1: #!/usr/bin/perl
2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
5: # $Id: loncnew,v 1.50 2004/07/02 09:28:14 albertel Exp $
6: #
7: # Copyright Michigan State University Board of Trustees
8: #
9: # This file is part of the LearningOnline Network with CAPA (LON-CAPA).
10: ## LON-CAPA is free software; you can redistribute it and/or modify
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: #
28: #
29: # new lonc handles n request out bver m connections to londs.
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).
36: # - Add general logic for dispatching requests and timeouts. (done).
37: # - Add support for the lonc/lond requests. (done).
38: # - Add logging/status monitoring. (done)
39: # - Add Signal handling - HUP restarts. USR1 status report. (done)
40: # - Add Configuration file I/O (done).
41: # - Add management/status request interface. (done)
42: # - Add deferred request capability. (done)
43: # - Detect transmission timeouts. (done)
44: #
45:
46: use strict;
47: use lib "/home/httpd/lib/perl/";
48: use Event qw(:DEFAULT );
49: use POSIX qw(:signal_h);
50: use POSIX;
51: use IO::Socket;
52: use IO::Socket::INET;
53: use IO::Socket::UNIX;
54: use IO::File;
55: use IO::Handle;
56: use Socket;
57: use Crypt::IDEA;
58: use LONCAPA::Queue;
59: use LONCAPA::Stack;
60: use LONCAPA::LondConnection;
61: use LONCAPA::LondTransaction;
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.
76: my %HostToPid; # By host -> pid.
77: my %HostHash; # by loncapaname -> IP.
78:
79:
80: my $MaxConnectionCount = 10; # Will get from config later.
81: my $ClientConnection = 0; # Uniquifier for client events.
82:
83: my $DebugLevel = 0;
84: my $NextDebugLevel= 2; # So Sigint can toggle this.
85: my $IdleTimeout= 600; # Wait 10 minutes before pruning connections.
86:
87: my $LogTransactions = 0; # When True, all transactions/replies get logged.
88:
89: #
90: # The variables below are only used by the child processes.
91: #
92: my $RemoteHost; # Name of host child is talking to.
93: my $UnixSocketDir= $perlvar{'lonSockDir'};
94: my $IdleConnections = Stack->new(); # Set of idle connections
95: my %ActiveConnections; # Connections to the remote lond.
96: my %ActiveTransactions; # LondTransactions in flight.
97: my %ActiveClients; # Serial numbers of active clients by socket.
98: my $WorkQueue = Queue->new(); # Queue of pending transactions.
99: my $ConnectionCount = 0;
100: my $IdleSeconds = 0; # Number of seconds idle.
101: my $Status = ""; # Current status string.
102: my $RecentLogEntry = "";
103: my $ConnectionRetries=2; # Number of connection retries allowed.
104: my $ConnectionRetriesLeft=2; # Number of connection retries remaining.
105: my $LondVersion = "unknown"; # Version of lond we talk with.
106: my $KeyMode = ""; # e.g. ssl, local, insecure from last connect.
107:
108: #
109: # The hash below gives the HTML format for log messages
110: # given a severity.
111: #
112: my %LogFormats;
113:
114: $LogFormats{"CRITICAL"} = "<font color='red'>CRITICAL: %s</font>";
115: $LogFormats{"SUCCESS"} = "<font color='green'>SUCCESS: %s</font>";
116: $LogFormats{"INFO"} = "<font color='yellow'>INFO: %s</font>";
117: $LogFormats{"WARNING"} = "<font color='blue'>WARNING: %s</font>";
118: $LogFormats{"DEFAULT"} = " %s ";
119:
120:
121:
122: =pod
123:
124: =head2 LogPerm
125:
126: Makes an entry into the permanent log file.
127:
128: =cut
129: sub LogPerm {
130: my $message=shift;
131: my $execdir=$perlvar{'lonDaemons'};
132: my $now=time;
133: my $local=localtime($now);
134: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
135: print $fh "$now:$message:$local\n";
136: }
137:
138: =pod
139:
140: =head2 Log
141:
142: Logs a message to the log file.
143: Parameters:
144:
145: =item severity
146:
147: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
148: format string used to format the message. if the severity is
149: not a defined severity the Default format string is used.
150:
151: =item message
152:
153: The base message. In addtion to the format string, the message
154: will be appended to a string containing the name of our remote
155: host and the time will be formatted into the message.
156:
157: =cut
158:
159: sub Log {
160:
161: my ($severity, $message) = @_;
162:
163: if(!$LogFormats{$severity}) {
164: $severity = "DEFAULT";
165: }
166:
167: my $format = $LogFormats{$severity};
168:
169: # Put the window dressing in in front of the message format:
170:
171: my $now = time;
172: my $local = localtime($now);
173: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
174: my $finalformat = $finalformat.$format."\n";
175:
176: # open the file and put the result.
177:
178: my $execdir = $perlvar{'lonDaemons'};
179: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
180: my $msg = sprintf($finalformat, $message);
181: $RecentLogEntry = $msg;
182: print $fh $msg;
183:
184:
185: }
186:
187:
188: =pod
189:
190: =head2 GetPeerName
191:
192: Returns the name of the host that a socket object is connected to.
193:
194: =cut
195:
196: sub GetPeername {
197:
198:
199: my ($connection, $AdrFamily) = @_;
200:
201: my $peer = $connection->peername();
202: my $peerport;
203: my $peerip;
204: if($AdrFamily == AF_INET) {
205: ($peerport, $peerip) = sockaddr_in($peer);
206: my $peername = gethostbyaddr($peerip, $AdrFamily);
207: return $peername;
208: } elsif ($AdrFamily == AF_UNIX) {
209: my $peerfile;
210: ($peerfile) = sockaddr_un($peer);
211: return $peerfile;
212: }
213: }
214: =pod
215:
216: =head2 Debug
217:
218: Invoked to issue a debug message.
219:
220: =cut
221:
222: sub Debug {
223:
224: my ($level, $message) = @_;
225:
226: if ($level <= $DebugLevel) {
227: Log("INFO", "-Debug- $message host = $RemoteHost");
228: }
229: }
230:
231: sub SocketDump {
232:
233: my ($level, $socket) = @_;
234:
235: if($level <= $DebugLevel) {
236: $socket->Dump(-1); # Ensure it will get dumped.
237: }
238: }
239:
240: =pod
241:
242: =head2 ShowStatus
243:
244: Place some text as our pid status.
245: and as what we return in a SIGUSR1
246:
247: =cut
248: sub ShowStatus {
249: my $state = shift;
250: my $now = time;
251: my $local = localtime($now);
252: $Status = $local.": ".$state;
253: $0='lonc: '.$state.' '.$local;
254: }
255:
256: =pod
257:
258: =head 2 SocketTimeout
259:
260: Called when an action on the socket times out. The socket is
261: destroyed and any active transaction is failed.
