1: #!/usr/bin/perl
2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
5: # $Id: loncnew,v 1.41 2004/02/09 13:39:28 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= 3600; # Wait an hour 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:
107: #
108: # The hash below gives the HTML format for log messages
109: # given a severity.
110: #
111: my %LogFormats;
112:
113: $LogFormats{"CRITICAL"} = "<font color=red>CRITICAL: %s</font>";
114: $LogFormats{"SUCCESS"} = "<font color=green>SUCCESS: %s</font>";
115: $LogFormats{"INFO"} = "<font color=yellow>INFO: %s</font>";
116: $LogFormats{"WARNING"} = "<font color=blue>WARNING: %s</font>";
117: $LogFormats{"DEFAULT"} = " %s ";
118:
119:
120:
121: =pod
122:
123: =head2 LogPerm
124:
125: Makes an entry into the permanent log file.
126:
127: =cut
128: sub LogPerm {
129: my $message=shift;
130: my $execdir=$perlvar{'lonDaemons'};
131: my $now=time;
132: my $local=localtime($now);
133: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
134: print $fh "$now:$message:$local\n";
135: }
136:
137: =pod
138:
139: =head2 Log
140:
141: Logs a message to the log file.
142: Parameters:
143:
144: =item severity
145:
146: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
147: format string used to format the message. if the severity is
148: not a defined severity the Default format string is used.
149:
150: =item message
151:
152: The base message. In addtion to the format string, the message
153: will be appended to a string containing the name of our remote
154: host and the time will be formatted into the message.
155:
156: =cut
157:
158: sub Log {
159: my $severity = shift;
160: my $message = shift;
161:
162: if(!$LogFormats{$severity}) {
163: $severity = "DEFAULT";
164: }
165:
166: my $format = $LogFormats{$severity};
167:
168: # Put the window dressing in in front of the message format:
169:
170: my $now = time;
171: my $local = localtime($now);
172: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
173: my $finalformat = $finalformat.$format."\n";
174:
175: # open the file and put the result.
176:
177: my $execdir = $perlvar{'lonDaemons'};
178: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
179: my $msg = sprintf($finalformat, $message);
180: $RecentLogEntry = $msg;
181: print $fh $msg;
182:
183:
184: }
185:
186:
187: =pod
188:
189: =head2 GetPeerName
190:
191: Returns the name of the host that a socket object is connected to.
192:
193: =cut
194:
195: sub GetPeername {
196: my $connection = shift;
197: my $AdrFamily = shift;
198: my $peer = $connection->peername();
199: my $peerport;
200: my $peerip;
201: if($AdrFamily == AF_INET) {
202: ($peerport, $peerip) = sockaddr_in($peer);
203: my $peername = gethostbyaddr($peerip, $AdrFamily);
204: return $peername;
205: } elsif ($AdrFamily == AF_UNIX) {
206: my $peerfile;
207: ($peerfile) = sockaddr_un($peer);
208: return $peerfile;
209: }
210: }
211: =pod
212:
213: =head2 Debug
214:
215: Invoked to issue a debug message.
216:
217: =cut
218:
219: sub Debug {
220: my $level = shift;
221: my $message = shift;
222: if ($level <= $DebugLevel) {
223: Log("INFO", "-Debug- $message host = $RemoteHost");
224: }
225: }
226:
227: sub SocketDump {
228: my $level = shift;
229: my $socket= shift;
230: if($level <= $DebugLevel) {
231: $socket->Dump();
232: }
233: }
234:
235: =pod
236:
237: =head2 ShowStatus
238:
239: Place some text as our pid status.
240: and as what we return in a SIGUSR1
241:
242: =cut
243: sub ShowStatus {
244: my $state = shift;
245: my $now = time;
246: my $local = localtime($now);
247: $Status = $local.": ".$state;
248: $0='lonc: '.$state.' '.$local;
249: }
250:
251: =pod
252:
253: =head 2 SocketTimeout
254:
255: Called when an action on the socket times out. The socket is
256: destroyed and any active transaction is failed.
257:
258:
259: =cut
260: sub SocketTimeout {
261: my $Socket = shift;
262: Log("WARNING", "A socket timeout was detected");
263: Debug(0, " SocketTimeout called: ");
264: $Socket->Dump();
265: KillSocket($Socket); # A transaction timeout also counts as
266: # a connection failure:
267: $ConnectionRetriesLeft--;
268: }
269: #----------------------------- Timer management ------------------------
270:
271: =pod
272:
273: =head2 Tick
274:
275: Invoked each timer tick.
276:
277: =cut
278:
279:
280: sub Tick {
281: my $client;
282: if($ConnectionRetriesLeft > 0) {
283: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
284: ." Retries remaining: ".$ConnectionRetriesLeft);
285: } else {
286: ShowStatus(GetServerHost()." >> DEAD <<");
287: }
288: # Is it time to prune connection count:
289:
290:
291: if($IdleConnections->Count() &&
292: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
293: $IdleSeconds++;
294: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
295: my $Socket = $IdleConnections->pop();
296: KillSocket($Socket);
297: }
298: } else {
299: $IdleSeconds = 0; # Reset idle count if not idle.
300: }
301: #
302: # For each inflight transaction, tick down its timeout counter.
303: #
304:
305: foreach my $item (keys %ActiveConnections) {
306: my $State = $ActiveConnections{$item}->data->GetState();
307: if ($State ne 'Idle') {
308: Debug(5,"Ticking Socket $State $item");
309: $ActiveConnections{$item}->data->Tick();
310: }
311: }
312: # Do we have work in the queue, but no connections to service them?
313: # If so, try to make some new connections to get things going again.
314: #
315:
316: my $Requests = $WorkQueue->Count();
317: if (($ConnectionCount == 0) && ($Requests > 0)) {
318: if ($ConnectionRetriesLeft > 0) {
319: my $Connections = ($Requests <= $MaxConnectionCount) ?
320: $Requests : $MaxConnectionCount;
321: Debug(5,"Work but no connections, start ".$Connections." of them");
322: my $successCount = 0;
323: for (my $i =0; $i < $Connections; $i++) {
324: $successCount += MakeLondConnection();
325: }
326: if($successCount == 0) { # All connections failed:
327: Debug(5,"Work in queue failed to make any connectiouns\n");
328: EmptyQueue(); # Fail pending transactions with con_lost.
329: }
330: } else {
331: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
332: Debug(5,"Work in queue, but gave up on connections..flushing\n");
333: EmptyQueue(); # Connections can't be established.
334: }
335:
336: }
337: }
338:
339: =pod
340:
341: =head2 SetupTimer
342:
343: Sets up a 1 per sec recurring timer event. The event handler is used to:
344:
345: =item
346:
347: Trigger timeouts on communications along active sockets.
