1: #!/usr/bin/perl
2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
5: # $Id: loncnew,v 1.105 2018/08/07 17:12:09 raeburn 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 over 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 Fcntl qw(:flock);
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 %ChildPid; # by pid -> host.
76: my %ChildHost; # by host.
77: my %ChildKeyMode; # by pid -> keymode
78: my %listening_to; # Socket->host table for who the parent
79: # is listening to.
80: my %parent_dispatchers; # host-> listener watcher events.
81:
82: my %parent_handlers; # Parent signal handlers...
83:
84: my $MaxConnectionCount = 10; # Will get from config later.
85: my $ClientConnection = 0; # Uniquifier for client events.
86:
87: my $DebugLevel = 0;
88: my $NextDebugLevel= 2; # So Sigint can toggle this.
89: my $IdleTimeout= 5*60; # Seconds to wait prior to pruning connections.
90:
91: my $LogTransactions = 0; # When True, all transactions/replies get logged.
92: my $executable = $0; # Get the full path to me.
93:
94: #
95: # The variables below are only used by the child processes.
96: #
97: my $RemoteHost; # Name of host child is talking to.
98: my $RemoteHostId; # default lonid of host child is talking to.
99: my @all_host_ids;
100: my $UnixSocketDir= $perlvar{'lonSockDir'};
101: my $IdleConnections = Stack->new(); # Set of idle connections
102: my %ActiveConnections; # Connections to the remote lond.
103: my %ActiveTransactions; # LondTransactions in flight.
104: my %ActiveClients; # Serial numbers of active clients by socket.
105: my $WorkQueue = Queue->new(); # Queue of pending transactions.
106: my $ConnectionCount = 0;
107: my $IdleSeconds = 0; # Number of seconds idle.
108: my $Status = ""; # Current status string.
109: my $RecentLogEntry = "";
110: my $ConnectionRetries=5; # Number of connection retries allowed.
111: my $ConnectionRetriesLeft=5; # Number of connection retries remaining.
112: my $LondVersion = "unknown"; # Version of lond we talk with.
113: my $KeyMode = ""; # e.g. ssl, local, insecure from last connect.
114: my $LondConnecting = 0; # True when a connection is being built.
115:
116:
117:
118: my $I_am_child = 0; # True if this is the child process.
119:
120: #
121: # The hash below gives the HTML format for log messages
122: # given a severity.
123: #
124: my %LogFormats;
125:
126: $LogFormats{"CRITICAL"} = "<font color='red'>CRITICAL: %s</font>";
127: $LogFormats{"SUCCESS"} = "<font color='green'>SUCCESS: %s</font>";
128: $LogFormats{"INFO"} = "<font color='yellow'>INFO: %s</font>";
129: $LogFormats{"WARNING"} = "<font color='blue'>WARNING: %s</font>";
130: $LogFormats{"DEFAULT"} = " %s ";
131:
132:
133: # UpdateStatus;
134: # Update the idle status display to show how many connections
135: # are left, retries and other stuff.
136: #
137: sub UpdateStatus {
138: if ($ConnectionRetriesLeft > 0) {
139: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
140: ." Retries remaining: ".$ConnectionRetriesLeft
141: ." ($KeyMode)");
142: } else {
143: ShowStatus(GetServerHost()." >> DEAD <<");
144: }
145: }
146:
147:
148: =pod
149:
150: =head2 LogPerm
151:
152: Makes an entry into the permanent log file.
153:
154: =cut
155:
156: sub LogPerm {
157: my $message=shift;
158: my $execdir=$perlvar{'lonDaemons'};
159: my $now=time;
160: my $local=localtime($now);
161: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
162: chomp($message);
163: print $fh "$now:$message:$local\n";
164: }
165:
166: =pod
167:
168: =head2 Log
169:
170: Logs a message to the log file.
171: Parameters:
172:
173: =item severity
174:
175: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
176: format string used to format the message. if the severity is
177: not a defined severity the Default format string is used.
178:
179: =item message
180:
181: The base message. In addtion to the format string, the message
182: will be appended to a string containing the name of our remote
183: host and the time will be formatted into the message.