262:
263:
264: =cut
265: sub SocketTimeout {
266: my $Socket = shift;
267: Log("WARNING", "A socket timeout was detected");
268: Debug(0, " SocketTimeout called: ");
269: $Socket->Dump(0);
270: if(exists($ActiveTransactions{$Socket})) {
271: FailTransaction($ActiveTransactions{$Socket});
272: }
273: KillSocket($Socket); # A transaction timeout also counts as
274: # a connection failure:
275: $ConnectionRetriesLeft--;
276: if($ConnectionRetriesLeft <= 0) {
277: Log("CRITICAL", "Host marked dead: ".GetServerHost());
278: }
279:
280: }
281: #----------------------------- Timer management ------------------------
282:
283: =pod
284:
285: =head2 Tick
286:
287: Invoked each timer tick.
288:
289: =cut
290:
291:
292: sub Tick {
293: my $client;
294: if($ConnectionRetriesLeft > 0) {
295: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
296: ." Retries remaining: ".$ConnectionRetriesLeft
297: ." ($KeyMode)");
298: } else {
299: ShowStatus(GetServerHost()." >> DEAD <<");
300: }
301: # Is it time to prune connection count:
302:
303:
304: if($IdleConnections->Count() &&
305: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
306: $IdleSeconds++;
307: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
308: my $Socket = $IdleConnections->pop();
309: KillSocket($Socket);
310: }
311: } else {
312: $IdleSeconds = 0; # Reset idle count if not idle.
313: }
314: #
315: # For each inflight transaction, tick down its timeout counter.
316: #
317:
318: foreach my $item (keys %ActiveConnections) {
319: my $State = $ActiveConnections{$item}->data->GetState();
320: if ($State ne 'Idle') {
321: Debug(5,"Ticking Socket $State $item");
322: $ActiveConnections{$item}->data->Tick();
323: }
324: }
325: # Do we have work in the queue, but no connections to service them?
326: # If so, try to make some new connections to get things going again.
327: #
328:
329: my $Requests = $WorkQueue->Count();
330: if (($ConnectionCount == 0) && ($Requests > 0)) {
331: if ($ConnectionRetriesLeft > 0) {
332: my $Connections = ($Requests <= $MaxConnectionCount) ?
333: $Requests : $MaxConnectionCount;
334: Debug(5,"Work but no connections, start ".$Connections." of them");
335: my $successCount = 0;
336: for (my $i =0; $i < $Connections; $i++) {
337: $successCount += MakeLondConnection();
338: }
339: if($successCount == 0) { # All connections failed:
340: Debug(5,"Work in queue failed to make any connectiouns\n");
341: EmptyQueue(); # Fail pending transactions with con_lost.
342: CloseAllLondConnections(); # Should all be closed but....
343: }
344: } else {
345: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
346: Debug(5,"Work in queue, but gave up on connections..flushing\n");
347: EmptyQueue(); # Connections can't be established.
348: CloseAllLondConnections(); # Should all already be closed but...
349: }
350:
351: }
352: if ($ConnectionCount == 0) {
353: $KeyMode = "";
354: }
355: }
356:
357: =pod
358:
359: =head2 SetupTimer
360:
361: Sets up a 1 per sec recurring timer event. The event handler is used to:
362:
363: =item
364:
365: Trigger timeouts on communications along active sockets.
366:
367: =item
368:
369: Trigger disconnections of idle sockets.
370:
371: =cut
372:
373: sub SetupTimer {
374: Debug(6, "SetupTimer");
375: Event->timer(interval => 1, cb => \&Tick );
376: }
377:
378: =pod
379:
380: =head2 ServerToIdle
381:
382: This function is called when a connection to the server is
383: ready for more work.
384:
385: If there is work in the Work queue the top element is dequeued
386: and the connection will start to work on it. If the work queue is
387: empty, the connection is pushed on the idle connection stack where
388: it will either get another work unit, or alternatively, if it sits there
389: long enough, it will be shut down and released.
390:
391: =cut
392:
393: sub ServerToIdle {
394: my $Socket = shift; # Get the socket.
395: $KeyMode = $Socket->{AuthenticationMode};
396: delete($ActiveTransactions{$Socket}); # Server has no transaction
397:
398: &Debug(5, "Server to idle");
399:
400: # If there's work to do, start the transaction:
401:
402: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
403: if ($reqdata ne undef) {
404: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
405: &StartRequest($Socket, $reqdata);
406:
407: } else {
408:
409: # There's no work waiting, so push the server to idle list.
410: &Debug(5, "No new work requests, server connection going idle");
411: $IdleConnections->push($Socket);
412: }
413: }
414:
415: =pod
416:
417: =head2 ClientWritable
418:
419: Event callback for when a client socket is writable.
420:
421: This callback is established when a transaction reponse is
422: avaiable from lond. The response is forwarded to the unix socket
423: as it becomes writable in this sub.
424:
425: Parameters:
426:
427: =item Event
428:
429: The event that has been triggered. Event->w->data is
430: the data and Event->w->fd is the socket to write.
431:
432: =cut
433:
434: sub ClientWritable {
435: my $Event = shift;
436: my $Watcher = $Event->w;
437: my $Data = $Watcher->data;
438: my $Socket = $Watcher->fd;
439:
440: # Try to send the data:
441:
442: &Debug(6, "ClientWritable writing".$Data);
443: &Debug(9, "Socket is: ".$Socket);
444:
445: if($Socket->connected) {
446: my $result = $Socket->send($Data, 0);
447:
448: # $result undefined: the write failed.
449: # otherwise $result is the number of bytes written.
450: # Remove that preceding string from the data.
451: # If the resulting data is empty, destroy the watcher
452: # and set up a read event handler to accept the next
453: # request.
454:
455: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
456: if($result ne undef) {
457: &Debug(9, "send result was defined");
458: if($result == length($Data)) { # Entire string sent.
459: &Debug(9, "ClientWritable data all written");
460: $Watcher->cancel();
461: #
462: # Set up to read next request from socket:
463:
464: my $descr = sprintf("Connection to lonc client %d",
465: $ActiveClients{$Socket});
466: Event->io(cb => \&ClientRequest,
467: poll => 'r',
468: desc => $descr,
469: data => "",
470: fd => $Socket);
471:
472: } else { # Partial string sent.
473: $Watcher->data(substr($Data, $result));
474: if($result == 0) { # client hung up on us!!
475: Log("INFO", "lonc pipe client hung up on us!");
476: $Watcher->cancel;
477: $Socket->shutdown(2);
478: $Socket->close();
479: }
480: }
481:
482: } else { # Error of some sort...
483:
484: # Some errnos are possible:
485: my $errno = $!;
486: if($errno == POSIX::EWOULDBLOCK ||
487: $errno == POSIX::EAGAIN ||
488: $errno == POSIX::EINTR) {
489: # No action taken?
490: } else { # Unanticipated errno.
491: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
492: $Watcher->cancel; # Stop the watcher.
493: $Socket->shutdown(2); # Kill connection
494: $Socket->close(); # Close the socket.
495: }
496:
497: }
498: } else {
499: $Watcher->cancel(); # A delayed request...just cancel.
