Annotation of loncom/loncnew, revision 1.61
1.1 foxr 1: #!/usr/bin/perl
1.2 albertel 2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
1.61 ! foxr 5: # $Id: loncnew,v 1.60 2004/09/22 11:11:54 foxr Exp $
1.2 albertel 6: #
7: # Copyright Michigan State University Board of Trustees
8: #
9: # This file is part of the LearningOnline Network with CAPA (LON-CAPA).
1.17 foxr 10: ## LON-CAPA is free software; you can redistribute it and/or modify
1.2 albertel 11: # it under the terms of the GNU General Public License as published by
12: # the Free Software Foundation; either version 2 of the License, or
13: # (at your option) any later version.
14: #
15: # LON-CAPA is distributed in the hope that it will be useful,
16: # but WITHOUT ANY WARRANTY; without even the implied warranty of
17: # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18: # GNU General Public License for more details.
19: #
20: # You should have received a copy of the GNU General Public License
21: # along with LON-CAPA; if not, write to the Free Software
22: # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23: #
24: # /home/httpd/html/adm/gpl.txt
25: #
26: # http://www.lon-capa.org/
27: #
1.1 foxr 28: #
1.15 foxr 29: # new lonc handles n request out bver m connections to londs.
1.1 foxr 30: # This module is based on the Event class.
31: # Development iterations:
32: # - Setup basic event loop. (done)
33: # - Add timer dispatch. (done)
34: # - Add ability to accept lonc UNIX domain sockets. (done)
35: # - Add ability to create/negotiate lond connections (done).
1.7 foxr 36: # - Add general logic for dispatching requests and timeouts. (done).
37: # - Add support for the lonc/lond requests. (done).
1.38 foxr 38: # - Add logging/status monitoring. (done)
39: # - Add Signal handling - HUP restarts. USR1 status report. (done)
1.7 foxr 40: # - Add Configuration file I/O (done).
1.38 foxr 41: # - Add management/status request interface. (done)
1.8 foxr 42: # - Add deferred request capability. (done)
1.38 foxr 43: # - Detect transmission timeouts. (done)
1.7 foxr 44: #
45:
1.23 foxr 46: use strict;
1.1 foxr 47: use lib "/home/httpd/lib/perl/";
48: use Event qw(:DEFAULT );
49: use POSIX qw(:signal_h);
1.12 foxr 50: use POSIX;
1.1 foxr 51: use IO::Socket;
52: use IO::Socket::INET;
53: use IO::Socket::UNIX;
1.9 foxr 54: use IO::File;
1.6 foxr 55: use IO::Handle;
1.1 foxr 56: use Socket;
57: use Crypt::IDEA;
58: use LONCAPA::Queue;
59: use LONCAPA::Stack;
60: use LONCAPA::LondConnection;
1.7 foxr 61: use LONCAPA::LondTransaction;
1.1 foxr 62: use LONCAPA::Configuration;
63: use LONCAPA::HashIterator;
64:
65:
66: # Read the httpd configuration file to get perl variables
67: # normally set in apache modules:
68:
69: my $perlvarref = LONCAPA::Configuration::read_conf('loncapa.conf');
70: my %perlvar = %{$perlvarref};
71:
72: #
73: # parent and shared variables.
74:
75: my %ChildHash; # by pid -> host.
1.26 foxr 76: my %HostToPid; # By host -> pid.
77: my %HostHash; # by loncapaname -> IP.
1.1 foxr 78:
79:
1.9 foxr 80: my $MaxConnectionCount = 10; # Will get from config later.
1.1 foxr 81: my $ClientConnection = 0; # Uniquifier for client events.
82:
1.9 foxr 83: my $DebugLevel = 0;
1.29 foxr 84: my $NextDebugLevel= 2; # So Sigint can toggle this.
1.50 albertel 85: my $IdleTimeout= 600; # Wait 10 minutes before pruning connections.
1.1 foxr 86:
1.39 foxr 87: my $LogTransactions = 0; # When True, all transactions/replies get logged.
88:
1.1 foxr 89: #
90: # The variables below are only used by the child processes.
91: #
92: my $RemoteHost; # Name of host child is talking to.
1.20 albertel 93: my $UnixSocketDir= $perlvar{'lonSockDir'};
1.1 foxr 94: my $IdleConnections = Stack->new(); # Set of idle connections
95: my %ActiveConnections; # Connections to the remote lond.
1.7 foxr 96: my %ActiveTransactions; # LondTransactions in flight.
1.1 foxr 97: my %ActiveClients; # Serial numbers of active clients by socket.
98: my $WorkQueue = Queue->new(); # Queue of pending transactions.
99: my $ConnectionCount = 0;
1.4 foxr 100: my $IdleSeconds = 0; # Number of seconds idle.
1.9 foxr 101: my $Status = ""; # Current status string.
1.14 foxr 102: my $RecentLogEntry = "";
1.30 foxr 103: my $ConnectionRetries=2; # Number of connection retries allowed.
104: my $ConnectionRetriesLeft=2; # Number of connection retries remaining.
1.40 foxr 105: my $LondVersion = "unknown"; # Version of lond we talk with.
1.49 foxr 106: my $KeyMode = ""; # e.g. ssl, local, insecure from last connect.
1.54 foxr 107: my $LondConnecting = 0; # True when a connection is being built.
1.1 foxr 108:
1.60 foxr 109:
110: # DO NOT SET THE NEXT VARIABLE TO NON ZERO!!!!!!!!!!!!!!!
111:
1.57 foxr 112: my $DieWhenIdle = 0; # When true children die when trimmed -> 0.
113:
1.1 foxr 114: #
1.9 foxr 115: # The hash below gives the HTML format for log messages
116: # given a severity.
117: #
118: my %LogFormats;
119:
1.45 albertel 120: $LogFormats{"CRITICAL"} = "<font color='red'>CRITICAL: %s</font>";
121: $LogFormats{"SUCCESS"} = "<font color='green'>SUCCESS: %s</font>";
122: $LogFormats{"INFO"} = "<font color='yellow'>INFO: %s</font>";
123: $LogFormats{"WARNING"} = "<font color='blue'>WARNING: %s</font>";
1.9 foxr 124: $LogFormats{"DEFAULT"} = " %s ";
125:
1.10 foxr 126:
1.57 foxr 127: # UpdateStatus;
128: # Update the idle status display to show how many connections
129: # are left, retries and other stuff.
130: #
131: sub UpdateStatus {
132: if ($ConnectionRetriesLeft > 0) {
133: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
134: ." Retries remaining: ".$ConnectionRetriesLeft
135: ." ($KeyMode)");
136: } else {
137: ShowStatus(GetServerHost()." >> DEAD <<");
138: }
139: }
140:
1.10 foxr 141:
142: =pod
143:
144: =head2 LogPerm
145:
146: Makes an entry into the permanent log file.
147:
148: =cut
149: sub LogPerm {
150: my $message=shift;
151: my $execdir=$perlvar{'lonDaemons'};
152: my $now=time;
153: my $local=localtime($now);
154: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
155: print $fh "$now:$message:$local\n";
156: }
1.9 foxr 157:
158: =pod
159:
160: =head2 Log
161:
162: Logs a message to the log file.
163: Parameters:
164:
165: =item severity
166:
167: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
168: format string used to format the message. if the severity is
169: not a defined severity the Default format string is used.
170:
171: =item message
172:
173: The base message. In addtion to the format string, the message
174: will be appended to a string containing the name of our remote
175: host and the time will be formatted into the message.
176:
177: =cut
178:
179: sub Log {
1.47 foxr 180:
181: my ($severity, $message) = @_;
182:
1.9 foxr 183: if(!$LogFormats{$severity}) {
184: $severity = "DEFAULT";
185: }
186:
187: my $format = $LogFormats{$severity};
188:
189: # Put the window dressing in in front of the message format:
190:
191: my $now = time;
192: my $local = localtime($now);
193: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
194: my $finalformat = $finalformat.$format."\n";
195:
196: # open the file and put the result.
197:
198: my $execdir = $perlvar{'lonDaemons'};
199: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
200: my $msg = sprintf($finalformat, $message);
1.14 foxr 201: $RecentLogEntry = $msg;
1.9 foxr 202: print $fh $msg;
203:
1.10 foxr 204:
1.9 foxr 205: }
1.6 foxr 206:
1.3 albertel 207:
1.1 foxr 208: =pod
1.3 albertel 209:
210: =head2 GetPeerName
211:
212: Returns the name of the host that a socket object is connected to.
213:
1.1 foxr 214: =cut
215:
216: sub GetPeername {
1.47 foxr 217:
218:
219: my ($connection, $AdrFamily) = @_;
220:
1.1 foxr 221: my $peer = $connection->peername();
222: my $peerport;
223: my $peerip;
224: if($AdrFamily == AF_INET) {
225: ($peerport, $peerip) = sockaddr_in($peer);
1.23 foxr 226: my $peername = gethostbyaddr($peerip, $AdrFamily);
1.1 foxr 227: return $peername;
228: } elsif ($AdrFamily == AF_UNIX) {
229: my $peerfile;
230: ($peerfile) = sockaddr_un($peer);
231: return $peerfile;
232: }
233: }
234: =pod
1.3 albertel 235:
1.1 foxr 236: =head2 Debug
1.3 albertel 237:
238: Invoked to issue a debug message.
239:
1.1 foxr 240: =cut
1.3 albertel 241:
1.1 foxr 242: sub Debug {
1.47 foxr 243:
244: my ($level, $message) = @_;
245:
1.1 foxr 246: if ($level <= $DebugLevel) {
1.23 foxr 247: Log("INFO", "-Debug- $message host = $RemoteHost");
1.1 foxr 248: }
249: }
250:
251: sub SocketDump {
1.47 foxr 252:
253: my ($level, $socket) = @_;
254:
1.1 foxr 255: if($level <= $DebugLevel) {
1.48 foxr 256: $socket->Dump(-1); # Ensure it will get dumped.
