| 334 | |||
| 335 | debug("rcvd update (sceptical) about $id"); | ||
| 336 | $up->{id} = $id; | ||
| 337 | $DATA->{sceptical}{$id} = $up; | ||
| 338 | } | ||
| 339 | |||
| 340 | sub update { | ||
| 341 | my $class = shift; | ||
| 342 | my $id = shift; # -> server_id | ||
| 343 | my $up = shift; | ||
| 344 | |||
| 345 | return unless $class->_env_ok($id, $up); | ||
| 346 | |||
| 347 | # only keep it if it is relatively fresh | ||
| 348 | return unless $up->{timestamp} > $^T - $KEEPLOST; | ||
| 349 | |||
| 350 | $up->{id} = $id; | ||
| 351 | my $previnfo = $DATA->{allpeer}{$id}; | ||
| 352 | verbose("discovered new peer: $id ($up->{hostname})") unless $previnfo; | ||
| 353 | |||
| 354 | # only keep it if it is newer than what we have | ||
| 355 | return if $previnfo && $up->{timestamp} <= $previnfo->{timestamp}; | ||
| 356 | |||
| 357 | $up->{path} .= ' ' . my_server_id(); | ||
| 358 | |||
| 359 | # debug("updating $id => $up->{status} => " . dumper($up)); | ||
| 360 | debug("updating $id => $up->{status}"); | ||
| 361 | |||
| 362 | $DATA->{allpeer}{$id} = $up; | ||
| 363 | |||
| 364 | if( $up->{status} != 200 ){ | ||
| 365 | _maybe_remove( $id ); | ||
| 366 | return ; | ||
| 367 | } | ||
| 368 | |||
| 369 | # update datacenter info | ||
| 370 | unless( $DATA->{datacenter}{$up->{datacenter}}{$id} ){ | ||
| 371 | my $pdc = $previnfo->{datacenter}; | ||
| 372 | delete $DATA->{datacenter}{$pdc}{$id} if $pdc; | ||
| 373 | $DATA->{datacenter}{$up->{datacenter}}{$id} = $id; | ||
| 374 | } | ||
| 375 | |||
| 376 | # update subsystem info | ||
| 377 | unless( $DATA->{peertype}{$up->{subsystem}}{$id} ){ | ||
| 378 | my $ss = $previnfo->{subsystem}; | ||
| 379 | delete $DATA->{peertype}{$ss}{$id} if $ss; | ||
| 380 | $DATA->{peertype}{$up->{subsystem}}{$id} = $id; | ||
| 381 | } | ||
| 382 | |||
| 383 | # update map info | ||
| 384 | $DATA->{peermap}{$id} ||= []; | ||
| 385 | $up->{map} ||= []; | ||
| 386 | my @curmap = @{$DATA->{peermap}{$id}}; | ||
| 387 | my @newmap = sort @{$up->{map}}; | ||
| 388 | |||
| 389 | return if "@curmap" eq "@newmap"; # unchanged | ||
| 390 | |||
| 391 | # what do we need to add/remove | ||
| 392 | my (%remove, %add); | ||
| 393 | @remove{@curmap} = @curmap; | ||
| 394 | @add{@newmap} = @newmap; | ||
| 395 | delete $remove{$_} for @newmap; | ||
| 396 | delete $add{$_} for @curmap; | ||
| 397 | |||
| 398 | for my $map (keys %remove){ | ||
| 399 | debug("removing $map from $id"); | ||
| 400 | delete $DATA->{mappeer}{$map}{$id}; | ||
| 401 | } | ||
| 402 | for my $map (keys %add){ | ||
| 403 | debug("adding $map to $id"); | ||
| 404 | $DATA->{mappeer}{$map}{$id} = $id; | ||
| 405 | } | ||
| 406 | $DATA->{peermap}{$id} = \@newmap; | ||
| 407 | } | ||
| 408 |