348:
349: =item
350:
351: Trigger disconnections of idle sockets.
352:
353: =cut
354:
355: sub SetupTimer {
356: Debug(6, "SetupTimer");
357: Event->timer(interval => 1, cb => \&Tick );
358: }
359:
360: =pod
361:
362: =head2 ServerToIdle
363:
364: This function is called when a connection to the server is
365: ready for more work.
366:
367: If there is work in the Work queue the top element is dequeued
368: and the connection will start to work on it. If the work queue is
369: empty, the connection is pushed on the idle connection stack where
370: it will either get another work unit, or alternatively, if it sits there
371: long enough, it will be shut down and released.
372:
373: =cut
374:
375: sub ServerToIdle {
376: my $Socket = shift; # Get the socket.
377: delete($ActiveTransactions{$Socket}); # Server has no transaction
378:
379: &Debug(5, "Server to idle");
380:
381: # If there's work to do, start the transaction:
382:
383: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
384: if ($reqdata ne undef) {
385: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
386: &StartRequest($Socket, $reqdata);
387:
388: } else {
389:
390: # There's no work waiting, so push the server to idle list.
391: &Debug(5, "No new work requests, server connection going idle");
392: $IdleConnections->push($Socket);
393: }
394: }
395:
396: =pod
397:
398: =head2 ClientWritable
399:
400: Event callback for when a client socket is writable.
401:
402: This callback is established when a transaction reponse is
403: avaiable from lond. The response is forwarded to the unix socket
404: as it becomes writable in this sub.
405:
406: Parameters:
407:
408: =item Event
409:
410: The event that has been triggered. Event->w->data is
411: the data and Event->w->fd is the socket to write.
412:
413: =cut
414:
415: sub ClientWritable {
416: my $Event = shift;
417: my $Watcher = $Event->w;
418: my $Data = $Watcher->data;
419: my $Socket = $Watcher->fd;
420:
421: # Try to send the data:
422:
423: &Debug(6, "ClientWritable writing".$Data);
424: &Debug(9, "Socket is: ".$Socket);
425:
426: if($Socket->connected) {
427: my $result = $Socket->send($Data, 0);
428:
429: # $result undefined: the write failed.
430: # otherwise $result is the number of bytes written.
431: # Remove that preceding string from the data.
432: # If the resulting data is empty, destroy the watcher
433: # and set up a read event handler to accept the next
434: # request.
435:
436: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
437: if($result ne undef) {
438: &Debug(9, "send result was defined");
439: if($result == length($Data)) { # Entire string sent.
440: &Debug(9, "ClientWritable data all written");
441: $Watcher->cancel();
442: #
443: # Set up to read next request from socket:
444:
445: my $descr = sprintf("Connection to lonc client %d",
446: $ActiveClients{$Socket});
447: Event->io(cb => \&ClientRequest,
448: poll => 'r',
449: desc => $descr,
450: data => "",
451: fd => $Socket);
452:
453: } else { # Partial string sent.
454: $Watcher->data(substr($Data, $result));
455: if($result == 0) { # client hung up on us!!
456: Log("INFO", "lonc pipe client hung up on us!");
457: $Watcher->cancel;
458: $Socket->shutdown(2);
459: $Socket->close();
460: }
461: }
462:
463: } else { # Error of some sort...
464:
465: # Some errnos are possible:
466: my $errno = $!;
467: if($errno == POSIX::EWOULDBLOCK ||
468: $errno == POSIX::EAGAIN ||
469: $errno == POSIX::EINTR) {
470: # No action taken?
471: } else { # Unanticipated errno.
472: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
473: $Watcher->cancel; # Stop the watcher.
474: $Socket->shutdown(2); # Kill connection
475: $Socket->close(); # Close the socket.
476: }
477:
478: }
479: } else {
480: $Watcher->cancel(); # A delayed request...just cancel.
481: }
482: }
483:
484: =pod
485:
486: =head2 CompleteTransaction
487:
488: Called when the reply data has been received for a lond
489: transaction. The reply data must now be sent to the
490: ultimate client on the other end of the Unix socket. This is
491: done by setting up a writable event for the socket with the
492: data the reply data.
493:
494: Parameters:
495:
496: =item Socket
497:
498: Socket on which the lond transaction occured. This is a
499: LondConnection. The data received is in the TransactionReply member.
500:
501: =item Transaction
502:
503: The transaction that is being completed.
504:
505: =cut
506:
507: sub CompleteTransaction {
508: &Debug(5,"Complete transaction");
509: my $Socket = shift;
510: my $Transaction = shift;
511:
512: if (!$Transaction->isDeferred()) { # Normal transaction
513: my $data = $Socket->GetReply(); # Data to send.
514: if($LogTransactions) {
515: Log("SUCCESS", "Reply from lond: '$data'");
516: }
517: StartClientReply($Transaction, $data);
518: } else { # Delete deferred transaction file.
519: Log("SUCCESS", "A delayed transaction was completed");
520: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
521: unlink $Transaction->getFile();
522: }
523: }
524: =pod
525: =head1 StartClientReply
526:
527: Initiates a reply to a client where the reply data is a parameter.
528:
529: =head2 parameters:
530:
531: =item Transaction
532:
533: The transaction for which we are responding to the client.
534:
535: =item data
536:
537: The data to send to apached client.
538:
539: =cut
540: sub StartClientReply {
541: my $Transaction = shift;
542: my $data = shift;
543:
544:
545: my $Client = $Transaction->getClient();
546:
547: &Debug(8," Reply was: ".$data);
548: my $Serial = $ActiveClients{$Client};
549: my $desc = sprintf("Connection to lonc client %d",
550: $Serial);
551: Event->io(fd => $Client,
552: poll => "w",
553: desc => $desc,
554: cb => \&ClientWritable,
555: data => $data);
556: }
557: =pod
558: =head2 FailTransaction
559:
560: Finishes a transaction with failure because the associated lond socket
561: disconnected. There are two possibilities:
562: - The transaction is deferred: in which case we just quietly
563: delete the transaction since there is no client connection.
564: - The transaction is 'live' in which case we initiate the sending
565: of "con_lost" to the client.
566:
567: Deleting the transaction means killing it from the
568: %ActiveTransactions hash.
569:
570: Parameters:
571:
572: =item client
573:
574: The LondTransaction we are failing.