184:
185: =cut
186:
187: sub Log {
188:
189: my ($severity, $message) = @_;
190:
191: if(!$LogFormats{$severity}) {
192: $severity = "DEFAULT";
193: }
194:
195: my $format = $LogFormats{$severity};
196:
197: # Put the window dressing in in front of the message format:
198:
199: my $now = time;
200: my $local = localtime($now);
201: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
202: $finalformat = $finalformat.$format."\n";
203:
204: # open the file and put the result.
205:
206: my $execdir = $perlvar{'lonDaemons'};
207: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
208: my $msg = sprintf($finalformat, $message);
209: $RecentLogEntry = $msg;
210: print $fh $msg;
211:
212:
213: }
214:
215:
216: =pod
217:
218: =head2 GetPeerName
219:
220: Returns the name of the host that a socket object is connected to.
221:
222: =cut
223:
224: sub GetPeername {
225:
226:
227: my ($connection, $AdrFamily) = @_;
228:
229: my $peer = $connection->peername();
230: my $peerport;
231: my $peerip;
232: if($AdrFamily == AF_INET) {
233: ($peerport, $peerip) = sockaddr_in($peer);
234: my $peername = gethostbyaddr($peerip, $AdrFamily);
235: return $peername;
236: } elsif ($AdrFamily == AF_UNIX) {
237: my $peerfile;
238: ($peerfile) = sockaddr_un($peer);
239: return $peerfile;
240: }
241: }
242: =pod
243:
244: =head2 Debug
245:
246: Invoked to issue a debug message.
247:
248: =cut
249:
250: sub Debug {
251:
252: my ($level, $message) = @_;
253:
254: if ($level <= $DebugLevel) {
255: Log("INFO", "-Debug- $message host = $RemoteHost");
256: }
257: }
258:
259: sub SocketDump {
260:
261: my ($level, $socket) = @_;
262:
263: if($level <= $DebugLevel) {
264: $socket->Dump(-1); # Ensure it will get dumped.
265: }
266: }
267:
268: =pod
269:
270: =head2 ShowStatus
271:
272: Place some text as our pid status.
273: and as what we return in a SIGUSR1
274:
275: =cut
276:
277: sub ShowStatus {
278: my $state = shift;
279: my $now = time;
280: my $local = localtime($now);
281: $Status = $local.": ".$state;
282: $0='lonc: '.$state.' '.$local;
283: }
284:
285: =pod
286:
287: =head2 SocketTimeout
288:
289: Called when an action on the socket times out. The socket is
290: destroyed and any active transaction is failed.
291:
292:
293: =cut
294:
295: sub SocketTimeout {
296: my $Socket = shift;
297: Log("WARNING", "A socket timeout was detected");
298: Debug(5, " SocketTimeout called: ");
299: $Socket->Dump(0);
300: if(exists($ActiveTransactions{$Socket})) {
301: FailTransaction($ActiveTransactions{$Socket});
302: }
303: KillSocket($Socket); # A transaction timeout also counts as
304: # a connection failure:
305: $ConnectionRetriesLeft--;
306: if($ConnectionRetriesLeft <= 0) {
307: Log("CRITICAL", "Host marked DEAD: ".GetServerHost());
308: $LondConnecting = 0;
309: }
310:
311: }
312:
313: #
314: # This function should be called by the child in all cases where it must
315: # exit. The child process must create a lock file for the AF_UNIX socket
316: # in order to prevent connection requests from lonnet in the time between
317: # process exit and the parent picking up the listen again.
318: #
319: # Parameters:
320: # exit_code - Exit status value, however see the next parameter.
321: # message - If this optional parameter is supplied, the exit
322: # is via a die with this message.
323: #
324: sub child_exit {
325: my ($exit_code, $message) = @_;
326:
327: # Regardless of how we exit, we may need to do the lock thing:
328:
329: #
330: # Create a lock file since there will be a time window
331: # between our exit and the parent's picking up the listen
332: # during which no listens will be done on the
333: # lonnet client socket.