1.1 foxr 257: }
258: }
1.3 albertel 259:
1.1 foxr 260: =pod
1.3 albertel 261:
1.5 foxr 262: =head2 ShowStatus
263:
264: Place some text as our pid status.
1.10 foxr 265: and as what we return in a SIGUSR1
1.5 foxr 266:
267: =cut
268: sub ShowStatus {
1.10 foxr 269: my $state = shift;
270: my $now = time;
271: my $local = localtime($now);
272: $Status = $local.": ".$state;
273: $0='lonc: '.$state.' '.$local;
1.5 foxr 274: }
275:
276: =pod
277:
1.15 foxr 278: =head 2 SocketTimeout
279:
280: Called when an action on the socket times out. The socket is
281: destroyed and any active transaction is failed.
282:
283:
284: =cut
285: sub SocketTimeout {
286: my $Socket = shift;
1.38 foxr 287: Log("WARNING", "A socket timeout was detected");
1.52 foxr 288: Debug(5, " SocketTimeout called: ");
1.48 foxr 289: $Socket->Dump(0);
1.42 foxr 290: if(exists($ActiveTransactions{$Socket})) {
1.43 albertel 291: FailTransaction($ActiveTransactions{$Socket});
1.42 foxr 292: }
1.22 foxr 293: KillSocket($Socket); # A transaction timeout also counts as
294: # a connection failure:
295: $ConnectionRetriesLeft--;
1.42 foxr 296: if($ConnectionRetriesLeft <= 0) {
1.52 foxr 297: Log("CRITICAL", "Host marked DEAD: ".GetServerHost());
1.56 foxr 298: $LondConnecting = 0;
1.42 foxr 299: }
300:
1.15 foxr 301: }
1.35 foxr 302: #----------------------------- Timer management ------------------------
1.15 foxr 303:
304: =pod
305:
1.1 foxr 306: =head2 Tick
1.3 albertel 307:
308: Invoked each timer tick.
309:
1.1 foxr 310: =cut
311:
1.5 foxr 312:
1.1 foxr 313: sub Tick {
1.52 foxr 314: my ($Event) = @_;
315: my $clock_watcher = $Event->w;
316:
1.1 foxr 317: my $client;
1.57 foxr 318: UpdateStatus();
319:
1.4 foxr 320: # Is it time to prune connection count:
321:
322:
323: if($IdleConnections->Count() &&
324: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
1.52 foxr 325: $IdleSeconds++;
1.4 foxr 326: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
1.23 foxr 327: my $Socket = $IdleConnections->pop();
1.6 foxr 328: KillSocket($Socket);
1.54 foxr 329: $IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
1.57 foxr 330: UpdateStatus();
331: if(($ConnectionCount == 0) && $DieWhenIdle) {
332: #
333: # Create a lock file since there will be a time window
334: # between our exit and the parent's picking up the listen
335: # during which no listens will be done on the
336: # lonnet client socket.
337: #
338: my $lock_file = GetLoncSocketPath().".lock";
339: open(LOCK,">$lock_file");
340: print LOCK "Contents not important";
341: close(LOCK);
342:
343: exit(0);
344: }
1.4 foxr 345: }
346: } else {
347: $IdleSeconds = 0; # Reset idle count if not idle.
348: }
1.15 foxr 349: #
350: # For each inflight transaction, tick down its timeout counter.
351: #
1.35 foxr 352:
1.34 albertel 353: foreach my $item (keys %ActiveConnections) {
354: my $State = $ActiveConnections{$item}->data->GetState();
1.35 foxr 355: if ($State ne 'Idle') {
1.34 albertel 356: Debug(5,"Ticking Socket $State $item");
357: $ActiveConnections{$item}->data->Tick();
358: }
1.15 foxr 359: }
1.5 foxr 360: # Do we have work in the queue, but no connections to service them?
361: # If so, try to make some new connections to get things going again.
362: #
1.57 foxr 363: # Note this code is dead now...
364: #
1.5 foxr 365: my $Requests = $WorkQueue->Count();
1.56 foxr 366: if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) {
1.10 foxr 367: if ($ConnectionRetriesLeft > 0) {
1.56 foxr 368: Debug(5,"Work but no connections, Make a new one");
369: my $success;
370: $success = &MakeLondConnection;
371: if($success == 0) { # All connections failed:
1.29 foxr 372: Debug(5,"Work in queue failed to make any connectiouns\n");
1.22 foxr 373: EmptyQueue(); # Fail pending transactions with con_lost.
1.42 foxr 374: CloseAllLondConnections(); # Should all be closed but....
1.10 foxr 375: }
376: } else {
1.56 foxr 377: $LondConnecting = 0;
1.22 foxr 378: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
1.29 foxr 379: Debug(5,"Work in queue, but gave up on connections..flushing\n");
1.10 foxr 380: EmptyQueue(); # Connections can't be established.
1.42 foxr 381: CloseAllLondConnections(); # Should all already be closed but...
1.5 foxr 382: }
383:
384: }
1.49 foxr 385: if ($ConnectionCount == 0) {
386: $KeyMode = "";
1.52 foxr 387: $clock_watcher->cancel();
1.49 foxr 388: }
1.1 foxr 389: }
390:
391: =pod
1.3 albertel 392:
1.1 foxr 393: =head2 SetupTimer
394:
1.3 albertel 395: Sets up a 1 per sec recurring timer event. The event handler is used to:
1.1 foxr 396:
1.3 albertel 397: =item
398:
399: Trigger timeouts on communications along active sockets.
400:
401: =item
402:
403: Trigger disconnections of idle sockets.
1.1 foxr 404:
405: =cut
406:
407: sub SetupTimer {
1.52 foxr 408: Debug(6, "SetupTimer");
409: Event->timer(interval => 1, cb => \&Tick );
1.1 foxr 410: }
1.3 albertel 411:
1.1 foxr 412: =pod
1.3 albertel 413:
1.1 foxr 414: =head2 ServerToIdle
1.3 albertel 415:
416: This function is called when a connection to the server is
417: ready for more work.
418:
419: If there is work in the Work queue the top element is dequeued
1.1 foxr 420: and the connection will start to work on it. If the work queue is
421: empty, the connection is pushed on the idle connection stack where
422: it will either get another work unit, or alternatively, if it sits there
423: long enough, it will be shut down and released.
424:
1.3 albertel 425: =cut
1.1 foxr 426:
427: sub ServerToIdle {
428: my $Socket = shift; # Get the socket.
1.49 foxr 429: $KeyMode = $Socket->{AuthenticationMode};
1.7 foxr 430: delete($ActiveTransactions{$Socket}); # Server has no transaction
1.1 foxr 431:
1.29 foxr 432: &Debug(5, "Server to idle");
1.1 foxr 433:
434: # If there's work to do, start the transaction:
435:
1.23 foxr 436: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
1.29 foxr 437: if ($reqdata ne undef) {
438: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
1.7 foxr 439: &StartRequest($Socket, $reqdata);
1.8 foxr 440:
1.1 foxr 441: } else {
442:
443: # There's no work waiting, so push the server to idle list.
1.29 foxr 444: &Debug(5, "No new work requests, server connection going idle");
1.1 foxr 445: $IdleConnections->push($Socket);
446: }
447: }
1.3 albertel 448:
1.1 foxr 449: =pod
1.3 albertel 450:
1.1 foxr 451: =head2 ClientWritable
1.3 albertel 452:
453: Event callback for when a client socket is writable.
454:
455: This callback is established when a transaction reponse is
456: avaiable from lond. The response is forwarded to the unix socket
457: as it becomes writable in this sub.
458:
1.1 foxr 459: Parameters:
460:
1.3 albertel 461: =item Event
462:
463: The event that has been triggered. Event->w->data is
464: the data and Event->w->fd is the socket to write.
1.1 foxr 465:
466: =cut
1.3 albertel 467:
1.1 foxr 468: sub ClientWritable {
469: my $Event = shift;
470: my $Watcher = $Event->w;
471: my $Data = $Watcher->data;
472: my $Socket = $Watcher->fd;
473:
474: # Try to send the data:
475:
476: &Debug(6, "ClientWritable writing".$Data);
477: &Debug(9, "Socket is: ".$Socket);
478:
1.6 foxr 479: if($Socket->connected) {
480: my $result = $Socket->send($Data, 0);
481:
482: # $result undefined: the write failed.
483: # otherwise $result is the number of bytes written.
484: # Remove that preceding string from the data.
485: # If the resulting data is empty, destroy the watcher
486: # and set up a read event handler to accept the next
487: # request.
488:
489: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
1.29 foxr 490: if($result ne undef) {
1.6 foxr 491: &Debug(9, "send result was defined");
492: if($result == length($Data)) { # Entire string sent.
493: &Debug(9, "ClientWritable data all written");
494: $Watcher->cancel();
495: #
496: # Set up to read next request from socket:
497:
498: my $descr = sprintf("Connection to lonc client %d",
499: $ActiveClients{$Socket});
500: Event->io(cb => \&ClientRequest,
501: poll => 'r',
502: desc => $descr,
503: data => "",
504: fd => $Socket);
505:
506: } else { # Partial string sent.
507: $Watcher->data(substr($Data, $result));
1.15 foxr 508: if($result == 0) { # client hung up on us!!
1.52 foxr 509: # Log("INFO", "lonc pipe client hung up on us!");
1.15 foxr 510: $Watcher->cancel;
511: $Socket->shutdown(2);
512: $Socket->close();
513: }
1.6 foxr 514: }
515:
516: } else { # Error of some sort...