500: }
501: }
502:
503: =pod
504:
505: =head2 CompleteTransaction
506:
507: Called when the reply data has been received for a lond
508: transaction. The reply data must now be sent to the
509: ultimate client on the other end of the Unix socket. This is
510: done by setting up a writable event for the socket with the
511: data the reply data.
512:
513: Parameters:
514:
515: =item Socket
516:
517: Socket on which the lond transaction occured. This is a
518: LondConnection. The data received is in the TransactionReply member.
519:
520: =item Transaction
521:
522: The transaction that is being completed.
523:
524: =cut
525:
526: sub CompleteTransaction {
527: &Debug(5,"Complete transaction");
528:
529: my ($Socket, $Transaction) = @_;
530:
531: if (!$Transaction->isDeferred()) { # Normal transaction
532: my $data = $Socket->GetReply(); # Data to send.
533: if($LogTransactions) {
534: Log("SUCCESS", "Reply from lond: '$data'");
535: }
536: StartClientReply($Transaction, $data);
537: } else { # Delete deferred transaction file.
538: Log("SUCCESS", "A delayed transaction was completed");
539: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
540: unlink $Transaction->getFile();
541: }
542: }
543:
544: =pod
545:
546: =head1 StartClientReply
547:
548: Initiates a reply to a client where the reply data is a parameter.
549:
550: =head2 parameters:
551:
552: =item Transaction
553:
554: The transaction for which we are responding to the client.
555:
556: =item data
557:
558: The data to send to apached client.
559:
560: =cut
561:
562: sub StartClientReply {
563:
564: my ($Transaction, $data) = @_;
565:
566: my $Client = $Transaction->getClient();
567:
568: &Debug(8," Reply was: ".$data);
569: my $Serial = $ActiveClients{$Client};
570: my $desc = sprintf("Connection to lonc client %d",
571: $Serial);
572: Event->io(fd => $Client,
573: poll => "w",
574: desc => $desc,
575: cb => \&ClientWritable,
576: data => $data);
577: }
578:
579: =pod
580:
581: =head2 FailTransaction
582:
583: Finishes a transaction with failure because the associated lond socket
584: disconnected. There are two possibilities:
585: - The transaction is deferred: in which case we just quietly
586: delete the transaction since there is no client connection.
587: - The transaction is 'live' in which case we initiate the sending
588: of "con_lost" to the client.
589:
590: Deleting the transaction means killing it from the %ActiveTransactions hash.
591:
592: Parameters:
593:
594: =item client
595:
596: The LondTransaction we are failing.
597:
598:
599: =cut
600:
601: sub FailTransaction {
602: my $transaction = shift;
603: Log("WARNING", "Failing transaction ".$transaction->getRequest());
604: Debug(1, "Failing transaction: ".$transaction->getRequest());
605: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
606: my $client = $transaction->getClient();
607: Debug(1," Replying con_lost to ".$transaction->getRequest());
608: StartClientReply($transaction, "con_lost\n");
609: }
610:
611: }
612:
613: =pod
614: =head1 EmptyQueue
615:
616: Fails all items in the work queue with con_lost.
617: Note that each item in the work queue is a transaction.
618:
619: =cut
620: sub EmptyQueue {
621: $ConnectionRetriesLeft--; # Counts as connection failure too.
622: while($WorkQueue->Count()) {
623: my $request = $WorkQueue->dequeue(); # This is a transaction
624: FailTransaction($request);
625: }
626: }
627:
628: =pod
629:
630: =head2 CloseAllLondConnections
631:
632: Close all connections open on lond prior to exit e.g.
633:
634: =cut
635: sub CloseAllLondConnections {
636: foreach my $Socket (keys %ActiveConnections) {
637: if(exists($ActiveTransactions{$Socket})) {
638: FailTransaction($ActiveTransactions{$Socket});
639: }
640: KillSocket($Socket);
641: }
642: }
643: =cut
644:
645: =pod
646:
647: =head2 KillSocket
648:
649: Destroys a socket. This function can be called either when a socket
650: has died of 'natural' causes or because a socket needs to be pruned due to
651: idleness. If the socket has died naturally, if there are no longer any
652: live connections a new connection is created (in case there are transactions
653: in the queue). If the socket has been pruned, it is never re-created.
654:
655: Parameters:
656:
657: =item Socket
658:
659: The socket to kill off.
660:
661: =item Restart
662:
663: nonzero if we are allowed to create a new connection.
664:
665:
666: =cut
667: sub KillSocket {
668: my $Socket = shift;
669:
670: Log("WARNING", "Shutting down a socket");
671: $Socket->Shutdown();
672:
673: # If the socket came from the active connection set,
674: # delete its transaction... note that FailTransaction should
675: # already have been called!!!
676: # otherwise it came from the idle set.
677: #
678:
679: if(exists($ActiveTransactions{$Socket})) {
680: delete ($ActiveTransactions{$Socket});
681: }
682: if(exists($ActiveConnections{$Socket})) {
683: delete($ActiveConnections{$Socket});
684: $ConnectionCount--;
685: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
686: }
687: # If the connection count has gone to zero and there is work in the
688: # work queue, the work all gets failed with con_lost.
689: #
690: if($ConnectionCount == 0) {
691: EmptyQueue();
692: CloseAllLondConnections; # Should all already be closed but...
693: }
694: }
695:
696: =pod
697:
698: =head2 LondReadable
699:
700: This function is called whenever a lond connection
701: is readable. The action is state dependent:
702:
703: =head3 State=Initialized
704:
705: We''re waiting for the challenge, this is a no-op until the
706: state changes.
707:
708: =head3 State=Challenged
709:
710: The challenge has arrived we need to transition to Writable.
711: The connection must echo the challenge back.
712:
713: =head3 State=ChallengeReplied
714:
715: The challenge has been replied to. The we are receiveing the
716: 'ok' from the partner.
717:
718: =head3 State=ReadingVersionString
719:
720: We have requested the lond version and are reading the
721: version back. Upon completion, we'll store the version away
722: for future use(?).
723:
724: =head3 State=HostSet
725:
726: We have selected the domain name of our peer (multhomed hosts)
727: and are getting the reply (presumably ok) back.
728:
729: =head3 State=RequestingKey
730:
731: The ok has been received and we need to send the request for
732: an encryption key. Transition to writable for that.
733:
734: =head3 State=ReceivingKey
735:
736: The the key has been requested, now we are reading the new key.
737:
738: =head3 State=Idle
739:
740: The encryption key has been negotiated or we have finished
741: reading data from the a transaction. If the callback data has
742: a client as well as the socket iformation, then we are
743: doing a transaction and the data received is relayed to the client
744: before the socket is put on the idle list.
745:
746: =head3 State=SendingRequest
747:
748: I do not think this state can be received here, but if it is,
749: the appropriate thing to do is to transition to writable, and send
750: the request.
751:
752: =head3 State=ReceivingReply
753:
754: We finished sending the request to the server and now transition
755: to readable to receive the reply.
756:
757: The parameter to this function are:
758:
759: The event. Implicit in this is the watcher and its data. The data
760: contains at least the lond connection object and, if a
761: transaction is in progress, the socket attached to the local client.