575:
576: =cut
577:
578: sub FailTransaction {
579: my $transaction = shift;
580: Log("WARNING", "Failing transaction ".$transaction->getRequest());
581: Debug(1, "Failing transaction: ".$transaction->getRequest());
582: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
583: my $client = $transaction->getClient();
584: Debug(1," Replying con_lost to ".$transaction->getRequest());
585: StartClientReply($transaction, "con_lost\n");
586: }
587: if($ConnectionRetriesLeft <= 0) {
588: Log("CRITICAL", "Host marked dead: ".GetServerHost());
589: }
590:
591: }
592:
593: =pod
594: =head1 EmptyQueue
595:
596: Fails all items in the work queue with con_lost.
597: Note that each item in the work queue is a transaction.
598:
599: =cut
600: sub EmptyQueue {
601: $ConnectionRetriesLeft--; # Counts as connection failure too.
602: while($WorkQueue->Count()) {
603: my $request = $WorkQueue->dequeue(); # This is a transaction
604: FailTransaction($request);
605: }
606: }
607:
608: =pod
609:
610: =head2 CloseAllLondConnections
611:
612: Close all connections open on lond prior to exit e.g.
613:
614: =cut
615: sub CloseAllLondConnections {
616: foreach my $Socket (keys %ActiveConnections) {
617: KillSocket($Socket);
618: }
619: }
620: =cut
621:
622: =pod
623:
624: =head2 KillSocket
625:
626: Destroys a socket. This function can be called either when a socket
627: has died of 'natural' causes or because a socket needs to be pruned due to
628: idleness. If the socket has died naturally, if there are no longer any
629: live connections a new connection is created (in case there are transactions
630: in the queue). If the socket has been pruned, it is never re-created.
631:
632: Parameters:
633:
634: =item Socket
635:
636: The socket to kill off.
637:
638: =item Restart
639:
640: nonzero if we are allowed to create a new connection.
641:
642:
643: =cut
644: sub KillSocket {
645: my $Socket = shift;
646:
647: Log("WARNING", "Shutting down a socket");
648: $Socket->Shutdown();
649:
650: # If the socket came from the active connection set,
651: # delete its transaction... note that FailTransaction should
652: # already have been called!!!
653: # otherwise it came from the idle set.
654: #
655:
656: if(exists($ActiveTransactions{$Socket})) {
657: delete ($ActiveTransactions{$Socket});
658: }
659: if(exists($ActiveConnections{$Socket})) {
660: delete($ActiveConnections{$Socket});
661: $ConnectionCount--;
662: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
663: }
664: # If the connection count has gone to zero and there is work in the
665: # work queue, the work all gets failed with con_lost.
666: #
667: if($ConnectionCount == 0) {
668: EmptyQueue();
669: }
670: }
671:
672: =pod
673:
674: =head2 LondReadable
675:
676: This function is called whenever a lond connection
677: is readable. The action is state dependent:
678:
679: =head3 State=Initialized
680:
681: We''re waiting for the challenge, this is a no-op until the
682: state changes.
683:
684: =head3 State=Challenged
685:
686: The challenge has arrived we need to transition to Writable.
687: The connection must echo the challenge back.
688:
689: =head3 State=ChallengeReplied
690:
691: The challenge has been replied to. The we are receiveing the
692: 'ok' from the partner.
693:
694: =head3 State=ReadingVersionString
695:
696: We have requested the lond version and are reading the
697: version back. Upon completion, we'll store the version away
698: for future use(?).
699:
700: =head3 State=HostSet
701:
702: We have selected the domain name of our peer (multhomed hosts)
703: and are getting the reply (presumably ok) back.
704:
705: =head3 State=RequestingKey
706:
707: The ok has been received and we need to send the request for
708: an encryption key. Transition to writable for that.
709:
710: =head3 State=ReceivingKey
711:
712: The the key has been requested, now we are reading the new key.
713:
714: =head3 State=Idle
715:
716: The encryption key has been negotiated or we have finished
717: reading data from the a transaction. If the callback data has
718: a client as well as the socket iformation, then we are
719: doing a transaction and the data received is relayed to the client
720: before the socket is put on the idle list.
721:
722: =head3 State=SendingRequest
723:
724: I do not think this state can be received here, but if it is,
725: the appropriate thing to do is to transition to writable, and send
726: the request.
727:
728: =head3 State=ReceivingReply
729:
730: We finished sending the request to the server and now transition
731: to readable to receive the reply.
732:
733: The parameter to this function are:
734:
735: The event. Implicit in this is the watcher and its data. The data
736: contains at least the lond connection object and, if a
737: transaction is in progress, the socket attached to the local client.
738:
739: =cut
740:
741: sub LondReadable {
742:
743: my $Event = shift;
744: my $Watcher = $Event->w;
745: my $Socket = $Watcher->data;
746: my $client = undef;
747:
748: &Debug(6,"LondReadable called state = ".$Socket->GetState());
749:
750:
751: my $State = $Socket->GetState(); # All action depends on the state.
752:
753: SocketDump(6, $Socket);
754: my $status = $Socket->Readable();
755:
756: &Debug(2, "Socket->Readable returned: $status");
757:
758: if($status != 0) {
759: # bad return from socket read. Currently this means that
760: # The socket has become disconnected. We fail the transaction.
761:
762: Log("WARNING",
763: "Lond connection lost.");
764: if(exists($ActiveTransactions{$Socket})) {
765: FailTransaction($ActiveTransactions{$Socket});
766: }
767: $Watcher->cancel();
768: KillSocket($Socket);
769: $ConnectionRetriesLeft--; # Counts as connection failure
770: return;
771: }
772: SocketDump(6,$Socket);
773:
774: $State = $Socket->GetState(); # Update in case of transition.
775: &Debug(6, "After read, state is ".$State);
776:
777: if($State eq "Initialized") {
778:
779:
780: } elsif ($State eq "ChallengeReceived") {
781: # The challenge must be echoed back; The state machine
782: # in the connection takes care of setting that up. Just
783: # need to transition to writable:
784:
785: $Watcher->cb(\&LondWritable);
786: $Watcher->poll("w");
787:
788: } elsif ($State eq "ChallengeReplied") {
789:
790: } elsif ($State eq "RequestingVersion") {
791: # Need to ask for the version... that is writiability:
792:
793: $Watcher->cb(\&LondWritable);
794: $Watcher->poll("w");
795:
796: } elsif ($State eq "ReadingVersionString") {
797: # Read the rest of the version string...
798: } elsif ($State eq "SetHost") {
799: # Need to request the actual domain get set...
800:
801: $Watcher->cb(\&LondWritable);
802: $Watcher->poll("w");
803: } elsif ($State eq "HostSet") {
804: # Reading the 'ok' from the peer.