517:
518: # Some errnos are possible:
519: my $errno = $!;
520: if($errno == POSIX::EWOULDBLOCK ||
521: $errno == POSIX::EAGAIN ||
522: $errno == POSIX::EINTR) {
523: # No action taken?
524: } else { # Unanticipated errno.
525: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
526: $Watcher->cancel; # Stop the watcher.
527: $Socket->shutdown(2); # Kill connection
528: $Socket->close(); # Close the socket.
529: }
1.1 foxr 530:
531: }
1.6 foxr 532: } else {
533: $Watcher->cancel(); # A delayed request...just cancel.
1.1 foxr 534: }
535: }
536:
537: =pod
1.3 albertel 538:
1.1 foxr 539: =head2 CompleteTransaction
1.3 albertel 540:
541: Called when the reply data has been received for a lond
1.1 foxr 542: transaction. The reply data must now be sent to the
543: ultimate client on the other end of the Unix socket. This is
544: done by setting up a writable event for the socket with the
545: data the reply data.
1.3 albertel 546:
1.1 foxr 547: Parameters:
1.3 albertel 548:
549: =item Socket
550:
551: Socket on which the lond transaction occured. This is a
552: LondConnection. The data received is in the TransactionReply member.
553:
1.7 foxr 554: =item Transaction
1.3 albertel 555:
1.7 foxr 556: The transaction that is being completed.
1.1 foxr 557:
558: =cut
1.3 albertel 559:
1.1 foxr 560: sub CompleteTransaction {
1.29 foxr 561: &Debug(5,"Complete transaction");
1.47 foxr 562:
563: my ($Socket, $Transaction) = @_;
1.1 foxr 564:
1.7 foxr 565: if (!$Transaction->isDeferred()) { # Normal transaction
566: my $data = $Socket->GetReply(); # Data to send.
1.39 foxr 567: if($LogTransactions) {
568: Log("SUCCESS", "Reply from lond: '$data'");
569: }
1.7 foxr 570: StartClientReply($Transaction, $data);
571: } else { # Delete deferred transaction file.
1.9 foxr 572: Log("SUCCESS", "A delayed transaction was completed");
1.23 foxr 573: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
1.7 foxr 574: unlink $Transaction->getFile();
575: }
1.6 foxr 576: }
1.42 foxr 577:
1.6 foxr 578: =pod
1.42 foxr 579:
1.6 foxr 580: =head1 StartClientReply
581:
582: Initiates a reply to a client where the reply data is a parameter.
583:
1.7 foxr 584: =head2 parameters:
585:
586: =item Transaction
587:
588: The transaction for which we are responding to the client.
589:
590: =item data
591:
592: The data to send to apached client.
593:
1.6 foxr 594: =cut
1.42 foxr 595:
1.6 foxr 596: sub StartClientReply {
1.1 foxr 597:
1.47 foxr 598: my ($Transaction, $data) = @_;
1.12 foxr 599:
1.7 foxr 600: my $Client = $Transaction->getClient();
601:
1.1 foxr 602: &Debug(8," Reply was: ".$data);
603: my $Serial = $ActiveClients{$Client};
604: my $desc = sprintf("Connection to lonc client %d",
605: $Serial);
606: Event->io(fd => $Client,
607: poll => "w",
608: desc => $desc,
609: cb => \&ClientWritable,
610: data => $data);
611: }
1.42 foxr 612:
1.4 foxr 613: =pod
1.42 foxr 614:
1.4 foxr 615: =head2 FailTransaction
616:
617: Finishes a transaction with failure because the associated lond socket
1.7 foxr 618: disconnected. There are two possibilities:
619: - The transaction is deferred: in which case we just quietly
620: delete the transaction since there is no client connection.
621: - The transaction is 'live' in which case we initiate the sending
622: of "con_lost" to the client.
623:
1.42 foxr 624: Deleting the transaction means killing it from the %ActiveTransactions hash.
1.4 foxr 625:
626: Parameters:
627:
628: =item client
629:
1.7 foxr 630: The LondTransaction we are failing.
631:
1.42 foxr 632:
1.4 foxr 633: =cut
634:
635: sub FailTransaction {
1.7 foxr 636: my $transaction = shift;
1.52 foxr 637:
638: # If the socket is dead, that's already logged.
639:
640: if ($ConnectionRetriesLeft > 0) {
641: Log("WARNING", "Failing transaction "
642: .$transaction->getRequest());
643: }
1.30 foxr 644: Debug(1, "Failing transaction: ".$transaction->getRequest());
1.10 foxr 645: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
1.11 foxr 646: my $client = $transaction->getClient();
1.30 foxr 647: Debug(1," Replying con_lost to ".$transaction->getRequest());
1.11 foxr 648: StartClientReply($transaction, "con_lost\n");
1.7 foxr 649: }
1.4 foxr 650:
651: }
652:
653: =pod
1.6 foxr 654: =head1 EmptyQueue
1.7 foxr 655:
1.6 foxr 656: Fails all items in the work queue with con_lost.
1.7 foxr 657: Note that each item in the work queue is a transaction.
658:
1.6 foxr 659: =cut
660: sub EmptyQueue {
1.22 foxr 661: $ConnectionRetriesLeft--; # Counts as connection failure too.
1.6 foxr 662: while($WorkQueue->Count()) {
1.10 foxr 663: my $request = $WorkQueue->dequeue(); # This is a transaction
1.7 foxr 664: FailTransaction($request);
1.6 foxr 665: }
666: }
667:
668: =pod
1.4 foxr 669:
1.9 foxr 670: =head2 CloseAllLondConnections
671:
672: Close all connections open on lond prior to exit e.g.
673:
674: =cut
675: sub CloseAllLondConnections {
1.23 foxr 676: foreach my $Socket (keys %ActiveConnections) {
1.42 foxr 677: if(exists($ActiveTransactions{$Socket})) {
678: FailTransaction($ActiveTransactions{$Socket});
679: }
680: KillSocket($Socket);
1.9 foxr 681: }
682: }
683: =cut
684:
685: =pod
686:
1.4 foxr 687: =head2 KillSocket
688:
689: Destroys a socket. This function can be called either when a socket
690: has died of 'natural' causes or because a socket needs to be pruned due to
691: idleness. If the socket has died naturally, if there are no longer any
692: live connections a new connection is created (in case there are transactions
693: in the queue). If the socket has been pruned, it is never re-created.
694:
695: Parameters:
1.1 foxr 696:
1.4 foxr 697: =item Socket
698:
699: The socket to kill off.
700:
701: =item Restart
702:
703: nonzero if we are allowed to create a new connection.
704:
705:
706: =cut
707: sub KillSocket {
708: my $Socket = shift;
709:
1.17 foxr 710: Log("WARNING", "Shutting down a socket");
1.9 foxr 711: $Socket->Shutdown();
712:
1.7 foxr 713: # If the socket came from the active connection set,
714: # delete its transaction... note that FailTransaction should
715: # already have been called!!!
716: # otherwise it came from the idle set.
717: #
1.4 foxr 718:
719: if(exists($ActiveTransactions{$Socket})) {
720: delete ($ActiveTransactions{$Socket});
721: }
722: if(exists($ActiveConnections{$Socket})) {
723: delete($ActiveConnections{$Socket});
1.37 albertel 724: $ConnectionCount--;
725: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
1.4 foxr 726: }
1.6 foxr 727: # If the connection count has gone to zero and there is work in the
728: # work queue, the work all gets failed with con_lost.
729: #
730: if($ConnectionCount == 0) {
1.22 foxr 731: EmptyQueue();
1.42 foxr 732: CloseAllLondConnections; # Should all already be closed but...
1.4 foxr 733: }
734: }
1.1 foxr 735:
736: =pod
1.3 albertel 737:
1.1 foxr 738: =head2 LondReadable
1.3 albertel 739:
1.1 foxr 740: This function is called whenever a lond connection
741: is readable. The action is state dependent:
742:
1.3 albertel 743: =head3 State=Initialized
744:
745: We''re waiting for the challenge, this is a no-op until the
1.1 foxr 746: state changes.
1.3 albertel 747:
1.1 foxr 748: =head3 State=Challenged
1.3 albertel 749:
750: The challenge has arrived we need to transition to Writable.
1.1 foxr 751: The connection must echo the challenge back.
1.3 albertel 752:
1.1 foxr 753: =head3 State=ChallengeReplied
1.3 albertel 754:
755: The challenge has been replied to. The we are receiveing the
1.1 foxr 756: 'ok' from the partner.
1.3 albertel 757:
1.40 foxr 758: =head3 State=ReadingVersionString
759:
760: We have requested the lond version and are reading the
761: version back. Upon completion, we'll store the version away
762: for future use(?).
763:
764: =head3 State=HostSet
765:
766: We have selected the domain name of our peer (multhomed hosts)
767: and are getting the reply (presumably ok) back.
768:
1.1 foxr 769: =head3 State=RequestingKey
1.3 albertel 770:
771: The ok has been received and we need to send the request for
1.1 foxr 772: an encryption key. Transition to writable for that.
1.3 albertel 773:
1.1 foxr 774: =head3 State=ReceivingKey
1.3 albertel 775:
776: The the key has been requested, now we are reading the new key.
777:
1.1 foxr 778: =head3 State=Idle
1.3 albertel 779:
780: The encryption key has been negotiated or we have finished
1.1 foxr 781: reading data from the a transaction. If the callback data has
782: a client as well as the socket iformation, then we are
783: doing a transaction and the data received is relayed to the client
784: before the socket is put on the idle list.
1.3 albertel 785:
1.1 foxr 786: =head3 State=SendingRequest
1.3 albertel 787:
788: I do not think this state can be received here, but if it is,
1.1 foxr 789: the appropriate thing to do is to transition to writable, and send
790: the request.
1.3 albertel 791:
1.1 foxr 792: =head3 State=ReceivingReply
1.3 albertel 793:
794: We finished sending the request to the server and now transition
1.1 foxr 795: to readable to receive the reply.