762:
763: =cut
764:
765: sub LondReadable {
766:
767: my $Event = shift;
768: my $Watcher = $Event->w;
769: my $Socket = $Watcher->data;
770: my $client = undef;
771:
772: &Debug(6,"LondReadable called state = ".$Socket->GetState());
773:
774:
775: my $State = $Socket->GetState(); # All action depends on the state.
776:
777: SocketDump(6, $Socket);
778: my $status = $Socket->Readable();
779:
780: &Debug(2, "Socket->Readable returned: $status");
781:
782: if($status != 0) {
783: # bad return from socket read. Currently this means that
784: # The socket has become disconnected. We fail the transaction.
785:
786: Log("WARNING",
787: "Lond connection lost.");
788: if(exists($ActiveTransactions{$Socket})) {
789: FailTransaction($ActiveTransactions{$Socket});
790: }
791: $Watcher->cancel();
792: KillSocket($Socket);
793: $ConnectionRetriesLeft--; # Counts as connection failure
794: return;
795: }
796: SocketDump(6,$Socket);
797:
798: $State = $Socket->GetState(); # Update in case of transition.
799: &Debug(6, "After read, state is ".$State);
800:
801: if($State eq "Initialized") {
802:
803:
804: } elsif ($State eq "ChallengeReceived") {
805: # The challenge must be echoed back; The state machine
806: # in the connection takes care of setting that up. Just
807: # need to transition to writable:
808:
809: $Watcher->cb(\&LondWritable);
810: $Watcher->poll("w");
811:
812: } elsif ($State eq "ChallengeReplied") {
813:
814: } elsif ($State eq "RequestingVersion") {
815: # Need to ask for the version... that is writiability:
816:
817: $Watcher->cb(\&LondWritable);
818: $Watcher->poll("w");
819:
820: } elsif ($State eq "ReadingVersionString") {
821: # Read the rest of the version string...
822: } elsif ($State eq "SetHost") {
823: # Need to request the actual domain get set...
824:
825: $Watcher->cb(\&LondWritable);
826: $Watcher->poll("w");
827: } elsif ($State eq "HostSet") {
828: # Reading the 'ok' from the peer.
829:
830: } elsif ($State eq "RequestingKey") {
831: # The ok was received. Now we need to request the key
832: # That requires us to be writable:
833:
834: $Watcher->cb(\&LondWritable);
835: $Watcher->poll("w");
836:
837: } elsif ($State eq "ReceivingKey") {
838:
839: } elsif ($State eq "Idle") {
840:
841: # This is as good a spot as any to get the peer version
842: # string:
843:
844: if($LondVersion eq "unknown") {
845: $LondVersion = $Socket->PeerVersion();
846: Log("INFO", "Connected to lond version: $LondVersion");
847: }
848: # If necessary, complete a transaction and then go into the
849: # idle queue.
850: # Note that a trasition to idle indicates a live lond
851: # on the other end so reset the connection retries.
852: #
853: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
854: $Watcher->cancel();
855: if(exists($ActiveTransactions{$Socket})) {
856: Debug(5,"Completing transaction!!");
857: CompleteTransaction($Socket,
858: $ActiveTransactions{$Socket});
859: } else {
860: Log("SUCCESS", "Connection ".$ConnectionCount." to "
861: .$RemoteHost." now ready for action");
862: }
863: ServerToIdle($Socket); # Next work unit or idle.
864:
865: } elsif ($State eq "SendingRequest") {
866: # We need to be writable for this and probably don't belong
867: # here inthe first place.
868:
869: Deubg(6, "SendingRequest state encountered in readable");
870: $Watcher->poll("w");
871: $Watcher->cb(\&LondWritable);
872:
873: } elsif ($State eq "ReceivingReply") {
874:
875:
876: } else {
877: # Invalid state.
878: Debug(4, "Invalid state in LondReadable");
879: }
880: }
881:
882: =pod
883:
884: =head2 LondWritable
885:
886: This function is called whenever a lond connection
887: becomes writable while there is a writeable monitoring
888: event. The action taken is very state dependent:
889:
890: =head3 State = Connected
891:
892: The connection is in the process of sending the 'init' hailing to the
893: lond on the remote end. The connection object''s Writable member is
894: called. On error, ConnectionError is called to destroy the connection
895: and remove it from the ActiveConnections hash
896:
897: =head3 Initialized
898:
899: 'init' has been sent, writability monitoring is removed and
900: readability monitoring is started with LondReadable as the callback.
901:
902: =head3 ChallengeReceived
903:
904: The connection has received the who are you challenge from the remote
905: system, and is in the process of sending the challenge
906: response. Writable is called.
907:
908: =head3 ChallengeReplied
909:
910: The connection has replied to the initial challenge The we switch to
911: monitoring readability looking for the server to reply with 'ok'.
912:
913: =head3 RequestingKey
914:
915: The connection is in the process of requesting its encryption key.
916: Writable is called.
917:
918: =head3 ReceivingKey
919:
920: The connection has sent the request for a key. Switch to readability
921: monitoring to accept the key
922:
923: =head3 SendingRequest
924:
925: The connection is in the process of sending a request to the server.
926: This request is part of a client transaction. All the states until
927: now represent the client setup protocol. Writable is called.
928:
929: =head3 ReceivingReply
930:
931: The connection has sent a request. Now it must receive a reply.
932: Readability monitoring is requested.
933:
934: This function is an event handler and therefore receives as
935: a parameter the event that has fired. The data for the watcher
936: of this event is a reference to a list of one or two elements,
937: depending on state. The first (and possibly only) element is the
938: socket. The second (present only if a request is in progress)
939: is the socket on which to return a reply to the caller.
940:
941: =cut
942:
943: sub LondWritable {
944: my $Event = shift;
945: my $Watcher = $Event->w;
946: my $Socket = $Watcher->data;
947: my $State = $Socket->GetState();
948:
949: Debug(6,"LondWritable State = ".$State."\n");
950:
951:
952: # Figure out what to do depending on the state of the socket:
953:
954:
955:
956:
957: SocketDump(6,$Socket);
958:
959: # If the socket is writable, we must always write.
960: # Only by writing will we undergo state transitions.
961: # Old logic wrote in state specific code below, however
962: # That forces us at least through another invocation of
963: # this function after writability is possible again.
964: # This logic also factors out common code for handling
965: # write failures... in all cases, write failures
966: # Kill the socket.
967: # This logic makes the branches of the >big< if below
968: # so that the writing states are actually NO-OPs.
969:
970: if ($Socket->Writable() != 0) {
971: # The write resulted in an error.
972: # We'll treat this as if the socket got disconnected:
973: Log("WARNING", "Connection to ".$RemoteHost.
974: " has been disconnected");
975: if(exists($ActiveTransactions{$Socket})) {
976: FailTransaction($ActiveTransactions{$Socket});
977: }
978: $Watcher->cancel();
979: KillSocket($Socket);
980: return;
981: }
982:
983:
984:
985: if ($State eq "Connected") {
986:
987: # "init" is being sent...