805:
806: } elsif ($State eq "RequestingKey") {
807: # The ok was received. Now we need to request the key
808: # That requires us to be writable:
809:
810: $Watcher->cb(\&LondWritable);
811: $Watcher->poll("w");
812:
813: } elsif ($State eq "ReceivingKey") {
814:
815: } elsif ($State eq "Idle") {
816:
817: # This is as good a spot as any to get the peer version
818: # string:
819:
820: if($LondVersion eq "unknown") {
821: $LondVersion = $Socket->PeerVersion();
822: Log("INFO", "Connected to lond version: $LondVersion");
823: }
824: # If necessary, complete a transaction and then go into the
825: # idle queue.
826: # Note that a trasition to idle indicates a live lond
827: # on the other end so reset the connection retries.
828: #
829: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
830: $Watcher->cancel();
831: if(exists($ActiveTransactions{$Socket})) {
832: Debug(5,"Completing transaction!!");
833: CompleteTransaction($Socket,
834: $ActiveTransactions{$Socket});
835: } else {
836: Log("SUCCESS", "Connection ".$ConnectionCount." to "
837: .$RemoteHost." now ready for action");
838: }
839: ServerToIdle($Socket); # Next work unit or idle.
840:
841: } elsif ($State eq "SendingRequest") {
842: # We need to be writable for this and probably don't belong
843: # here inthe first place.
844:
845: Deubg(6, "SendingRequest state encountered in readable");
846: $Watcher->poll("w");
847: $Watcher->cb(\&LondWritable);
848:
849: } elsif ($State eq "ReceivingReply") {
850:
851:
852: } else {
853: # Invalid state.
854: Debug(4, "Invalid state in LondReadable");
855: }
856: }
857:
858: =pod
859:
860: =head2 LondWritable
861:
862: This function is called whenever a lond connection
863: becomes writable while there is a writeable monitoring
864: event. The action taken is very state dependent:
865:
866: =head3 State = Connected
867:
868: The connection is in the process of sending the 'init' hailing to the
869: lond on the remote end. The connection object''s Writable member is
870: called. On error, ConnectionError is called to destroy the connection
871: and remove it from the ActiveConnections hash
872:
873: =head3 Initialized
874:
875: 'init' has been sent, writability monitoring is removed and
876: readability monitoring is started with LondReadable as the callback.
877:
878: =head3 ChallengeReceived
879:
880: The connection has received the who are you challenge from the remote
881: system, and is in the process of sending the challenge
882: response. Writable is called.
883:
884: =head3 ChallengeReplied
885:
886: The connection has replied to the initial challenge The we switch to
887: monitoring readability looking for the server to reply with 'ok'.
888:
889: =head3 RequestingKey
890:
891: The connection is in the process of requesting its encryption key.
892: Writable is called.
893:
894: =head3 ReceivingKey
895:
896: The connection has sent the request for a key. Switch to readability
897: monitoring to accept the key
898:
899: =head3 SendingRequest
900:
901: The connection is in the process of sending a request to the server.
902: This request is part of a client transaction. All the states until
903: now represent the client setup protocol. Writable is called.
904:
905: =head3 ReceivingReply
906:
907: The connection has sent a request. Now it must receive a reply.
908: Readability monitoring is requested.
909:
910: This function is an event handler and therefore receives as
911: a parameter the event that has fired. The data for the watcher
912: of this event is a reference to a list of one or two elements,
913: depending on state. The first (and possibly only) element is the
914: socket. The second (present only if a request is in progress)
915: is the socket on which to return a reply to the caller.
916:
917: =cut
918:
919: sub LondWritable {
920: my $Event = shift;
921: my $Watcher = $Event->w;
922: my $Socket = $Watcher->data;
923: my $State = $Socket->GetState();
924:
925: Debug(6,"LondWritable State = ".$State."\n");
926:
927:
928: # Figure out what to do depending on the state of the socket:
929:
930:
931:
932:
933: SocketDump(6,$Socket);
934:
935: if ($State eq "Connected") {
936:
937: if ($Socket->Writable() != 0) {
938: # The write resulted in an error.
939: # We'll treat this as if the socket got disconnected:
940: Log("WARNING", "Connection to ".$RemoteHost.
941: " has been disconnected");
942: FailTransaction($ActiveTransactions{$Socket});
943: $Watcher->cancel();
944: KillSocket($Socket);
945: return;
946: }
947:
948: # "init" is being sent...
949:
950: } elsif ($State eq "Initialized") {
951:
952: # Now that init was sent, we switch
953: # to watching for readability:
954:
955: $Watcher->cb(\&LondReadable);
956: $Watcher->poll("r");
957:
958: } elsif ($State eq "ChallengeReceived") {
959: # We received the challenge, now we
960: # are echoing it back. This is a no-op,
961: # we're waiting for the state to change
962:
963: if($Socket->Writable() != 0) {
964:
965: $Watcher->cancel();
966: KillSocket($Socket);
967: return;
968: }
969:
970: } elsif ($State eq "ChallengeReplied") {
971: # The echo was sent back, so we switch
972: # to watching readability.
973:
974: $Watcher->cb(\&LondReadable);
975: $Watcher->poll("r");
976: } elsif ($State eq "RequestingVersion") {
977: # Sending the peer a version request...
978:
979: if($Socket->Writable() != 0) {
980: $Watcher->cancel();
981: KillSocket($Socket);
982: return;
983: }
984: } elsif ($State eq "ReadingVersionString") {
985: # Transition to read since we have sent the
986: # version command and now just need to read the
987: # version string from the peer:
988:
989: $Watcher->cb(\&LondReadable);
990: $Watcher->poll("r");
991:
992: } elsif ($State eq "SetHost") {
993: # Setting the remote domain...
994:
995: if($Socket->Writable() != 0) {
996: $Watcher->cancel();
997: KillSocket($Socket);
998: return;
999: }
1000: } elsif ($State eq "HostSet") {
1001: # Back to readable to get the ok.
1002:
1003: $Watcher->cb(\&LondReadable);
1004: $Watcher->poll("r");
1005:
1006:
1007: } elsif ($State eq "RequestingKey") {
1008: # At this time we're requesting the key.
1009: # again, this is essentially a no-op.
1010: # we'll write the next chunk until the
1011: # state changes.
1012:
1013: if($Socket->Writable() != 0) {
1014: # Write resulted in an error.