796:
797: The parameter to this function are:
1.3 albertel 798:
1.1 foxr 799: The event. Implicit in this is the watcher and its data. The data
800: contains at least the lond connection object and, if a
801: transaction is in progress, the socket attached to the local client.
802:
1.3 albertel 803: =cut
1.1 foxr 804:
805: sub LondReadable {
1.8 foxr 806:
1.41 albertel 807: my $Event = shift;
808: my $Watcher = $Event->w;
809: my $Socket = $Watcher->data;
810: my $client = undef;
1.40 foxr 811:
1.41 albertel 812: &Debug(6,"LondReadable called state = ".$Socket->GetState());
1.40 foxr 813:
814:
1.41 albertel 815: my $State = $Socket->GetState(); # All action depends on the state.
1.40 foxr 816:
1.41 albertel 817: SocketDump(6, $Socket);
818: my $status = $Socket->Readable();
1.40 foxr 819:
1.41 albertel 820: &Debug(2, "Socket->Readable returned: $status");
1.40 foxr 821:
1.41 albertel 822: if($status != 0) {
823: # bad return from socket read. Currently this means that
824: # The socket has become disconnected. We fail the transaction.
1.40 foxr 825:
1.41 albertel 826: Log("WARNING",
827: "Lond connection lost.");
828: if(exists($ActiveTransactions{$Socket})) {
829: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 830: } else {
831: # Socket is connecting and failed... need to mark
832: # no longer connecting.
833:
834: $LondConnecting = 0;
1.41 albertel 835: }
836: $Watcher->cancel();
837: KillSocket($Socket);
838: $ConnectionRetriesLeft--; # Counts as connection failure
839: return;
840: }
841: SocketDump(6,$Socket);
1.17 foxr 842:
1.41 albertel 843: $State = $Socket->GetState(); # Update in case of transition.
844: &Debug(6, "After read, state is ".$State);
1.1 foxr 845:
1.41 albertel 846: if($State eq "Initialized") {
1.1 foxr 847:
848:
1.41 albertel 849: } elsif ($State eq "ChallengeReceived") {
1.1 foxr 850: # The challenge must be echoed back; The state machine
851: # in the connection takes care of setting that up. Just
852: # need to transition to writable:
1.41 albertel 853:
854: $Watcher->cb(\&LondWritable);
855: $Watcher->poll("w");
1.1 foxr 856:
1.41 albertel 857: } elsif ($State eq "ChallengeReplied") {
1.1 foxr 858:
1.41 albertel 859: } elsif ($State eq "RequestingVersion") {
860: # Need to ask for the version... that is writiability:
1.1 foxr 861:
1.41 albertel 862: $Watcher->cb(\&LondWritable);
863: $Watcher->poll("w");
864:
865: } elsif ($State eq "ReadingVersionString") {
866: # Read the rest of the version string...
867: } elsif ($State eq "SetHost") {
868: # Need to request the actual domain get set...
869:
870: $Watcher->cb(\&LondWritable);
871: $Watcher->poll("w");
872: } elsif ($State eq "HostSet") {
873: # Reading the 'ok' from the peer.
874:
875: } elsif ($State eq "RequestingKey") {
1.1 foxr 876: # The ok was received. Now we need to request the key
877: # That requires us to be writable:
878:
1.41 albertel 879: $Watcher->cb(\&LondWritable);
880: $Watcher->poll("w");
1.1 foxr 881:
1.41 albertel 882: } elsif ($State eq "ReceivingKey") {
1.1 foxr 883:
1.41 albertel 884: } elsif ($State eq "Idle") {
1.40 foxr 885:
1.41 albertel 886: # This is as good a spot as any to get the peer version
887: # string:
1.40 foxr 888:
1.41 albertel 889: if($LondVersion eq "unknown") {
890: $LondVersion = $Socket->PeerVersion();
891: Log("INFO", "Connected to lond version: $LondVersion");
892: }
1.1 foxr 893: # If necessary, complete a transaction and then go into the
894: # idle queue.
1.22 foxr 895: # Note that a trasition to idle indicates a live lond
896: # on the other end so reset the connection retries.
897: #
1.41 albertel 898: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
899: $Watcher->cancel();
900: if(exists($ActiveTransactions{$Socket})) {
901: Debug(5,"Completing transaction!!");
902: CompleteTransaction($Socket,
903: $ActiveTransactions{$Socket});
904: } else {
905: Log("SUCCESS", "Connection ".$ConnectionCount." to "
906: .$RemoteHost." now ready for action");
907: }
908: ServerToIdle($Socket); # Next work unit or idle.
1.54 foxr 909:
910: #
911: $LondConnecting = 0; # Best spot I can think of for this.
912: #
1.6 foxr 913:
1.41 albertel 914: } elsif ($State eq "SendingRequest") {
1.1 foxr 915: # We need to be writable for this and probably don't belong
916: # here inthe first place.
917:
1.41 albertel 918: Deubg(6, "SendingRequest state encountered in readable");
919: $Watcher->poll("w");
920: $Watcher->cb(\&LondWritable);
1.1 foxr 921:
1.41 albertel 922: } elsif ($State eq "ReceivingReply") {
1.1 foxr 923:
924:
1.41 albertel 925: } else {
926: # Invalid state.
927: Debug(4, "Invalid state in LondReadable");
928: }
1.1 foxr 929: }
1.3 albertel 930:
1.1 foxr 931: =pod
1.3 albertel 932:
1.1 foxr 933: =head2 LondWritable
1.3 albertel 934:
1.1 foxr 935: This function is called whenever a lond connection
936: becomes writable while there is a writeable monitoring
937: event. The action taken is very state dependent:
1.3 albertel 938:
1.1 foxr 939: =head3 State = Connected
1.3 albertel 940:
941: The connection is in the process of sending the 'init' hailing to the
942: lond on the remote end. The connection object''s Writable member is
943: called. On error, ConnectionError is called to destroy the connection
944: and remove it from the ActiveConnections hash
945:
1.1 foxr 946: =head3 Initialized
1.3 albertel 947:
948: 'init' has been sent, writability monitoring is removed and
949: readability monitoring is started with LondReadable as the callback.
950:
1.1 foxr 951: =head3 ChallengeReceived
1.3 albertel 952:
953: The connection has received the who are you challenge from the remote
954: system, and is in the process of sending the challenge
955: response. Writable is called.
956:
1.1 foxr 957: =head3 ChallengeReplied
1.3 albertel 958:
959: The connection has replied to the initial challenge The we switch to
960: monitoring readability looking for the server to reply with 'ok'.
961:
1.1 foxr 962: =head3 RequestingKey
1.3 albertel 963:
964: The connection is in the process of requesting its encryption key.
965: Writable is called.
966:
1.1 foxr 967: =head3 ReceivingKey
1.3 albertel 968:
969: The connection has sent the request for a key. Switch to readability
970: monitoring to accept the key
971:
1.1 foxr 972: =head3 SendingRequest
1.3 albertel 973:
974: The connection is in the process of sending a request to the server.
975: This request is part of a client transaction. All the states until
976: now represent the client setup protocol. Writable is called.
977:
1.1 foxr 978: =head3 ReceivingReply
979:
1.3 albertel 980: The connection has sent a request. Now it must receive a reply.
981: Readability monitoring is requested.
982:
983: This function is an event handler and therefore receives as
1.1 foxr 984: a parameter the event that has fired. The data for the watcher
985: of this event is a reference to a list of one or two elements,
986: depending on state. The first (and possibly only) element is the
987: socket. The second (present only if a request is in progress)
988: is the socket on which to return a reply to the caller.
989:
990: =cut
1.3 albertel 991:
1.1 foxr 992: sub LondWritable {
993: my $Event = shift;
994: my $Watcher = $Event->w;
1.8 foxr 995: my $Socket = $Watcher->data;
996: my $State = $Socket->GetState();
1.1 foxr 997:
1.8 foxr 998: Debug(6,"LondWritable State = ".$State."\n");
1.1 foxr 999:
1.8 foxr 1000:
1.1 foxr 1001: # Figure out what to do depending on the state of the socket:
1002:
1003:
1004:
1005:
1006: SocketDump(6,$Socket);
1007:
1.42 foxr 1008: # If the socket is writable, we must always write.
1009: # Only by writing will we undergo state transitions.
1010: # Old logic wrote in state specific code below, however
1011: # That forces us at least through another invocation of
1012: # this function after writability is possible again.
1013: # This logic also factors out common code for handling
1014: # write failures... in all cases, write failures
1015: # Kill the socket.
1016: # This logic makes the branches of the >big< if below
1017: # so that the writing states are actually NO-OPs.
1018:
1019: if ($Socket->Writable() != 0) {
1.43 albertel 1020: # The write resulted in an error.
1021: # We'll treat this as if the socket got disconnected:
1022: Log("WARNING", "Connection to ".$RemoteHost.
1023: " has been disconnected");
1024: if(exists($ActiveTransactions{$Socket})) {
1025: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 1026: } else {
1027: # In the process of conneting, so need to turn that off.
1028:
1029: $LondConnecting = 0;
1.43 albertel 1030: }
1031: $Watcher->cancel();
1032: KillSocket($Socket);
1033: return;
1.42 foxr 1034: }
1035:
1036:
1037:
1.41 albertel 1038: if ($State eq "Connected") {
1.1 foxr 1039:
1.41 albertel 1040: # "init" is being sent...