988:
989: } elsif ($State eq "Initialized") {
990:
991: # Now that init was sent, we switch
992: # to watching for readability:
993:
994: $Watcher->cb(\&LondReadable);
995: $Watcher->poll("r");
996:
997: } elsif ($State eq "ChallengeReceived") {
998: # We received the challenge, now we
999: # are echoing it back. This is a no-op,
1000: # we're waiting for the state to change
1001:
1002: } elsif ($State eq "ChallengeReplied") {
1003: # The echo was sent back, so we switch
1004: # to watching readability.
1005:
1006: $Watcher->cb(\&LondReadable);
1007: $Watcher->poll("r");
1008: } elsif ($State eq "RequestingVersion") {
1009: # Sending the peer a version request...
1010:
1011: } elsif ($State eq "ReadingVersionString") {
1012: # Transition to read since we have sent the
1013: # version command and now just need to read the
1014: # version string from the peer:
1015:
1016: $Watcher->cb(\&LondReadable);
1017: $Watcher->poll("r");
1018:
1019: } elsif ($State eq "SetHost") {
1020: # Setting the remote domain...
1021:
1022: } elsif ($State eq "HostSet") {
1023: # Back to readable to get the ok.
1024:
1025: $Watcher->cb(\&LondReadable);
1026: $Watcher->poll("r");
1027:
1028:
1029: } elsif ($State eq "RequestingKey") {
1030: # At this time we're requesting the key.
1031: # again, this is essentially a no-op.
1032:
1033: } elsif ($State eq "ReceivingKey") {
1034: # Now we need to wait for the key
1035: # to come back from the peer:
1036:
1037: $Watcher->cb(\&LondReadable);
1038: $Watcher->poll("r");
1039:
1040: } elsif ($State eq "SendingRequest") {
1041:
1042: # At this time we are sending a request to the
1043: # peer... write the next chunk:
1044:
1045:
1046: } elsif ($State eq "ReceivingReply") {
1047: # The send has completed. Wait for the
1048: # data to come in for a reply.
1049: Debug(8,"Writable sent request/receiving reply");
1050: $Watcher->cb(\&LondReadable);
1051: $Watcher->poll("r");
1052:
1053: } else {
1054: # Control only passes here on an error:
1055: # the socket state does not match any
1056: # of the known states... so an error
1057: # must be logged.
1058:
1059: &Debug(4, "Invalid socket state ".$State."\n");
1060: }
1061:
1062: }
1063: =pod
1064:
1065: =cut
1066: sub QueueDelayed {
1067: Debug(3,"QueueDelayed called");
1068:
1069: my $path = "$perlvar{'lonSockDir'}/delayed";
1070:
1071: Debug(4, "Delayed path: ".$path);
1072: opendir(DIRHANDLE, $path);
1073:
1074: my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
1075: closedir(DIRHANDLE);
1076: my $dfname;
1077: my $reqfile;
1078: foreach $dfname (sort @alldelayed) {
1079: $reqfile = "$path/$dfname";
1080: Debug(4, "queueing ".$reqfile);
1081: my $Handle = IO::File->new($reqfile);
1082: my $cmd = <$Handle>;
1083: chomp $cmd; # There may or may not be a newline...
1084: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1085: my $Transaction = LondTransaction->new($cmd);
1086: $Transaction->SetDeferred($reqfile);
1087: QueueTransaction($Transaction);
1088: }
1089:
1090: }
1091:
1092: =pod
1093:
1094: =head2 MakeLondConnection
1095:
1096: Create a new lond connection object, and start it towards its initial
1097: idleness. Once idle, it becomes elligible to receive transactions
1098: from the work queue. If the work queue is not empty when the
1099: connection is completed and becomes idle, it will dequeue an entry and
1100: start off on it.
1101:
1102: =cut
1103:
1104: sub MakeLondConnection {
1105: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1106: .GetServerPort());
1107:
1108: my $Connection = LondConnection->new(&GetServerHost(),
1109: &GetServerPort());
1110:
1111: if($Connection eq undef) { # Needs to be more robust later.
1112: Log("CRITICAL","Failed to make a connection with lond.");
1113: $ConnectionRetriesLeft--;
1114: return 0; # Failure.
1115: } else {
1116:
1117: # The connection needs to have writability
1118: # monitored in order to send the init sequence
1119: # that starts the whole authentication/key
1120: # exchange underway.
1121: #
1122: my $Socket = $Connection->GetSocket();
1123: if($Socket eq undef) {
1124: die "did not get a socket from the connection";
1125: } else {
1126: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1127: }
1128:
1129: $Connection->SetTimeoutCallback(\&SocketTimeout);
1130:
1131: my $event = Event->io(fd => $Socket,
1132: poll => 'w',
1133: cb => \&LondWritable,
1134: data => $Connection,
1135: desc => 'Connection to lond server');
1136: $ActiveConnections{$Connection} = $event;
1137:
1138: $ConnectionCount++;
1139: Debug(4, "Connection count = ".$ConnectionCount);
1140: if($ConnectionCount == 1) { # First Connection:
1141: QueueDelayed;
1142: }
1143: Log("SUCESS", "Created connection ".$ConnectionCount
1144: ." to host ".GetServerHost());
1145: return 1; # Return success.
1146: }
1147:
1148: }
1149:
1150: =pod
1151:
1152: =head2 StartRequest
1153:
1154: Starts a lond request going on a specified lond connection.
1155: parameters are:
1156:
1157: =item $Lond
1158:
1159: Connection to the lond that will send the transaction and receive the
1160: reply.
1161:
1162: =item $Client
1163:
1164: Connection to the client that is making this request We got the
1165: request from this socket, and when the request has been relayed to
1166: lond and we get a reply back from lond it will get sent to this
1167: socket.
1168:
1169: =item $Request
1170:
1171: The text of the request to send.
1172:
1173: =cut
1174:
1175: sub StartRequest {
1176:
1177: my ($Lond, $Request) = @_;
1178:
1179: Debug(6, "StartRequest: ".$Request->getRequest());
1180:
1181: my $Socket = $Lond->GetSocket();
1182:
1183: $Request->Activate($Lond);
1184: $ActiveTransactions{$Lond} = $Request;
1185:
1186: $Lond->InitiateTransaction($Request->getRequest());
1187: my $event = Event->io(fd => $Socket,
1188: poll => "w",
1189: cb => \&LondWritable,
1190: data => $Lond,
1191: desc => "lond transaction connection");
1192: $ActiveConnections{$Lond} = $event;
1193: Debug(8," Start Request made watcher data with ".$event->data."\n");
1194: }
1195:
1196: =pod
1197:
1198: =head2 QueueTransaction
1199:
1200: If there is an idle lond connection, it is put to work doing this
1201: transaction. Otherwise, the transaction is placed in the work queue.
1202: If placed in the work queue and the maximum number of connections has
1203: not yet been created, a new connection will be started. Our goal is
1204: to eventually have a sufficient number of connections that the work
1205: queue will typically be empty. parameters are:
1206:
1207: =item Socket
1208:
1209: open on the lonc client.