334: #
335: my $lock_file = &GetLoncSocketPath().".lock";
336: open(LOCK,">$lock_file");
337: print LOCK "Contents not important";
338: close(LOCK);
339: unlink(&GetLoncSocketPath());
340:
341: if ($message) {
342: die($message);
343: } else {
344: exit($exit_code);
345: }
346: }
347: #----------------------------- Timer management ------------------------
348:
349: =pod
350:
351: =head2 Tick
352:
353: Invoked each timer tick.
354:
355: =cut
356:
357:
358: sub Tick {
359: my ($Event) = @_;
360: my $clock_watcher = $Event->w;
361:
362: my $client;
363: UpdateStatus();
364:
365: # Is it time to prune connection count:
366:
367:
368: if($IdleConnections->Count() &&
369: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
370: $IdleSeconds++;
371: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
372: my $Socket = $IdleConnections->pop();
373: KillSocket($Socket);
374: $IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
375: UpdateStatus();
376: if(($ConnectionCount == 0)) {
377: &child_exit(0);
378:
379: }
380: }
381: } else {
382: $IdleSeconds = 0; # Reset idle count if not idle.
383: }
384: #
385: # For each inflight transaction, tick down its timeout counter.
386: #
387:
388: foreach my $item (keys %ActiveConnections) {
389: my $State = $ActiveConnections{$item}->data->GetState();
390: if ($State ne 'Idle') {
391: Debug(5,"Ticking Socket $State $item");
392: $ActiveConnections{$item}->data->Tick();
393: }
394: }
395: # Do we have work in the queue, but no connections to service them?
396: # If so, try to make some new connections to get things going again.
397: #
398: # Note this code is dead now...
399: #
400: my $Requests = $WorkQueue->Count();
401: if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) {
402: if ($ConnectionRetriesLeft > 0) {
403: Debug(5,"Work but no connections, Make a new one");
404: my $success;
405: $success = &MakeLondConnection;
406: if($success == 0) { # All connections failed:
407: Debug(5,"Work in queue failed to make any connectiouns\n");
408: EmptyQueue(); # Fail pending transactions with con_lost.
409: CloseAllLondConnections(); # Should all be closed but....
410: }
411: } else {
412: $LondConnecting = 0;
413: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
414: Debug(5,"Work in queue, but gave up on connections..flushing\n");
415: EmptyQueue(); # Connections can't be established.
416: CloseAllLondConnections(); # Should all already be closed but...
417: }
418:
419: }
420: if ($ConnectionCount == 0) {
421: $KeyMode = "";
422: $clock_watcher->cancel();
423: }
424: &UpdateStatus();
425: }
426:
427: =pod
428:
429: =head2 SetupTimer
430:
431: Sets up a 1 per sec recurring timer event. The event handler is used to:
432:
433: =item
434:
435: Trigger timeouts on communications along active sockets.
436:
437: =item
438:
439: Trigger disconnections of idle sockets.
440:
441: =cut
442:
443: sub SetupTimer {
444: Debug(6, "SetupTimer");
445: Event->timer(interval => 1, cb => \&Tick,
446: hard => 1);
447: }
448:
449: =pod
450:
451: =head2 ServerToIdle
452:
453: This function is called when a connection to the server is
454: ready for more work.
455:
456: If there is work in the Work queue the top element is dequeued
457: and the connection will start to work on it. If the work queue is
458: empty, the connection is pushed on the idle connection stack where
459: it will either get another work unit, or alternatively, if it sits there
460: long enough, it will be shut down and released.
461:
462: =cut
463:
464: sub ServerToIdle {
465: my $Socket = shift; # Get the socket.
466: $KeyMode = $Socket->{AuthenticationMode};
467: delete($ActiveTransactions{$Socket}); # Server has no transaction
468:
469: &Debug(5, "Server to idle");
470:
471: # If there's work to do, start the transaction:
472:
473: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
474: if ($reqdata ne undef) {
475: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
476: &StartRequest($Socket, $reqdata);
477:
478: } else {
479:
480: # There's no work waiting, so push the server to idle list.
481: &Debug(5, "No new work requests, server connection going idle");
482: $IdleConnections->push($Socket);
483: }
484: }
485:
486: =pod
487:
488: =head2 ClientWritable
489:
490: Event callback for when a client socket is writable.