1.42 foxr 1041:
1.41 albertel 1042: } elsif ($State eq "Initialized") {
1.4 foxr 1043:
1.41 albertel 1044: # Now that init was sent, we switch
1045: # to watching for readability:
1.1 foxr 1046:
1.41 albertel 1047: $Watcher->cb(\&LondReadable);
1048: $Watcher->poll("r");
1049:
1050: } elsif ($State eq "ChallengeReceived") {
1051: # We received the challenge, now we
1052: # are echoing it back. This is a no-op,
1053: # we're waiting for the state to change
1.1 foxr 1054:
1.41 albertel 1055: } elsif ($State eq "ChallengeReplied") {
1056: # The echo was sent back, so we switch
1057: # to watching readability.
1058:
1059: $Watcher->cb(\&LondReadable);
1060: $Watcher->poll("r");
1061: } elsif ($State eq "RequestingVersion") {
1062: # Sending the peer a version request...
1.42 foxr 1063:
1.41 albertel 1064: } elsif ($State eq "ReadingVersionString") {
1065: # Transition to read since we have sent the
1066: # version command and now just need to read the
1067: # version string from the peer:
1.40 foxr 1068:
1.41 albertel 1069: $Watcher->cb(\&LondReadable);
1070: $Watcher->poll("r");
1.40 foxr 1071:
1.41 albertel 1072: } elsif ($State eq "SetHost") {
1073: # Setting the remote domain...
1.42 foxr 1074:
1.41 albertel 1075: } elsif ($State eq "HostSet") {
1076: # Back to readable to get the ok.
1.40 foxr 1077:
1.41 albertel 1078: $Watcher->cb(\&LondReadable);
1079: $Watcher->poll("r");
1.40 foxr 1080:
1081:
1.41 albertel 1082: } elsif ($State eq "RequestingKey") {
1083: # At this time we're requesting the key.
1084: # again, this is essentially a no-op.
1085:
1086: } elsif ($State eq "ReceivingKey") {
1087: # Now we need to wait for the key
1088: # to come back from the peer:
1089:
1090: $Watcher->cb(\&LondReadable);
1091: $Watcher->poll("r");
1092:
1093: } elsif ($State eq "SendingRequest") {
1.40 foxr 1094:
1.41 albertel 1095: # At this time we are sending a request to the
1.1 foxr 1096: # peer... write the next chunk:
1097:
1.41 albertel 1098:
1099: } elsif ($State eq "ReceivingReply") {
1100: # The send has completed. Wait for the
1101: # data to come in for a reply.
1102: Debug(8,"Writable sent request/receiving reply");
1103: $Watcher->cb(\&LondReadable);
1104: $Watcher->poll("r");
1.1 foxr 1105:
1.41 albertel 1106: } else {
1107: # Control only passes here on an error:
1108: # the socket state does not match any
1109: # of the known states... so an error
1110: # must be logged.
1.1 foxr 1111:
1.41 albertel 1112: &Debug(4, "Invalid socket state ".$State."\n");
1113: }
1.1 foxr 1114:
1115: }
1.6 foxr 1116: =pod
1117:
1118: =cut
1119: sub QueueDelayed {
1.8 foxr 1120: Debug(3,"QueueDelayed called");
1121:
1.6 foxr 1122: my $path = "$perlvar{'lonSockDir'}/delayed";
1.8 foxr 1123:
1124: Debug(4, "Delayed path: ".$path);
1.6 foxr 1125: opendir(DIRHANDLE, $path);
1.8 foxr 1126:
1.23 foxr 1127: my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
1.6 foxr 1128: closedir(DIRHANDLE);
1129: my $dfname;
1.8 foxr 1130: my $reqfile;
1131: foreach $dfname (sort @alldelayed) {
1132: $reqfile = "$path/$dfname";
1133: Debug(4, "queueing ".$reqfile);
1.6 foxr 1134: my $Handle = IO::File->new($reqfile);
1135: my $cmd = <$Handle>;
1.8 foxr 1136: chomp $cmd; # There may or may not be a newline...
1.12 foxr 1137: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1.7 foxr 1138: my $Transaction = LondTransaction->new($cmd);
1139: $Transaction->SetDeferred($reqfile);
1140: QueueTransaction($Transaction);
1.6 foxr 1141: }
1142:
1143: }
1.1 foxr 1144:
1145: =pod
1.3 albertel 1146:
1.1 foxr 1147: =head2 MakeLondConnection
1.3 albertel 1148:
1149: Create a new lond connection object, and start it towards its initial
1150: idleness. Once idle, it becomes elligible to receive transactions
1151: from the work queue. If the work queue is not empty when the
1152: connection is completed and becomes idle, it will dequeue an entry and
1153: start off on it.
1154:
1.1 foxr 1155: =cut
1.3 albertel 1156:
1.1 foxr 1157: sub MakeLondConnection {
1158: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1159: .GetServerPort());
1160:
1161: my $Connection = LondConnection->new(&GetServerHost(),
1162: &GetServerPort());
1163:
1.30 foxr 1164: if($Connection eq undef) { # Needs to be more robust later.
1.9 foxr 1165: Log("CRITICAL","Failed to make a connection with lond.");
1.10 foxr 1166: $ConnectionRetriesLeft--;
1167: return 0; # Failure.
1.5 foxr 1168: } else {
1.22 foxr 1169:
1.5 foxr 1170: # The connection needs to have writability
1171: # monitored in order to send the init sequence
1172: # that starts the whole authentication/key
1173: # exchange underway.
1174: #
1175: my $Socket = $Connection->GetSocket();
1.30 foxr 1176: if($Socket eq undef) {
1.5 foxr 1177: die "did not get a socket from the connection";
1178: } else {
1179: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1180: }
1.1 foxr 1181:
1.21 foxr 1182: $Connection->SetTimeoutCallback(\&SocketTimeout);
1183:
1.23 foxr 1184: my $event = Event->io(fd => $Socket,
1.5 foxr 1185: poll => 'w',
1186: cb => \&LondWritable,
1.8 foxr 1187: data => $Connection,
1.5 foxr 1188: desc => 'Connection to lond server');
1189: $ActiveConnections{$Connection} = $event;
1.52 foxr 1190: if ($ConnectionCount == 0) {
1191: &SetupTimer; # Need to handle timeouts with connections...
1192: }
1.5 foxr 1193: $ConnectionCount++;
1.8 foxr 1194: Debug(4, "Connection count = ".$ConnectionCount);
1.6 foxr 1195: if($ConnectionCount == 1) { # First Connection:
1196: QueueDelayed;
1197: }
1.9 foxr 1198: Log("SUCESS", "Created connection ".$ConnectionCount
1199: ." to host ".GetServerHost());
1.54 foxr 1200: $LondConnecting = 1; # Connection in progress.
1.10 foxr 1201: return 1; # Return success.
1.1 foxr 1202: }
1203:
1204: }
1.3 albertel 1205:
1.1 foxr 1206: =pod
1.3 albertel 1207:
1.1 foxr 1208: =head2 StartRequest
1.3 albertel 1209:
1210: Starts a lond request going on a specified lond connection.
1211: parameters are:
1212:
1213: =item $Lond
1214:
1215: Connection to the lond that will send the transaction and receive the
1216: reply.
1217:
1218: =item $Client
1219:
1220: Connection to the client that is making this request We got the
1221: request from this socket, and when the request has been relayed to
1222: lond and we get a reply back from lond it will get sent to this
1223: socket.
1224:
1225: =item $Request
1226:
1227: The text of the request to send.
1228:
1.1 foxr 1229: =cut
1230:
1231: sub StartRequest {
1.47 foxr 1232:
1233: my ($Lond, $Request) = @_;
1.1 foxr 1234:
1.7 foxr 1235: Debug(6, "StartRequest: ".$Request->getRequest());
1.1 foxr 1236:
1237: my $Socket = $Lond->GetSocket();
1238:
1.7 foxr 1239: $Request->Activate($Lond);
1240: $ActiveTransactions{$Lond} = $Request;
1.1 foxr 1241:
1.7 foxr 1242: $Lond->InitiateTransaction($Request->getRequest());
1.23 foxr 1243: my $event = Event->io(fd => $Socket,
1.1 foxr 1244: poll => "w",
1245: cb => \&LondWritable,
1246: data => $Lond,
1247: desc => "lond transaction connection");
1248: $ActiveConnections{$Lond} = $event;
1249: Debug(8," Start Request made watcher data with ".$event->data."\n");
1250: }
1251:
1252: =pod
1.3 albertel 1253:
1.1 foxr 1254: =head2 QueueTransaction
1.3 albertel 1255:
1256: If there is an idle lond connection, it is put to work doing this
1257: transaction. Otherwise, the transaction is placed in the work queue.
1258: If placed in the work queue and the maximum number of connections has
1259: not yet been created, a new connection will be started. Our goal is
1260: to eventually have a sufficient number of connections that the work
1261: queue will typically be empty. parameters are:
1262:
1263: =item Socket
1264:
1265: open on the lonc client.
1266:
1267: =item Request
1268:
1269: data to send to the lond.
1.1 foxr 1270:
1271: =cut
1.3 albertel 1272:
1.1 foxr 1273: sub QueueTransaction {
1274:
1.7 foxr 1275: my $requestData = shift; # This is a LondTransaction.
1276: my $cmd = $requestData->getRequest();
1277:
1278: Debug(6,"QueueTransaction: ".$cmd);
1.1 foxr 1279:
1280: my $LondSocket = $IdleConnections->pop();
1281: if(!defined $LondSocket) { # Need to queue request.
1.29 foxr 1282: Debug(5,"Must queue...");
1.1 foxr 1283: $WorkQueue->enqueue($requestData);
1.56 foxr 1284: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1285: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1286:
1.22 foxr 1287: if($ConnectionRetriesLeft > 0) {
1.29 foxr 1288: Debug(5,"Starting additional lond connection");
1.56 foxr 1289: if(&MakeLondConnection() == 0) {
1.22 foxr 1290: EmptyQueue(); # Fail transactions, can't make connection.
1.42 foxr 1291: CloseAllLondConnections; # Should all be closed but...