1210:
1211: =item Request
1212:
1213: data to send to the lond.
1214:
1215: =cut
1216:
1217: sub QueueTransaction {
1218:
1219: my $requestData = shift; # This is a LondTransaction.
1220: my $cmd = $requestData->getRequest();
1221:
1222: Debug(6,"QueueTransaction: ".$cmd);
1223:
1224: my $LondSocket = $IdleConnections->pop();
1225: if(!defined $LondSocket) { # Need to queue request.
1226: Debug(5,"Must queue...");
1227: $WorkQueue->enqueue($requestData);
1228: if($ConnectionCount < $MaxConnectionCount) {
1229: if($ConnectionRetriesLeft > 0) {
1230: Debug(5,"Starting additional lond connection");
1231: if(MakeLondConnection() == 0) {
1232: EmptyQueue(); # Fail transactions, can't make connection.
1233: CloseAllLondConnections; # Should all be closed but...
1234: }
1235: } else {
1236: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1237: EmptyQueue(); # It's worse than that ... he's dead Jim.
1238: CloseAllLondConnections; # Should all be closed but..
1239: }
1240: }
1241: } else { # Can start the request:
1242: Debug(8,"Can start...");
1243: StartRequest($LondSocket, $requestData);
1244: }
1245: }
1246:
1247: #-------------------------- Lonc UNIX socket handling ---------------------
1248:
1249: =pod
1250:
1251: =head2 ClientRequest
1252: Callback that is called when data can be read from the UNIX domain
1253: socket connecting us with an apache server process.
1254:
1255: =cut
1256:
1257: sub ClientRequest {
1258: Debug(6, "ClientRequest");
1259: my $event = shift;
1260: my $watcher = $event->w;
1261: my $socket = $watcher->fd;
1262: my $data = $watcher->data;
1263: my $thisread;
1264:
1265: Debug(9, " Watcher named: ".$watcher->desc);
1266:
1267: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1268: Debug(8, "rcv: data length = ".length($thisread)
1269: ." read =".$thisread);
1270: unless (defined $rv && length($thisread)) {
1271: # Likely eof on socket.
1272: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1273: close($socket);
1274: $watcher->cancel();
1275: delete($ActiveClients{$socket});
1276: return;
1277: }
1278: Debug(8,"Data: ".$data." this read: ".$thisread);
1279: $data = $data.$thisread; # Append new data.
1280: $watcher->data($data);
1281: if($data =~ /\n$/) { # Request entirely read.
1282: if($data eq "close_connection_exit\n") {
1283: Log("CRITICAL",
1284: "Request Close Connection ... exiting");
1285: CloseAllLondConnections();
1286: exit;
1287: }
1288: Debug(8, "Complete transaction received: ".$data);
1289: if($LogTransactions) {
1290: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1291: }
1292: my $Transaction = LondTransaction->new($data);
1293: $Transaction->SetClient($socket);
1294: QueueTransaction($Transaction);
1295: $watcher->cancel(); # Done looking for input data.
1296: }
1297:
1298: }
1299:
1300:
1301: =pod
1302:
1303: =head2 NewClient
1304:
1305: Callback that is called when a connection is received on the unix
1306: socket for a new client of lonc. The callback is parameterized by the
1307: event.. which is a-priori assumed to be an io event, and therefore has
1308: an fd member that is the Listener socket. We Accept the connection
1309: and register a new event on the readability of that socket:
1310:
1311: =cut
1312:
1313: sub NewClient {
1314: Debug(6, "NewClient");
1315: my $event = shift; # Get the event parameters.
1316: my $watcher = $event->w;
1317: my $socket = $watcher->fd; # Get the event' socket.
1318: my $connection = $socket->accept(); # Accept the client connection.
1319: Debug(8,"Connection request accepted from "
1320: .GetPeername($connection, AF_UNIX));
1321:
1322:
1323: my $description = sprintf("Connection to lonc client %d",
1324: $ClientConnection);
1325: Debug(9, "Creating event named: ".$description);
1326: Event->io(cb => \&ClientRequest,
1327: poll => 'r',
1328: desc => $description,
1329: data => "",
1330: fd => $connection);
1331: $ActiveClients{$connection} = $ClientConnection;
1332: $ClientConnection++;
1333: }
1334:
1335: =pod
1336:
1337: =head2 GetLoncSocketPath
1338:
1339: Returns the name of the UNIX socket on which to listen for client
1340: connections.
1341:
1342: =cut
1343:
1344: sub GetLoncSocketPath {
1345: return $UnixSocketDir."/".GetServerHost();
1346: }
1347:
1348: =pod
1349:
1350: =head2 GetServerHost
1351:
1352: Returns the host whose lond we talk with.
1353:
1354: =cut
1355:
1356: sub GetServerHost {
1357: return $RemoteHost; # Setup by the fork.
1358: }
1359:
1360: =pod
1361:
1362: =head2 GetServerPort
1363:
1364: Returns the lond port number.
1365:
1366: =cut
1367:
1368: sub GetServerPort {
1369: return $perlvar{londPort};
1370: }
1371:
1372: =pod
1373:
1374: =head2 SetupLoncListener
1375:
1376: Setup a lonc listener event. The event is called when the socket
1377: becomes readable.. that corresponds to the receipt of a new
1378: connection. The event handler established will accept the connection
1379: (creating a communcations channel), that int turn will establish
1380: another event handler to subess requests.
1381:
1382: =cut
1383:
1384: sub SetupLoncListener {
1385:
1386: my $socket;
1387: my $SocketName = GetLoncSocketPath();
1388: unlink($SocketName);
1389: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1390: Listen => 10,
1391: Type => SOCK_STREAM)) {
1392: die "Failed to create a lonc listner socket";
1393: }
1394: Event->io(cb => \&NewClient,
1395: poll => 'r',
1396: desc => 'Lonc listener Unix Socket',
1397: fd => $socket);
1398: }
1399:
1400: #
1401: # Toggle transaction logging.
1402: # Implicit inputs:
1403: # LogTransactions
1404: # Implicit Outputs:
1405: # LogTransactions
1406: sub ToggleTransactionLogging {
1407: print STDERR "Toggle transaction logging...\n";
1408: if(!$LogTransactions) {
1409: $LogTransactions = 1;
1410: } else {
1411: $LogTransactions = 0;
1412: }
1413:
1414:
1415: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1416: }
1417:
1418: =pod
1419:
1420: =head2 ChildStatus
1421:
1422: Child USR1 signal handler to report the most recent status
1423: into the status file.
1424:
1425: We also use this to reset the retries count in order to allow the
1426: client to retry connections with a previously dead server.
1427: =cut
1428:
1429: sub ChildStatus {
1430: my $event = shift;
1431: my $watcher = $event->w;
1432:
1433: Debug(2, "Reporting child status because : ".$watcher->data);
1434: my $docdir = $perlvar{'lonDocRoot'};
1435: my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
1436: print $fh $$."\t".$RemoteHost."\t".$Status."\t".