1015:
1016: $Watcher->cancel();
1017: KillSocket($Socket);
1018: return;
1019:
1020: }
1021: } elsif ($State eq "ReceivingKey") {
1022: # Now we need to wait for the key
1023: # to come back from the peer:
1024:
1025: $Watcher->cb(\&LondReadable);
1026: $Watcher->poll("r");
1027:
1028: } elsif ($State eq "SendingRequest") {
1029:
1030: # At this time we are sending a request to the
1031: # peer... write the next chunk:
1032:
1033: if($Socket->Writable() != 0) {
1034:
1035: if(exists($ActiveTransactions{$Socket})) {
1036: Debug(3, "Lond connection lost, failing transactions");
1037: FailTransaction($ActiveTransactions{$Socket});
1038: }
1039: $Watcher->cancel();
1040: KillSocket($Socket);
1041: return;
1042:
1043: }
1044:
1045: } elsif ($State eq "ReceivingReply") {
1046: # The send has completed. Wait for the
1047: # data to come in for a reply.
1048: Debug(8,"Writable sent request/receiving reply");
1049: $Watcher->cb(\&LondReadable);
1050: $Watcher->poll("r");
1051:
1052: } else {
1053: # Control only passes here on an error:
1054: # the socket state does not match any
1055: # of the known states... so an error
1056: # must be logged.
1057:
1058: &Debug(4, "Invalid socket state ".$State."\n");
1059: }
1060:
1061: }
1062: =pod
1063:
1064: =cut
1065: sub QueueDelayed {
1066: Debug(3,"QueueDelayed called");
1067:
1068: my $path = "$perlvar{'lonSockDir'}/delayed";
1069:
1070: Debug(4, "Delayed path: ".$path);
1071: opendir(DIRHANDLE, $path);
1072:
1073: my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
1074: closedir(DIRHANDLE);
1075: my $dfname;
1076: my $reqfile;
1077: foreach $dfname (sort @alldelayed) {
1078: $reqfile = "$path/$dfname";
1079: Debug(4, "queueing ".$reqfile);
1080: my $Handle = IO::File->new($reqfile);
1081: my $cmd = <$Handle>;
1082: chomp $cmd; # There may or may not be a newline...
1083: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1084: my $Transaction = LondTransaction->new($cmd);
1085: $Transaction->SetDeferred($reqfile);
1086: QueueTransaction($Transaction);
1087: }
1088:
1089: }
1090:
1091: =pod
1092:
1093: =head2 MakeLondConnection
1094:
1095: Create a new lond connection object, and start it towards its initial
1096: idleness. Once idle, it becomes elligible to receive transactions
1097: from the work queue. If the work queue is not empty when the
1098: connection is completed and becomes idle, it will dequeue an entry and
1099: start off on it.
1100:
1101: =cut
1102:
1103: sub MakeLondConnection {
1104: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1105: .GetServerPort());
1106:
1107: my $Connection = LondConnection->new(&GetServerHost(),
1108: &GetServerPort());
1109:
1110: if($Connection eq undef) { # Needs to be more robust later.
1111: Log("CRITICAL","Failed to make a connection with lond.");
1112: $ConnectionRetriesLeft--;
1113: return 0; # Failure.
1114: } else {
1115:
1116: # The connection needs to have writability
1117: # monitored in order to send the init sequence
1118: # that starts the whole authentication/key
1119: # exchange underway.
1120: #
1121: my $Socket = $Connection->GetSocket();
1122: if($Socket eq undef) {
1123: die "did not get a socket from the connection";
1124: } else {
1125: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1126: }
1127:
1128: $Connection->SetTimeoutCallback(\&SocketTimeout);
1129:
1130: my $event = Event->io(fd => $Socket,
1131: poll => 'w',
1132: cb => \&LondWritable,
1133: data => $Connection,
1134: desc => 'Connection to lond server');
1135: $ActiveConnections{$Connection} = $event;
1136:
1137: $ConnectionCount++;
1138: Debug(4, "Connection count = ".$ConnectionCount);
1139: if($ConnectionCount == 1) { # First Connection:
1140: QueueDelayed;
1141: }
1142: Log("SUCESS", "Created connection ".$ConnectionCount
1143: ." to host ".GetServerHost());
1144: return 1; # Return success.
1145: }
1146:
1147: }
1148:
1149: =pod
1150:
1151: =head2 StartRequest
1152:
1153: Starts a lond request going on a specified lond connection.
1154: parameters are:
1155:
1156: =item $Lond
1157:
1158: Connection to the lond that will send the transaction and receive the
1159: reply.
1160:
1161: =item $Client
1162:
1163: Connection to the client that is making this request We got the
1164: request from this socket, and when the request has been relayed to
1165: lond and we get a reply back from lond it will get sent to this
1166: socket.
1167:
1168: =item $Request
1169:
1170: The text of the request to send.
1171:
1172: =cut
1173:
1174: sub StartRequest {
1175: my $Lond = shift;
1176: my $Request = shift; # This is a LondTransaction.
1177:
1178: Debug(6, "StartRequest: ".$Request->getRequest());
1179:
1180: my $Socket = $Lond->GetSocket();
1181:
1182: $Request->Activate($Lond);
1183: $ActiveTransactions{$Lond} = $Request;
1184:
1185: $Lond->InitiateTransaction($Request->getRequest());
1186: my $event = Event->io(fd => $Socket,
1187: poll => "w",
1188: cb => \&LondWritable,
1189: data => $Lond,
1190: desc => "lond transaction connection");
1191: $ActiveConnections{$Lond} = $event;
1192: Debug(8," Start Request made watcher data with ".$event->data."\n");
1193: }
1194:
1195: =pod
1196:
1197: =head2 QueueTransaction
1198:
1199: If there is an idle lond connection, it is put to work doing this
1200: transaction. Otherwise, the transaction is placed in the work queue.
1201: If placed in the work queue and the maximum number of connections has
1202: not yet been created, a new connection will be started. Our goal is
1203: to eventually have a sufficient number of connections that the work
1204: queue will typically be empty. parameters are:
1205:
1206: =item Socket
1207:
1208: open on the lonc client.
1209:
1210: =item Request
1211:
1212: data to send to the lond.
1213:
1214: =cut
1215:
1216: sub QueueTransaction {
1217:
1218: my $requestData = shift; # This is a LondTransaction.
1219: my $cmd = $requestData->getRequest();
1220:
1221: Debug(6,"QueueTransaction: ".$cmd);
1222:
1223: my $LondSocket = $IdleConnections->pop();
1224: if(!defined $LondSocket) { # Need to queue request.
1225: Debug(5,"Must queue...");
1226: $WorkQueue->enqueue($requestData);
1227: if($ConnectionCount < $MaxConnectionCount) {
1228: if($ConnectionRetriesLeft > 0) {
1229: Debug(5,"Starting additional lond connection");
1230: if(MakeLondConnection() == 0) {
1231: EmptyQueue(); # Fail transactions, can't make connection.
1232: }
1233: } else {
1234: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1235: EmptyQueue(); # It's worse than that ... he's dead Jim.