1.22 foxr 1292: }
1293: } else {
1294: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1.56 foxr 1295: $LondConnecting = 0;
1.22 foxr 1296: EmptyQueue(); # It's worse than that ... he's dead Jim.
1.42 foxr 1297: CloseAllLondConnections; # Should all be closed but..
1.17 foxr 1298: }
1.1 foxr 1299: }
1300: } else { # Can start the request:
1301: Debug(8,"Can start...");
1.7 foxr 1302: StartRequest($LondSocket, $requestData);
1.1 foxr 1303: }
1304: }
1305:
1306: #-------------------------- Lonc UNIX socket handling ---------------------
1.3 albertel 1307:
1.1 foxr 1308: =pod
1.3 albertel 1309:
1.1 foxr 1310: =head2 ClientRequest
1.3 albertel 1311: Callback that is called when data can be read from the UNIX domain
1312: socket connecting us with an apache server process.
1.1 foxr 1313:
1314: =cut
1315:
1316: sub ClientRequest {
1317: Debug(6, "ClientRequest");
1318: my $event = shift;
1319: my $watcher = $event->w;
1320: my $socket = $watcher->fd;
1321: my $data = $watcher->data;
1322: my $thisread;
1323:
1324: Debug(9, " Watcher named: ".$watcher->desc);
1325:
1326: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1327: Debug(8, "rcv: data length = ".length($thisread)
1328: ." read =".$thisread);
1.29 foxr 1329: unless (defined $rv && length($thisread)) {
1.1 foxr 1330: # Likely eof on socket.
1331: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1332: close($socket);
1333: $watcher->cancel();
1334: delete($ActiveClients{$socket});
1.10 foxr 1335: return;
1.1 foxr 1336: }
1337: Debug(8,"Data: ".$data." this read: ".$thisread);
1338: $data = $data.$thisread; # Append new data.
1339: $watcher->data($data);
1.44 albertel 1340: if($data =~ /\n$/) { # Request entirely read.
1.10 foxr 1341: if($data eq "close_connection_exit\n") {
1.9 foxr 1342: Log("CRITICAL",
1343: "Request Close Connection ... exiting");
1344: CloseAllLondConnections();
1345: exit;
1346: }
1.1 foxr 1347: Debug(8, "Complete transaction received: ".$data);
1.39 foxr 1348: if($LogTransactions) {
1349: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1350: }
1.8 foxr 1351: my $Transaction = LondTransaction->new($data);
1.7 foxr 1352: $Transaction->SetClient($socket);
1353: QueueTransaction($Transaction);
1.1 foxr 1354: $watcher->cancel(); # Done looking for input data.
1355: }
1356:
1357: }
1358:
1359:
1360: =pod
1.3 albertel 1361:
1.1 foxr 1362: =head2 NewClient
1.3 albertel 1363:
1364: Callback that is called when a connection is received on the unix
1365: socket for a new client of lonc. The callback is parameterized by the
1366: event.. which is a-priori assumed to be an io event, and therefore has
1367: an fd member that is the Listener socket. We Accept the connection
1368: and register a new event on the readability of that socket:
1369:
1.1 foxr 1370: =cut
1.3 albertel 1371:
1.1 foxr 1372: sub NewClient {
1373: Debug(6, "NewClient");
1374: my $event = shift; # Get the event parameters.
1375: my $watcher = $event->w;
1376: my $socket = $watcher->fd; # Get the event' socket.
1377: my $connection = $socket->accept(); # Accept the client connection.
1378: Debug(8,"Connection request accepted from "
1379: .GetPeername($connection, AF_UNIX));
1380:
1381:
1382: my $description = sprintf("Connection to lonc client %d",
1383: $ClientConnection);
1384: Debug(9, "Creating event named: ".$description);
1385: Event->io(cb => \&ClientRequest,
1386: poll => 'r',
1387: desc => $description,
1388: data => "",
1389: fd => $connection);
1390: $ActiveClients{$connection} = $ClientConnection;
1391: $ClientConnection++;
1392: }
1.3 albertel 1393:
1394: =pod
1395:
1396: =head2 GetLoncSocketPath
1397:
1398: Returns the name of the UNIX socket on which to listen for client
1399: connections.
1.1 foxr 1400:
1.58 foxr 1401: =head2 Parameters:
1402:
1403: host (optional) - Name of the host socket to return.. defaults to
1404: the return from GetServerHost().
1405:
1.1 foxr 1406: =cut
1.3 albertel 1407:
1.1 foxr 1408: sub GetLoncSocketPath {
1.58 foxr 1409:
1410: my $host = GetServerHost(); # Default host.
1411: if (@_) {
1412: ($host) = @_; # Override if supplied.
1413: }
1414: return $UnixSocketDir."/".$host;
1.1 foxr 1415: }
1416:
1.3 albertel 1417: =pod
1418:
1419: =head2 GetServerHost
1420:
1421: Returns the host whose lond we talk with.
1422:
1.1 foxr 1423: =cut
1.3 albertel 1424:
1.7 foxr 1425: sub GetServerHost {
1.1 foxr 1426: return $RemoteHost; # Setup by the fork.
1427: }
1.3 albertel 1428:
1429: =pod
1430:
1431: =head2 GetServerPort
1432:
1433: Returns the lond port number.
1434:
1.1 foxr 1435: =cut
1.3 albertel 1436:
1.7 foxr 1437: sub GetServerPort {
1.1 foxr 1438: return $perlvar{londPort};
1439: }
1.3 albertel 1440:
1441: =pod
1442:
1443: =head2 SetupLoncListener
1444:
1445: Setup a lonc listener event. The event is called when the socket
1446: becomes readable.. that corresponds to the receipt of a new
1447: connection. The event handler established will accept the connection
1448: (creating a communcations channel), that int turn will establish
1449: another event handler to subess requests.
1.1 foxr 1450:
1.58 foxr 1451: =head2 Parameters:
1452:
1453: host (optional) Name of the host to set up a unix socket to.
1454:
1.1 foxr 1455: =cut
1.3 albertel 1456:
1.1 foxr 1457: sub SetupLoncListener {
1458:
1.58 foxr 1459: my $host = GetServerHost(); # Default host.
1460: if (@_) {
1461: ($host) = @_ # Override host with parameter.
1462: }
1463:
1.1 foxr 1464: my $socket;
1.58 foxr 1465: my $SocketName = GetLoncSocketPath($host);
1.1 foxr 1466: unlink($SocketName);
1.7 foxr 1467: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1.55 albertel 1468: Listen => 250,
1.1 foxr 1469: Type => SOCK_STREAM)) {
1470: die "Failed to create a lonc listner socket";
1471: }
1.59 foxr 1472: return $socket;
1.1 foxr 1473: }
1474:
1.39 foxr 1475: #
1476: # Toggle transaction logging.
1477: # Implicit inputs:
1478: # LogTransactions
1479: # Implicit Outputs:
1480: # LogTransactions
1481: sub ToggleTransactionLogging {
1482: print STDERR "Toggle transaction logging...\n";
1483: if(!$LogTransactions) {
1484: $LogTransactions = 1;
1485: } else {
1486: $LogTransactions = 0;
1487: }
1488:
1489:
1490: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1491: }
1492:
1.14 foxr 1493: =pod
1494:
1495: =head2 ChildStatus
1496:
1497: Child USR1 signal handler to report the most recent status
1498: into the status file.
1499:
1.22 foxr 1500: We also use this to reset the retries count in order to allow the
1501: client to retry connections with a previously dead server.
1.14 foxr 1502: =cut
1.46 albertel 1503:
1.14 foxr 1504: sub ChildStatus {
1505: my $event = shift;
1506: my $watcher = $event->w;
1507:
1508: Debug(2, "Reporting child status because : ".$watcher->data);
1509: my $docdir = $perlvar{'lonDocRoot'};
1510: my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
1511: print $fh $$."\t".$RemoteHost."\t".$Status."\t".
1512: $RecentLogEntry."\n";
1.38 foxr 1513: #
1514: # Write out information about each of the connections:
1515: #
1.46 albertel 1516: if ($DebugLevel > 2) {
1517: print $fh "Active connection statuses: \n";
1518: my $i = 1;
1519: print STDERR "================================= Socket Status Dump:\n";
1520: foreach my $item (keys %ActiveConnections) {
1521: my $Socket = $ActiveConnections{$item}->data;
1522: my $state = $Socket->GetState();
1523: print $fh "Connection $i State: $state\n";
1524: print STDERR "---------------------- Connection $i \n";
1.48 foxr 1525: $Socket->Dump(-1); # Ensure it gets dumped..
1.46 albertel 1526: $i++;
1527: }
1.38 foxr 1528: }
1.22 foxr 1529: $ConnectionRetriesLeft = $ConnectionRetries;
1.14 foxr 1530: }
1531:
1.1 foxr 1532: =pod
1.3 albertel 1533:
1.10 foxr 1534: =head2 SignalledToDeath
1535:
1536: Called in response to a signal that causes a chid process to die.
1537:
1538: =cut
1539:
1540:
1541: sub SignalledToDeath {
1.14 foxr 1542: my $event = shift;
1543: my $watcher= $event->w;
1544:
1545: Debug(2,"Signalled to death! via ".$watcher->data);
1.17 foxr 1546: my ($signal) = $watcher->data;
1.10 foxr 1547: chomp($signal);
1548: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1549: ."died through "."\"$signal\"");
1550: LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1551: ."\"$signal\"");
1.12 foxr 1552: exit 0;
1.10 foxr 1553:
1554: }
1.16 foxr 1555:
1556: =head2 ToggleDebug
1557:
1558: This sub toggles trace debugging on and off.