1437: $RecentLogEntry."\n";
1438: #
1439: # Write out information about each of the connections:
1440: #
1441: if ($DebugLevel > 2) {
1442: print $fh "Active connection statuses: \n";
1443: my $i = 1;
1444: print STDERR "================================= Socket Status Dump:\n";
1445: foreach my $item (keys %ActiveConnections) {
1446: my $Socket = $ActiveConnections{$item}->data;
1447: my $state = $Socket->GetState();
1448: print $fh "Connection $i State: $state\n";
1449: print STDERR "---------------------- Connection $i \n";
1450: $Socket->Dump(-1); # Ensure it gets dumped..
1451: $i++;
1452: }
1453: }
1454: $ConnectionRetriesLeft = $ConnectionRetries;
1455: }
1456:
1457: =pod
1458:
1459: =head2 SignalledToDeath
1460:
1461: Called in response to a signal that causes a chid process to die.
1462:
1463: =cut
1464:
1465:
1466: sub SignalledToDeath {
1467: my $event = shift;
1468: my $watcher= $event->w;
1469:
1470: Debug(2,"Signalled to death! via ".$watcher->data);
1471: my ($signal) = $watcher->data;
1472: chomp($signal);
1473: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1474: ."died through "."\"$signal\"");
1475: LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1476: ."\"$signal\"");
1477: exit 0;
1478:
1479: }
1480:
1481: =head2 ToggleDebug
1482:
1483: This sub toggles trace debugging on and off.
1484:
1485: =cut
1486:
1487: sub ToggleDebug {
1488: my $Current = $DebugLevel;
1489: $DebugLevel = $NextDebugLevel;
1490: $NextDebugLevel = $Current;
1491:
1492: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1493:
1494: }
1495:
1496: =head2 ChildProcess
1497:
1498: This sub implements a child process for a single lonc daemon.
1499:
1500: =cut
1501:
1502: sub ChildProcess {
1503:
1504:
1505: #
1506: # Signals must be handled by the Event framework...
1507: #
1508:
1509: Event->signal(signal => "QUIT",
1510: cb => \&SignalledToDeath,
1511: data => "QUIT");
1512: Event->signal(signal => "HUP",
1513: cb => \&ChildStatus,
1514: data => "HUP");
1515: Event->signal(signal => "USR1",
1516: cb => \&ChildStatus,
1517: data => "USR1");
1518: Event->signal(signal => "USR2",
1519: cb => \&ToggleTransactionLogging);
1520: Event->signal(signal => "INT",
1521: cb => \&ToggleDebug,
1522: data => "INT");
1523:
1524: SetupTimer();
1525:
1526: SetupLoncListener();
1527:
1528: $Event::Debuglevel = $DebugLevel;
1529:
1530: Debug(9, "Making initial lond connection for ".$RemoteHost);
1531:
1532: # Setup the initial server connection:
1533:
1534: # &MakeLondConnection(); // let first work requirest do it.
1535:
1536:
1537: Debug(9,"Entering event loop");
1538: my $ret = Event::loop(); # Start the main event loop.
1539:
1540:
1541: die "Main event loop exited!!!";
1542: }
1543:
1544: # Create a new child for host passed in:
1545:
1546: sub CreateChild {
1547: my $sigset = POSIX::SigSet->new(SIGINT);
1548: sigprocmask(SIG_BLOCK, $sigset);
1549: my $host = shift;
1550: $RemoteHost = $host;
1551: Log("CRITICAL", "Forking server for ".$host);
1552: my $pid = fork;
1553: if($pid) { # Parent
1554: $RemoteHost = "Parent";
1555: $ChildHash{$pid} = $host;
1556: $HostToPid{$host}= $pid;
1557: sigprocmask(SIG_UNBLOCK, $sigset);
1558:
1559: } else { # child.
1560: ShowStatus("Connected to ".$RemoteHost);
1561: $SIG{INT} = 'DEFAULT';
1562: sigprocmask(SIG_UNBLOCK, $sigset);
1563: ChildProcess; # Does not return.
1564: }
1565:
1566: }
1567: #
1568: # Parent process logic pass 1:
1569: # For each entry in the hosts table, we will
1570: # fork off an instance of ChildProcess to service the transactions
1571: # to that host. Each pid will be entered in a global hash
1572: # with the value of the key, the host.
1573: # The parent will then enter a loop to wait for process exits.
1574: # Each exit gets logged and the child gets restarted.
1575: #
1576:
1577: #
1578: # Fork and start in new session so hang-up isn't going to
1579: # happen without intent.
1580: #
1581:
1582:
1583:
1584:
1585:
1586:
1587: ShowStatus("Forming new session");
1588: my $childpid = fork;
1589: if ($childpid != 0) {
1590: sleep 4; # Give child a chacne to break to
1591: exit 0; # a new sesion.
1592: }
1593: #
1594: # Write my pid into the pid file so I can be located
1595: #
1596:
1597: ShowStatus("Parent writing pid file:");
1598: my $execdir = $perlvar{'lonDaemons'};
1599: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1600: print PIDSAVE "$$\n";
1601: close(PIDSAVE);
1602:
1603:
1604:
1605: if (POSIX::setsid() < 0) {
1606: print "Could not create new session\n";
1607: exit -1;
1608: }
1609:
1610: ShowStatus("Forking node servers");
1611:
1612: Log("CRITICAL", "--------------- Starting children ---------------");
1613:
1614: LondConnection::ReadConfig; # Read standard config files.
1615: my $HostIterator = LondConnection::GetHostIterator;
1616: while (! $HostIterator->end()) {
1617:
1618: my $hostentryref = $HostIterator->get();
1619: CreateChild($hostentryref->[0]);
1620: $HostHash{$hostentryref->[0]} = $hostentryref->[4];
1621: $HostIterator->next();
1622: }
1623: $RemoteHost = "Parent Server";
1624:
1625: # Maintain the population:
1626:
1627: ShowStatus("Parent keeping the flock");
1628:
1629: #
1630: # Set up parent signals:
1631: #
1632:
1633: $SIG{INT} = \&Terminate;
1634: $SIG{TERM} = \&Terminate;
1635: $SIG{HUP} = \&Restart;
1636: $SIG{USR1} = \&CheckKids;
1637: $SIG{USR2} = \&UpdateKids; # LonManage update request.
1638:
1639: while(1) {
1640: my $deadchild = wait();
1641: if(exists $ChildHash{$deadchild}) { # need to restart.
1642: my $deadhost = $ChildHash{$deadchild};
1643: delete($HostToPid{$deadhost});
1644: delete($ChildHash{$deadchild});
1645: Log("WARNING","Lost child pid= ".$deadchild.
1646: "Connected to host ".$deadhost);
1647: Log("INFO", "Restarting child procesing ".$deadhost);
1648: CreateChild($deadhost);
1649: }
1650: }
1651:
1652:
1653:
1654: =pod
1655:
1656: =head1 CheckKids
1657:
1658: Since kids do not die as easily in this implementation
1659: as the previous one, there is no need to restart the
1660: dead ones (all dead kids get restarted when they die!!)
1661: The only thing this function does is to pass USR1 to the
1662: kids so that they report their status.