1236: }
1237: }
1238: } else { # Can start the request:
1239: Debug(8,"Can start...");
1240: StartRequest($LondSocket, $requestData);
1241: }
1242: }
1243:
1244: #-------------------------- Lonc UNIX socket handling ---------------------
1245:
1246: =pod
1247:
1248: =head2 ClientRequest
1249: Callback that is called when data can be read from the UNIX domain
1250: socket connecting us with an apache server process.
1251:
1252: =cut
1253:
1254: sub ClientRequest {
1255: Debug(6, "ClientRequest");
1256: my $event = shift;
1257: my $watcher = $event->w;
1258: my $socket = $watcher->fd;
1259: my $data = $watcher->data;
1260: my $thisread;
1261:
1262: Debug(9, " Watcher named: ".$watcher->desc);
1263:
1264: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1265: Debug(8, "rcv: data length = ".length($thisread)
1266: ." read =".$thisread);
1267: unless (defined $rv && length($thisread)) {
1268: # Likely eof on socket.
1269: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1270: close($socket);
1271: $watcher->cancel();
1272: delete($ActiveClients{$socket});
1273: return;
1274: }
1275: Debug(8,"Data: ".$data." this read: ".$thisread);
1276: $data = $data.$thisread; # Append new data.
1277: $watcher->data($data);
1278: if($data =~ /(.*\n)/) { # Request entirely read.
1279: if($data eq "close_connection_exit\n") {
1280: Log("CRITICAL",
1281: "Request Close Connection ... exiting");
1282: CloseAllLondConnections();
1283: exit;
1284: }
1285: Debug(8, "Complete transaction received: ".$data);
1286: if($LogTransactions) {
1287: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1288: }
1289: my $Transaction = LondTransaction->new($data);
1290: $Transaction->SetClient($socket);
1291: QueueTransaction($Transaction);
1292: $watcher->cancel(); # Done looking for input data.
1293: }
1294:
1295: }
1296:
1297:
1298: =pod
1299:
1300: =head2 NewClient
1301:
1302: Callback that is called when a connection is received on the unix
1303: socket for a new client of lonc. The callback is parameterized by the
1304: event.. which is a-priori assumed to be an io event, and therefore has
1305: an fd member that is the Listener socket. We Accept the connection
1306: and register a new event on the readability of that socket:
1307:
1308: =cut
1309:
1310: sub NewClient {
1311: Debug(6, "NewClient");
1312: my $event = shift; # Get the event parameters.
1313: my $watcher = $event->w;
1314: my $socket = $watcher->fd; # Get the event' socket.
1315: my $connection = $socket->accept(); # Accept the client connection.
1316: Debug(8,"Connection request accepted from "
1317: .GetPeername($connection, AF_UNIX));
1318:
1319:
1320: my $description = sprintf("Connection to lonc client %d",
1321: $ClientConnection);
1322: Debug(9, "Creating event named: ".$description);
1323: Event->io(cb => \&ClientRequest,
1324: poll => 'r',
1325: desc => $description,
1326: data => "",
1327: fd => $connection);
1328: $ActiveClients{$connection} = $ClientConnection;
1329: $ClientConnection++;
1330: }
1331:
1332: =pod
1333:
1334: =head2 GetLoncSocketPath
1335:
1336: Returns the name of the UNIX socket on which to listen for client
1337: connections.
1338:
1339: =cut
1340:
1341: sub GetLoncSocketPath {
1342: return $UnixSocketDir."/".GetServerHost();
1343: }
1344:
1345: =pod
1346:
1347: =head2 GetServerHost
1348:
1349: Returns the host whose lond we talk with.
1350:
1351: =cut
1352:
1353: sub GetServerHost {
1354: return $RemoteHost; # Setup by the fork.
1355: }
1356:
1357: =pod
1358:
1359: =head2 GetServerPort
1360:
1361: Returns the lond port number.
1362:
1363: =cut
1364:
1365: sub GetServerPort {
1366: return $perlvar{londPort};
1367: }
1368:
1369: =pod
1370:
1371: =head2 SetupLoncListener
1372:
1373: Setup a lonc listener event. The event is called when the socket
1374: becomes readable.. that corresponds to the receipt of a new
1375: connection. The event handler established will accept the connection
1376: (creating a communcations channel), that int turn will establish
1377: another event handler to subess requests.
1378:
1379: =cut
1380:
1381: sub SetupLoncListener {
1382:
1383: my $socket;
1384: my $SocketName = GetLoncSocketPath();
1385: unlink($SocketName);
1386: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1387: Listen => 10,
1388: Type => SOCK_STREAM)) {
1389: die "Failed to create a lonc listner socket";
1390: }
1391: Event->io(cb => \&NewClient,
1392: poll => 'r',
1393: desc => 'Lonc listener Unix Socket',
1394: fd => $socket);
1395: }
1396:
1397: #
1398: # Toggle transaction logging.
1399: # Implicit inputs:
1400: # LogTransactions
1401: # Implicit Outputs:
1402: # LogTransactions
1403: sub ToggleTransactionLogging {
1404: print STDERR "Toggle transaction logging...\n";
1405: if(!$LogTransactions) {
1406: $LogTransactions = 1;
1407: } else {
1408: $LogTransactions = 0;
1409: }
1410:
1411:
1412: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1413: }
1414:
1415: =pod
1416:
1417: =head2 ChildStatus
1418:
1419: Child USR1 signal handler to report the most recent status
1420: into the status file.
1421:
1422: We also use this to reset the retries count in order to allow the
1423: client to retry connections with a previously dead server.
1424: =cut
1425: sub ChildStatus {
1426: my $event = shift;
1427: my $watcher = $event->w;
1428:
1429: Debug(2, "Reporting child status because : ".$watcher->data);
1430: my $docdir = $perlvar{'lonDocRoot'};
1431: my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
1432: print $fh $$."\t".$RemoteHost."\t".$Status."\t".
1433: $RecentLogEntry."\n";
1434: #
1435: # Write out information about each of the connections:
1436: #
1437: print $fh "Active connection statuses: \n";
1438: my $i = 1;
1439: print STDERR "================================= Socket Status Dump:\n";
1440: foreach my $item (keys %ActiveConnections) {
1441: my $Socket = $ActiveConnections{$item}->data;
1442: my $state = $Socket->GetState();
1443: print $fh "Connection $i State: $state\n";
1444: print STDERR "---------------------- Connection $i \n";
1445: $Socket->Dump();
1446: $i++;
1447: }
1448: $ConnectionRetriesLeft = $ConnectionRetries;
1449: }
1450:
1451: =pod
1452:
1453: =head2 SignalledToDeath
1454:
1455: Called in response to a signal that causes a chid process to die.