1559:
1560: =cut
1561:
1562: sub ToggleDebug {
1563: my $Current = $DebugLevel;
1564: $DebugLevel = $NextDebugLevel;
1565: $NextDebugLevel = $Current;
1566:
1567: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1568:
1569: }
1570:
1.1 foxr 1571: =head2 ChildProcess
1572:
1573: This sub implements a child process for a single lonc daemon.
1.61 ! foxr 1574: Optional parameter:
! 1575: $socket - if provided, this is a socket already open for listen
! 1576: on the client socket. Otherwise, a new listen is set up.
1.1 foxr 1577:
1578: =cut
1579:
1580: sub ChildProcess {
1581:
1582:
1.14 foxr 1583: #
1584: # Signals must be handled by the Event framework...
1.61 ! foxr 1585: #
1.14 foxr 1586:
1587: Event->signal(signal => "QUIT",
1588: cb => \&SignalledToDeath,
1589: data => "QUIT");
1590: Event->signal(signal => "HUP",
1591: cb => \&ChildStatus,
1592: data => "HUP");
1593: Event->signal(signal => "USR1",
1594: cb => \&ChildStatus,
1595: data => "USR1");
1.39 foxr 1596: Event->signal(signal => "USR2",
1597: cb => \&ToggleTransactionLogging);
1.16 foxr 1598: Event->signal(signal => "INT",
1599: cb => \&ToggleDebug,
1600: data => "INT");
1.1 foxr 1601:
1602:
1.61 ! foxr 1603: my ($socket) = @_;
! 1604: if (!$socket) {
! 1605:
! 1606: $socket = SetupLoncListener();
! 1607: }
1.59 foxr 1608: Event->io(cb => \&NewClient,
1609: poll => 'r',
1610: desc => 'Lonc Listener Unix Socket',
1611: fd => $socket);
1.1 foxr 1612:
1613: $Event::Debuglevel = $DebugLevel;
1614:
1615: Debug(9, "Making initial lond connection for ".$RemoteHost);
1616:
1617: # Setup the initial server connection:
1618:
1.61 ! foxr 1619: # &MakeLondConnection(); // let first work requirest do it.
1.10 foxr 1620:
1.5 foxr 1621:
1.1 foxr 1622: Debug(9,"Entering event loop");
1623: my $ret = Event::loop(); # Start the main event loop.
1624:
1625:
1626: die "Main event loop exited!!!";
1627: }
1628:
1629: # Create a new child for host passed in:
1630:
1631: sub CreateChild {
1.52 foxr 1632: my $host = shift;
1633:
1.12 foxr 1634: my $sigset = POSIX::SigSet->new(SIGINT);
1635: sigprocmask(SIG_BLOCK, $sigset);
1.1 foxr 1636: $RemoteHost = $host;
1.9 foxr 1637: Log("CRITICAL", "Forking server for ".$host);
1.23 foxr 1638: my $pid = fork;
1.1 foxr 1639: if($pid) { # Parent
1.17 foxr 1640: $RemoteHost = "Parent";
1.27 foxr 1641: $ChildHash{$pid} = $host;
1.26 foxr 1642: $HostToPid{$host}= $pid;
1.12 foxr 1643: sigprocmask(SIG_UNBLOCK, $sigset);
1644:
1.1 foxr 1645: } else { # child.
1.5 foxr 1646: ShowStatus("Connected to ".$RemoteHost);
1.23 foxr 1647: $SIG{INT} = 'DEFAULT';
1.12 foxr 1648: sigprocmask(SIG_UNBLOCK, $sigset);
1649: ChildProcess; # Does not return.
1.1 foxr 1650: }
1.61 ! foxr 1651: }
1.1 foxr 1652:
1.61 ! foxr 1653: # parent_client_connection:
! 1654: # Event handler that processes client connections for the parent process.
! 1655: # This sub is called when the parent is listening on a socket and
! 1656: # a connection request arrives. We must:
! 1657: # Start a child process to accept the connection request.
! 1658: # Kill our listen on the socket.
! 1659: # Setup an event to handle the child process exit. (SIGCHLD).
! 1660: # Parameter:
! 1661: # event - The event object that was created to monitor this socket.
! 1662: # event->w->fd is the socket.
! 1663: # Returns:
! 1664: # NONE
! 1665: #
! 1666: sub parent_client_connection {
! 1667: die "DieWhenIdle processing not completely operational yet";
! 1668:
! 1669: }
! 1670:
! 1671: # parent_listen:
! 1672: # Opens a socket and starts a listen for the parent process on a client UNIX
! 1673: # domain socket.
! 1674: #
! 1675: # This involves:
! 1676: # Creating a socket for listen.
! 1677: # Removing any socket lock file
! 1678: # Adding an event handler for this socket becoming readable
! 1679: # To the parent's event dispatcher.
! 1680: # Parameters:
! 1681: # loncapa_host - LonCAPA cluster name of the host represented by the client
! 1682: # socket.
! 1683: # Returns:
! 1684: # NONE
! 1685: #
! 1686: sub parent_listen {
! 1687: my ($loncapa_host) = @_;
! 1688: Debug(5, "parent_listen: $loncapa_host");
! 1689:
! 1690: my $socket = &SetupLoncListener($loncapa_host);
! 1691: if (!$socket) {
! 1692: die "Unable to create a listen socket for $loncapa_host";
! 1693: }
! 1694:
! 1695: my $lock_file = &GetLoncSocketPath().".lock";
! 1696: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
! 1697:
! 1698: Event->io(cb => &parent_client_connection,
! 1699: poll => 'r',
! 1700: desc => 'Parent listener unix socket',
! 1701: fd => $socket);
! 1702:
! 1703: }
! 1704:
! 1705:
! 1706: # listen_on_all_unix_sockets:
! 1707: # This sub initiates a listen on all unix domain lonc client sockets.
! 1708: # This will be called in the case where we are trimming idle processes.
! 1709: # When idle processes are trimmed, loncnew starts up with no children,
! 1710: # and only spawns off children when a connection request occurs on the
! 1711: # client unix socket. The spawned child continues to run until it has
! 1712: # been idle a while at which point it eventually exits and once more
! 1713: # the parent picks up the listen.
! 1714: #
! 1715: # Parameters:
! 1716: # NONE
! 1717: # Implicit Inputs:
! 1718: # The configuration file that has been read in by LondConnection.
! 1719: # Returns:
! 1720: # NONE
! 1721: #
! 1722: sub listen_on_all_unix_sockets {
! 1723: Debug(5, "listen_on_all_unix_sockets");
! 1724: my $host_iterator = &LondConnection::GetHostIterator();
! 1725: while (!$host_iterator->end()) {
! 1726: my $host_entry_ref = $host_iterator->get();
! 1727: my $host_name = $host_entry_ref->[0];
! 1728: Debug(9, "Listen for $host_name");
! 1729: &parent_listen($host_name);
! 1730: $host_iterator->next();
! 1731: }
1.1 foxr 1732: }
1.61 ! foxr 1733:
1.1 foxr 1734: #
1735: # Parent process logic pass 1:
1736: # For each entry in the hosts table, we will
1737: # fork off an instance of ChildProcess to service the transactions
1738: # to that host. Each pid will be entered in a global hash
1739: # with the value of the key, the host.
1740: # The parent will then enter a loop to wait for process exits.
1741: # Each exit gets logged and the child gets restarted.
1742: #
1743:
1.5 foxr 1744: #
1745: # Fork and start in new session so hang-up isn't going to
1746: # happen without intent.
1747: #
1748:
1749:
1.6 foxr 1750:
1751:
1.8 foxr 1752:
1.6 foxr 1753:
1754: ShowStatus("Forming new session");
1755: my $childpid = fork;
1756: if ($childpid != 0) {
1757: sleep 4; # Give child a chacne to break to
1758: exit 0; # a new sesion.
1759: }
1.8 foxr 1760: #
1761: # Write my pid into the pid file so I can be located
1762: #
1763:
1764: ShowStatus("Parent writing pid file:");
1.23 foxr 1765: my $execdir = $perlvar{'lonDaemons'};
1.8 foxr 1766: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1767: print PIDSAVE "$$\n";
1768: close(PIDSAVE);
1.6 foxr 1769:
1.17 foxr 1770:
1771:
1.6 foxr 1772: if (POSIX::setsid() < 0) {
1773: print "Could not create new session\n";
1774: exit -1;
1775: }
1.5 foxr 1776:
1777: ShowStatus("Forking node servers");
1778:
1.9 foxr 1779: Log("CRITICAL", "--------------- Starting children ---------------");
1780:
1.31 foxr 1781: LondConnection::ReadConfig; # Read standard config files.
1.1 foxr 1782: my $HostIterator = LondConnection::GetHostIterator;
1783:
1.60 foxr 1784: if ($DieWhenIdle) {
1.61 ! foxr 1785: $RemoteHost = "[parent]";
! 1786: &listen_on_all_unix_sockets();
1.60 foxr 1787: } else {
1788:
1789: while (! $HostIterator->end()) {
1790:
1791: my $hostentryref = $HostIterator->get();
1792: CreateChild($hostentryref->[0]);
1793: $HostHash{$hostentryref->[0]} = $hostentryref->[4];
1794: $HostIterator->next();
1795: }
1.1 foxr 1796: }
1.60 foxr 1797:
1.12 foxr 1798: $RemoteHost = "Parent Server";
1.1 foxr 1799:
1800: # Maintain the population:
1.5 foxr 1801:
1802: ShowStatus("Parent keeping the flock");
1.1 foxr 1803:
1.12 foxr 1804:
1.60 foxr 1805: if ($DieWhenIdle) {
1.61 ! foxr 1806: $Event::DebugLevel = $DebugLevel;
! 1807: Debug(9, "Parent entering event loop");
! 1808: my $ret = Event::loop();
! 1809: die "Main Event loop exited: $ret";
! 1810:
! 1811:
1.60 foxr 1812: } else {
1.61 ! foxr 1813: #
! 1814: # Set up parent signals:
! 1815: #
1.60 foxr 1816:
1817: $SIG{INT} = \&Terminate;
1818: $SIG{TERM} = \&Terminate;
1819: $SIG{HUP} = \&Restart;
1820: $SIG{USR1} = \&CheckKids;
1821: $SIG{USR2} = \&UpdateKids; # LonManage update request.