1663:
1664: =cut
1665:
1666: sub CheckKids {
1667: Debug(2, "Checking status of children");
1668: my $docdir = $perlvar{'lonDocRoot'};
1669: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
1670: my $now=time;
1671: my $local=localtime($now);
1672: print $fh "LONC status $local - parent $$ \n\n";
1673: foreach my $pid (keys %ChildHash) {
1674: Debug(2, "Sending USR1 -> $pid");
1675: kill 'USR1' => $pid; # Tell Child to report status.
1676: sleep 1; # Wait so file doesn't intermix.
1677: }
1678: }
1679:
1680: =pod
1681:
1682: =head1 UpdateKids
1683:
1684: parent's SIGUSR2 handler. This handler:
1685:
1686: =item
1687:
1688: Rereads the hosts file.
1689:
1690: =item
1691:
1692: Kills off (via sigint) children for hosts that have disappeared.
1693:
1694: =item
1695:
1696: QUITs children for hosts that already exist (this just forces a status display
1697: and resets the connection retry count for that host.
1698:
1699: =item
1700:
1701: Starts new children for hosts that have been added to the hosts.tab file since
1702: the start of the master program and maintains them.
1703:
1704: =cut
1705:
1706: sub UpdateKids {
1707:
1708: Log("INFO", "Updating connections via SIGUSR2");
1709:
1710: # Just in case we need to kill our own lonc, we wait a few seconds to
1711: # give it a chance to receive and relay lond's response to the
1712: # re-init command.
1713: #
1714:
1715: sleep(2); # Wait a couple of seconds.
1716:
1717: my %hosts; # Indexed by loncapa hostname, value=ip.
1718:
1719: # Need to re-read the host table:
1720:
1721:
1722: LondConnection::ReadConfig();
1723: my $I = LondConnection::GetHostIterator;
1724: while (! $I->end()) {
1725: my $item = $I->get();
1726: $hosts{$item->[0]} = $item->[4];
1727: $I->next();
1728: }
1729:
1730: # The logic below is written for clarity not for efficiency.
1731: # Since I anticipate that this function is only rarely called, that's
1732: # appropriate. There are certainly ways to combine the loops below,
1733: # and anyone wishing to obscure the logic is welcome to go for it.
1734: # Note that we don't re-direct sigchild. Instead we do what's needed
1735: # to the data structures that keep track of children to ensure that
1736: # when sigchild is honored, no new child is born.
1737: #
1738:
1739: # For each existing child; if it's host doesn't exist, kill the child.
1740:
1741: foreach my $child (keys %ChildHash) {
1742: my $oldhost = $ChildHash{$child};
1743: if (!(exists $hosts{$oldhost})) {
1744: Log("CRITICAL", "Killing child for $oldhost host no longer exists");
1745: delete $ChildHash{$child};
1746: delete $HostToPid{$oldhost};
1747: kill 'QUIT' => $child;
1748: }
1749: }
1750: # For each remaining existing child; if it's host's ip has changed,
1751: # Restart the child on the new IP.
1752:
1753: foreach my $child (keys %ChildHash) {
1754: my $oldhost = $ChildHash{$child};
1755: my $oldip = $HostHash{$oldhost};
1756: if ($hosts{$oldhost} ne $oldip) {
1757:
1758: # kill the old child.
1759:
1760: Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
1761: delete $ChildHash{$child};
1762: delete $HostToPid{$oldhost};
1763: kill 'QUIT' => $child;
1764:
1765: # Do the book-keeping needed to start a new child on the
1766: # new ip.
1767:
1768: $HostHash{$oldhost} = $hosts{$oldhost};
1769: CreateChild($oldhost);
1770: }
1771: }
1772: # Finally, for each new host, not in the host hash, create a
1773: # enter the host and create a new child.
1774: # Force a status display of any existing process.
1775:
1776: foreach my $host (keys %hosts) {
1777: if(!(exists $HostHash{$host})) {
1778: Log("INFO", "New host $host discovered in hosts.tab...");
1779: $HostHash{$host} = $hosts{$host};
1780: CreateChild($host);
1781: } else {
1782: kill 'HUP' => $HostToPid{$host}; # status display.
1783: }
1784: }
1785: }
1786:
1787:
1788: =pod
1789:
1790: =head1 Restart
1791:
1792: Signal handler for HUP... all children are killed and
1793: we self restart. This is an el-cheapo way to re read
1794: the config file.
1795:
1796: =cut
1797:
1798: sub Restart {
1799: &KillThemAll; # First kill all the children.
1800: Log("CRITICAL", "Restarting");
1801: my $execdir = $perlvar{'lonDaemons'};
1802: unlink("$execdir/logs/lonc.pid");
1803: exec("$execdir/loncnew");
1804: }
1805:
1806: =pod
1807:
1808: =head1 KillThemAll
1809:
1810: Signal handler that kills all children by sending them a
1811: SIGHUP. Responds to sigint and sigterm.
1812:
1813: =cut
1814:
1815: sub KillThemAll {
1816: Debug(2, "Kill them all!!");
1817: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
1818: foreach my $pid (keys %ChildHash) {
1819: my $serving = $ChildHash{$pid};
1820: Debug(2, "Killing lonc for $serving pid = $pid");
1821: ShowStatus("Killing lonc for $serving pid = $pid");
1822: Log("CRITICAL", "Killing lonc for $serving pid = $pid");
1823: kill 'QUIT' => $pid;
1824: delete($ChildHash{$pid});
1825: }
1826: my $execdir = $perlvar{'lonDaemons'};
1827: unlink("$execdir/logs/lonc.pid");
1828:
1829: }
1830:
1831: =pod
1832:
1833: =head1 Terminate
1834:
1835: Terminate the system.
1836:
1837: =cut
1838:
1839: sub Terminate {
1840: KillThemAll;
1841: Log("CRITICAL","Master process exiting");
1842: exit 0;
1843:
1844: }
1845: =pod
1846:
1847: =head1 Theory
1848:
1849: The event class is used to build this as a single process with an
1850: event driven model. The following events are handled:
1851:
1852: =item UNIX Socket connection Received
1853:
1854: =item Request data arrives on UNIX data transfer socket.
1855:
1856: =item lond connection becomes writable.
1857:
1858: =item timer fires at 1 second intervals.
1859:
1860: All sockets are run in non-blocking mode. Timeouts managed by the timer
1861: handler prevents hung connections.
1862:
1863: Key data structures:
1864:
1865: =item RequestQueue
1866:
1867: A queue of requests received from UNIX sockets that are
1868: waiting for a chance to be forwarded on a lond connection socket.
1869:
1870: =item ActiveConnections
1871:
1872: A hash of lond connections that have transactions in process that are
1873: available to be timed out.
1874:
1875: =item ActiveTransactions
1876:
1877: A hash indexed by lond connections that contain the client reply
1878: socket for each connection that has an active transaction on it.
1879:
1880: =item IdleConnections
1881:
1882: A hash of lond connections that have no work to do. These connections
1883: can be closed if they are idle for a long enough time.
1884:
1885: =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