1456:
1457: =cut
1458:
1459:
1460: sub SignalledToDeath {
1461: my $event = shift;
1462: my $watcher= $event->w;
1463:
1464: Debug(2,"Signalled to death! via ".$watcher->data);
1465: my ($signal) = $watcher->data;
1466: chomp($signal);
1467: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1468: ."died through "."\"$signal\"");
1469: LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1470: ."\"$signal\"");
1471: exit 0;
1472:
1473: }
1474:
1475: =head2 ToggleDebug
1476:
1477: This sub toggles trace debugging on and off.
1478:
1479: =cut
1480:
1481: sub ToggleDebug {
1482: my $Current = $DebugLevel;
1483: $DebugLevel = $NextDebugLevel;
1484: $NextDebugLevel = $Current;
1485:
1486: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1487:
1488: }
1489:
1490: =head2 ChildProcess
1491:
1492: This sub implements a child process for a single lonc daemon.
1493:
1494: =cut
1495:
1496: sub ChildProcess {
1497:
1498:
1499: #
1500: # Signals must be handled by the Event framework...
1501: #
1502:
1503: Event->signal(signal => "QUIT",
1504: cb => \&SignalledToDeath,
1505: data => "QUIT");
1506: Event->signal(signal => "HUP",
1507: cb => \&ChildStatus,
1508: data => "HUP");
1509: Event->signal(signal => "USR1",
1510: cb => \&ChildStatus,
1511: data => "USR1");
1512: Event->signal(signal => "USR2",
1513: cb => \&ToggleTransactionLogging);
1514: Event->signal(signal => "INT",
1515: cb => \&ToggleDebug,
1516: data => "INT");
1517:
1518: SetupTimer();
1519:
1520: SetupLoncListener();
1521:
1522: $Event::Debuglevel = $DebugLevel;
1523:
1524: Debug(9, "Making initial lond connection for ".$RemoteHost);
1525:
1526: # Setup the initial server connection:
1527:
1528: # &MakeLondConnection(); // let first work requirest do it.
1529:
1530:
1531: Debug(9,"Entering event loop");
1532: my $ret = Event::loop(); # Start the main event loop.
1533:
1534:
1535: die "Main event loop exited!!!";
1536: }
1537:
1538: # Create a new child for host passed in:
1539:
1540: sub CreateChild {
1541: my $sigset = POSIX::SigSet->new(SIGINT);
1542: sigprocmask(SIG_BLOCK, $sigset);
1543: my $host = shift;
1544: $RemoteHost = $host;
1545: Log("CRITICAL", "Forking server for ".$host);
1546: my $pid = fork;
1547: if($pid) { # Parent
1548: $RemoteHost = "Parent";
1549: $ChildHash{$pid} = $host;
1550: $HostToPid{$host}= $pid;
1551: sigprocmask(SIG_UNBLOCK, $sigset);
1552:
1553: } else { # child.
1554: ShowStatus("Connected to ".$RemoteHost);
1555: $SIG{INT} = 'DEFAULT';
1556: sigprocmask(SIG_UNBLOCK, $sigset);
1557: ChildProcess; # Does not return.
1558: }
1559:
1560: }
1561: #
1562: # Parent process logic pass 1:
1563: # For each entry in the hosts table, we will
1564: # fork off an instance of ChildProcess to service the transactions
1565: # to that host. Each pid will be entered in a global hash
1566: # with the value of the key, the host.
1567: # The parent will then enter a loop to wait for process exits.
1568: # Each exit gets logged and the child gets restarted.
1569: #
1570:
1571: #
1572: # Fork and start in new session so hang-up isn't going to
1573: # happen without intent.
1574: #
1575:
1576:
1577:
1578:
1579:
1580:
1581: ShowStatus("Forming new session");
1582: my $childpid = fork;
1583: if ($childpid != 0) {
1584: sleep 4; # Give child a chacne to break to
1585: exit 0; # a new sesion.
1586: }
1587: #
1588: # Write my pid into the pid file so I can be located
1589: #
1590:
1591: ShowStatus("Parent writing pid file:");
1592: my $execdir = $perlvar{'lonDaemons'};
1593: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1594: print PIDSAVE "$$\n";
1595: close(PIDSAVE);
1596:
1597:
1598:
1599: if (POSIX::setsid() < 0) {
1600: print "Could not create new session\n";
1601: exit -1;
1602: }
1603:
1604: ShowStatus("Forking node servers");
1605:
1606: Log("CRITICAL", "--------------- Starting children ---------------");
1607:
1608: LondConnection::ReadConfig; # Read standard config files.
1609: my $HostIterator = LondConnection::GetHostIterator;
1610: while (! $HostIterator->end()) {
1611:
1612: my $hostentryref = $HostIterator->get();
1613: CreateChild($hostentryref->[0]);
1614: $HostHash{$hostentryref->[0]} = $hostentryref->[4];
1615: $HostIterator->next();
1616: }
1617: $RemoteHost = "Parent Server";
1618:
1619: # Maintain the population:
1620:
1621: ShowStatus("Parent keeping the flock");
1622:
1623: #
1624: # Set up parent signals:
1625: #
1626:
1627: $SIG{INT} = \&Terminate;
1628: $SIG{TERM} = \&Terminate;
1629: $SIG{HUP} = \&Restart;
1630: $SIG{USR1} = \&CheckKids;
1631: $SIG{USR2} = \&UpdateKids; # LonManage update request.
1632:
1633: while(1) {
1634: my $deadchild = wait();
1635: if(exists $ChildHash{$deadchild}) { # need to restart.
1636: my $deadhost = $ChildHash{$deadchild};
1637: delete($HostToPid{$deadhost});
1638: delete($ChildHash{$deadchild});
1639: Log("WARNING","Lost child pid= ".$deadchild.
1640: "Connected to host ".$deadhost);
1641: Log("INFO", "Restarting child procesing ".$deadhost);
1642: CreateChild($deadhost);
1643: }
1644: }
1645:
1646:
1647:
1648: =pod
1649:
1650: =head1 CheckKids
1651:
1652: Since kids do not die as easily in this implementation
1653: as the previous one, there is no need to restart the
1654: dead ones (all dead kids get restarted when they die!!)
1655: The only thing this function does is to pass USR1 to the
1656: kids so that they report their status.
1657:
1658: =cut
1659:
1660: sub CheckKids {
1661: Debug(2, "Checking status of children");
1662: my $docdir = $perlvar{'lonDocRoot'};
1663: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
1664: my $now=time;
1665: my $local=localtime($now);
1666: print $fh "LONC status $local - parent $$ \n\n";
1667: foreach my $pid (keys %ChildHash) {
1668: Debug(2, "Sending USR1 -> $pid");
1669: kill 'USR1' => $pid; # Tell Child to report status.