1822:
1823: while(1) {
1824: my $deadchild = wait();
1825: if(exists $ChildHash{$deadchild}) { # need to restart.
1826: my $deadhost = $ChildHash{$deadchild};
1827: delete($HostToPid{$deadhost});
1828: delete($ChildHash{$deadchild});
1829: Log("WARNING","Lost child pid= ".$deadchild.
1830: "Connected to host ".$deadhost);
1831: Log("INFO", "Restarting child procesing ".$deadhost);
1832: CreateChild($deadhost);
1833: }
1.1 foxr 1834: }
1.13 foxr 1835: }
1836:
1.14 foxr 1837:
1838: =pod
1839:
1840: =head1 CheckKids
1841:
1842: Since kids do not die as easily in this implementation
1843: as the previous one, there is no need to restart the
1844: dead ones (all dead kids get restarted when they die!!)
1845: The only thing this function does is to pass USR1 to the
1846: kids so that they report their status.
1847:
1848: =cut
1849:
1850: sub CheckKids {
1851: Debug(2, "Checking status of children");
1852: my $docdir = $perlvar{'lonDocRoot'};
1853: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
1854: my $now=time;
1855: my $local=localtime($now);
1856: print $fh "LONC status $local - parent $$ \n\n";
1.23 foxr 1857: foreach my $pid (keys %ChildHash) {
1.14 foxr 1858: Debug(2, "Sending USR1 -> $pid");
1859: kill 'USR1' => $pid; # Tell Child to report status.
1860: sleep 1; # Wait so file doesn't intermix.
1861: }
1862: }
1.24 foxr 1863:
1864: =pod
1865:
1866: =head1 UpdateKids
1867:
1.25 foxr 1868: parent's SIGUSR2 handler. This handler:
1.24 foxr 1869:
1870: =item
1871:
1872: Rereads the hosts file.
1873:
1874: =item
1875:
1876: Kills off (via sigint) children for hosts that have disappeared.
1877:
1878: =item
1879:
1.27 foxr 1880: QUITs children for hosts that already exist (this just forces a status display
1.24 foxr 1881: and resets the connection retry count for that host.
1882:
1883: =item
1884:
1885: Starts new children for hosts that have been added to the hosts.tab file since
1886: the start of the master program and maintains them.
1887:
1888: =cut
1889:
1890: sub UpdateKids {
1.27 foxr 1891:
1.25 foxr 1892: Log("INFO", "Updating connections via SIGUSR2");
1.27 foxr 1893:
1894: # Just in case we need to kill our own lonc, we wait a few seconds to
1895: # give it a chance to receive and relay lond's response to the
1896: # re-init command.
1897: #
1898:
1899: sleep(2); # Wait a couple of seconds.
1900:
1901: my %hosts; # Indexed by loncapa hostname, value=ip.
1902:
1903: # Need to re-read the host table:
1904:
1905:
1906: LondConnection::ReadConfig();
1907: my $I = LondConnection::GetHostIterator;
1908: while (! $I->end()) {
1909: my $item = $I->get();
1910: $hosts{$item->[0]} = $item->[4];
1911: $I->next();
1912: }
1913:
1914: # The logic below is written for clarity not for efficiency.
1915: # Since I anticipate that this function is only rarely called, that's
1916: # appropriate. There are certainly ways to combine the loops below,
1917: # and anyone wishing to obscure the logic is welcome to go for it.
1918: # Note that we don't re-direct sigchild. Instead we do what's needed
1919: # to the data structures that keep track of children to ensure that
1920: # when sigchild is honored, no new child is born.
1921: #
1922:
1923: # For each existing child; if it's host doesn't exist, kill the child.
1924:
1925: foreach my $child (keys %ChildHash) {
1926: my $oldhost = $ChildHash{$child};
1927: if (!(exists $hosts{$oldhost})) {
1928: Log("CRITICAL", "Killing child for $oldhost host no longer exists");
1929: delete $ChildHash{$child};
1930: delete $HostToPid{$oldhost};
1931: kill 'QUIT' => $child;
1932: }
1933: }
1934: # For each remaining existing child; if it's host's ip has changed,
1935: # Restart the child on the new IP.
1936:
1937: foreach my $child (keys %ChildHash) {
1938: my $oldhost = $ChildHash{$child};
1939: my $oldip = $HostHash{$oldhost};
1940: if ($hosts{$oldhost} ne $oldip) {
1941:
1942: # kill the old child.
1943:
1944: Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
1945: delete $ChildHash{$child};
1946: delete $HostToPid{$oldhost};
1947: kill 'QUIT' => $child;
1948:
1949: # Do the book-keeping needed to start a new child on the
1950: # new ip.
1951:
1952: $HostHash{$oldhost} = $hosts{$oldhost};
1953: CreateChild($oldhost);
1954: }
1955: }
1956: # Finally, for each new host, not in the host hash, create a
1957: # enter the host and create a new child.
1958: # Force a status display of any existing process.
1959:
1960: foreach my $host (keys %hosts) {
1961: if(!(exists $HostHash{$host})) {
1962: Log("INFO", "New host $host discovered in hosts.tab...");
1963: $HostHash{$host} = $hosts{$host};
1964: CreateChild($host);
1965: } else {
1966: kill 'HUP' => $HostToPid{$host}; # status display.
1967: }
1968: }
1.24 foxr 1969: }
1970:
1.14 foxr 1971:
1.13 foxr 1972: =pod
1973:
1974: =head1 Restart
1975:
1976: Signal handler for HUP... all children are killed and
1977: we self restart. This is an el-cheapo way to re read
1978: the config file.
1979:
1980: =cut
1981:
1982: sub Restart {
1.23 foxr 1983: &KillThemAll; # First kill all the children.
1.13 foxr 1984: Log("CRITICAL", "Restarting");
1985: my $execdir = $perlvar{'lonDaemons'};
1986: unlink("$execdir/logs/lonc.pid");
1.28 albertel 1987: exec("$execdir/loncnew");
1.10 foxr 1988: }
1.12 foxr 1989:
1990: =pod
1991:
1992: =head1 KillThemAll
1993:
1994: Signal handler that kills all children by sending them a
1.17 foxr 1995: SIGHUP. Responds to sigint and sigterm.
1.12 foxr 1996:
1997: =cut
1998:
1.10 foxr 1999: sub KillThemAll {
1.12 foxr 2000: Debug(2, "Kill them all!!");
2001: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
1.23 foxr 2002: foreach my $pid (keys %ChildHash) {
1.12 foxr 2003: my $serving = $ChildHash{$pid};
1.52 foxr 2004: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2005: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
1.17 foxr 2006: kill 'QUIT' => $pid;
1.12 foxr 2007: }
1.52 foxr 2008:
1.17 foxr 2009:
1.1 foxr 2010: }
1.12 foxr 2011:
1.52 foxr 2012:
2013: #
2014: # Kill all children via KILL. Just in case the
2015: # first shot didn't get them.
2016:
2017: sub really_kill_them_all_dammit
2018: {
2019: Debug(2, "Kill them all Dammit");
2020: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
2021: foreach my $pid (keys %ChildHash) {
2022: my $serving = $ChildHash{$pid};
2023: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2024: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2025: kill 'KILL' => $pid;
2026: delete($ChildHash{$pid});
2027: my $execdir = $perlvar{'lonDaemons'};
2028: unlink("$execdir/logs/lonc.pid");
2029: }
2030: }
1.14 foxr 2031: =pod
2032:
2033: =head1 Terminate
2034:
2035: Terminate the system.
2036:
2037: =cut
2038:
2039: sub Terminate {
1.52 foxr 2040: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2041: &KillThemAll;
2042: #
2043: # By now they really should all be dead.. but just in case
2044: # send them all SIGKILL's after a bit of waiting:
2045:
2046: sleep(4);
2047: &Log("CRITICAL", "Now kill children nasty");
2048: &really_kill_them_all_dammit;
1.17 foxr 2049: Log("CRITICAL","Master process exiting");
2050: exit 0;
1.14 foxr 2051:
2052: }
1.12 foxr 2053: =pod
1.1 foxr 2054:
2055: =head1 Theory
1.3 albertel 2056:
2057: The event class is used to build this as a single process with an
2058: event driven model. The following events are handled:
1.1 foxr 2059:
2060: =item UNIX Socket connection Received
2061:
2062: =item Request data arrives on UNIX data transfer socket.
2063:
2064: =item lond connection becomes writable.
2065:
2066: =item timer fires at 1 second intervals.
2067:
2068: All sockets are run in non-blocking mode. Timeouts managed by the timer
2069: handler prevents hung connections.
2070:
2071: Key data structures:
2072:
1.3 albertel 2073: =item RequestQueue
2074:
2075: A queue of requests received from UNIX sockets that are
2076: waiting for a chance to be forwarded on a lond connection socket.
2077:
2078: =item ActiveConnections
2079:
2080: A hash of lond connections that have transactions in process that are
2081: available to be timed out.
2082:
2083: =item ActiveTransactions
2084:
2085: A hash indexed by lond connections that contain the client reply
2086: socket for each connection that has an active transaction on it.
2087:
2088: =item IdleConnections
2089:
2090: A hash of lond connections that have no work to do. These connections
2091: can be closed if they are idle for a long enough time.
1.1 foxr 2092:
2093: =cut
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/back.gif){kind=icon}
Return to loncnew CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [LON-CAPA] / loncom