1670: sleep 1; # Wait so file doesn't intermix.
1671: }
1672: }
1673:
1674: =pod
1675:
1676: =head1 UpdateKids
1677:
1678: parent's SIGUSR2 handler. This handler:
1679:
1680: =item
1681:
1682: Rereads the hosts file.
1683:
1684: =item
1685:
1686: Kills off (via sigint) children for hosts that have disappeared.
1687:
1688: =item
1689:
1690: QUITs children for hosts that already exist (this just forces a status display
1691: and resets the connection retry count for that host.
1692:
1693: =item
1694:
1695: Starts new children for hosts that have been added to the hosts.tab file since
1696: the start of the master program and maintains them.
1697:
1698: =cut
1699:
1700: sub UpdateKids {
1701:
1702: Log("INFO", "Updating connections via SIGUSR2");
1703:
1704: # Just in case we need to kill our own lonc, we wait a few seconds to
1705: # give it a chance to receive and relay lond's response to the
1706: # re-init command.
1707: #
1708:
1709: sleep(2); # Wait a couple of seconds.
1710:
1711: my %hosts; # Indexed by loncapa hostname, value=ip.
1712:
1713: # Need to re-read the host table:
1714:
1715:
1716: LondConnection::ReadConfig();
1717: my $I = LondConnection::GetHostIterator;
1718: while (! $I->end()) {
1719: my $item = $I->get();
1720: $hosts{$item->[0]} = $item->[4];
1721: $I->next();
1722: }
1723:
1724: # The logic below is written for clarity not for efficiency.
1725: # Since I anticipate that this function is only rarely called, that's
1726: # appropriate. There are certainly ways to combine the loops below,
1727: # and anyone wishing to obscure the logic is welcome to go for it.
1728: # Note that we don't re-direct sigchild. Instead we do what's needed
1729: # to the data structures that keep track of children to ensure that
1730: # when sigchild is honored, no new child is born.
1731: #
1732:
1733: # For each existing child; if it's host doesn't exist, kill the child.
1734:
1735: foreach my $child (keys %ChildHash) {
1736: my $oldhost = $ChildHash{$child};
1737: if (!(exists $hosts{$oldhost})) {
1738: Log("CRITICAL", "Killing child for $oldhost host no longer exists");
1739: delete $ChildHash{$child};
1740: delete $HostToPid{$oldhost};
1741: kill 'QUIT' => $child;
1742: }
1743: }
1744: # For each remaining existing child; if it's host's ip has changed,
1745: # Restart the child on the new IP.
1746:
1747: foreach my $child (keys %ChildHash) {
1748: my $oldhost = $ChildHash{$child};
1749: my $oldip = $HostHash{$oldhost};
1750: if ($hosts{$oldhost} ne $oldip) {
1751:
1752: # kill the old child.
1753:
1754: Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
1755: delete $ChildHash{$child};
1756: delete $HostToPid{$oldhost};
1757: kill 'QUIT' => $child;
1758:
1759: # Do the book-keeping needed to start a new child on the
1760: # new ip.
1761:
1762: $HostHash{$oldhost} = $hosts{$oldhost};
1763: CreateChild($oldhost);
1764: }
1765: }
1766: # Finally, for each new host, not in the host hash, create a
1767: # enter the host and create a new child.
1768: # Force a status display of any existing process.
1769:
1770: foreach my $host (keys %hosts) {
1771: if(!(exists $HostHash{$host})) {
1772: Log("INFO", "New host $host discovered in hosts.tab...");
1773: $HostHash{$host} = $hosts{$host};
1774: CreateChild($host);
1775: } else {
1776: kill 'HUP' => $HostToPid{$host}; # status display.
1777: }
1778: }
1779: }
1780:
1781:
1782: =pod
1783:
1784: =head1 Restart
1785:
1786: Signal handler for HUP... all children are killed and
1787: we self restart. This is an el-cheapo way to re read
1788: the config file.
1789:
1790: =cut
1791:
1792: sub Restart {
1793: &KillThemAll; # First kill all the children.
1794: Log("CRITICAL", "Restarting");
1795: my $execdir = $perlvar{'lonDaemons'};
1796: unlink("$execdir/logs/lonc.pid");
1797: exec("$execdir/loncnew");
1798: }
1799:
1800: =pod
1801:
1802: =head1 KillThemAll
1803:
1804: Signal handler that kills all children by sending them a
1805: SIGHUP. Responds to sigint and sigterm.
1806:
1807: =cut
1808:
1809: sub KillThemAll {
1810: Debug(2, "Kill them all!!");
1811: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
1812: foreach my $pid (keys %ChildHash) {
1813: my $serving = $ChildHash{$pid};
1814: Debug(2, "Killing lonc for $serving pid = $pid");
1815: ShowStatus("Killing lonc for $serving pid = $pid");
1816: Log("CRITICAL", "Killing lonc for $serving pid = $pid");
1817: kill 'QUIT' => $pid;
1818: delete($ChildHash{$pid});
1819: }
1820: my $execdir = $perlvar{'lonDaemons'};
1821: unlink("$execdir/logs/lonc.pid");
1822:
1823: }
1824:
1825: =pod
1826:
1827: =head1 Terminate
1828:
1829: Terminate the system.
1830:
1831: =cut
1832:
1833: sub Terminate {
1834: KillThemAll;
1835: Log("CRITICAL","Master process exiting");
1836: exit 0;
1837:
1838: }
1839: =pod
1840:
1841: =head1 Theory
1842:
1843: The event class is used to build this as a single process with an
1844: event driven model. The following events are handled:
1845:
1846: =item UNIX Socket connection Received
1847:
1848: =item Request data arrives on UNIX data transfer socket.
1849:
1850: =item lond connection becomes writable.
1851:
1852: =item timer fires at 1 second intervals.
1853:
1854: All sockets are run in non-blocking mode. Timeouts managed by the timer
1855: handler prevents hung connections.
1856:
1857: Key data structures:
1858:
1859: =item RequestQueue
1860:
1861: A queue of requests received from UNIX sockets that are
1862: waiting for a chance to be forwarded on a lond connection socket.
1863:
1864: =item ActiveConnections
1865:
1866: A hash of lond connections that have transactions in process that are
1867: available to be timed out.
1868:
1869: =item ActiveTransactions
1870:
1871: A hash indexed by lond connections that contain the client reply
1872: socket for each connection that has an active transaction on it.
1873:
1874: =item IdleConnections
1875:
1876: A hash of lond connections that have no work to do. These connections
1877: can be closed if they are idle for a long enough time.
1878:
1879: =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