491:
492: This callback is established when a transaction response is
493: available from lond. The response is forwarded to the unix socket
494: as it becomes writable in this sub.
495:
496: Parameters:
497:
498: =item Event
499:
500: The event that has been triggered. Event->w->data is
501: the data and Event->w->fd is the socket to write.
502:
503: =cut
504:
505: sub ClientWritable {
506: my $Event = shift;
507: my $Watcher = $Event->w;
508: if (!defined($Watcher)) {
509: &child_exit(-1,'No watcher for event in ClientWritable');
510: }
511: my $Data = $Watcher->data;
512: my $Socket = $Watcher->fd;
513:
514: # Try to send the data:
515:
516: &Debug(6, "ClientWritable writing".$Data);
517: &Debug(9, "Socket is: ".$Socket);
518:
519: if($Socket->connected) {
520: my $result = $Socket->send($Data, 0);
521:
522: # $result undefined: the write failed.
523: # otherwise $result is the number of bytes written.
524: # Remove that preceding string from the data.
525: # If the resulting data is empty, destroy the watcher
526: # and set up a read event handler to accept the next
527: # request.
528:
529: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
530: if($result ne undef) {
531: &Debug(9, "send result was defined");
532: if($result == length($Data)) { # Entire string sent.
533: &Debug(9, "ClientWritable data all written");
534: $Watcher->cancel();
535: #
536: # Set up to read next request from socket:
537:
538: my $descr = sprintf("Connection to lonc client %d",
539: $ActiveClients{$Socket});
540: Event->io(cb => \&ClientRequest,
541: poll => 'r',
542: desc => $descr,
543: data => "",
544: fd => $Socket);
545:
546: } else { # Partial string sent.
547: $Watcher->data(substr($Data, $result));
548: if($result == 0) { # client hung up on us!!
549: # Log("INFO", "lonc pipe client hung up on us!");
550: $Watcher->cancel;
551: $Socket->shutdown(2);
552: $Socket->close();
553: }
554: }
555:
556: } else { # Error of some sort...
557:
558: # Some errnos are possible:
559: my $errno = $!;
560: if($errno == POSIX::EWOULDBLOCK ||
561: $errno == POSIX::EAGAIN ||
562: $errno == POSIX::EINTR) {
563: # No action taken...the socket will be writable firing the event again
564: # which will result in a retry of the write.
565: } else { # Unanticipated errno.
566: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
567: $Watcher->cancel; # Stop the watcher.
568: $Socket->shutdown(2); # Kill connection
569: $Socket->close(); # Close the socket.
570: }
571:
572: }
573: } else {
574: $Watcher->cancel(); # A delayed request...just cancel.
575: return;
576: }
577: }
578:
579: =pod
580:
581: =head2 CompleteTransaction
582:
583: Called when the reply data has been received for a lond
584: transaction. The reply data must now be sent to the
585: ultimate client on the other end of the Unix socket. This is
586: done by setting up a writable event for the socket with the
587: data the reply data.
588:
589: Parameters:
590:
591: =item Socket
592:
593: Socket on which the lond transaction occurred. This is a
594: LondConnection. The data received are in the TransactionReply member.
595:
596: =item Transaction
597:
598: The transaction that is being completed.
599:
600: =cut
601:
602: sub CompleteTransaction {
603: &Debug(5,"Complete transaction");
604:
605: my ($Socket, $Transaction) = @_;
606:
607: if (!$Transaction->isDeferred()) { # Normal transaction
608: my $data = $Socket->GetReply(); # Data to send.
609: if($LogTransactions) {
610: Log("SUCCESS", "Reply from lond: '$data'");
611: }
612: StartClientReply($Transaction, $data);
613: } else { # Delete deferred transaction file.
614: Log("SUCCESS", "A delayed transaction was completed");
615: LogPerm("S:".$Socket->PeerLoncapaHim().":".$Transaction->getRequest());
616: unlink($Transaction->getFile());
617: }
618: }
619:
620: =pod
621:
622: =head1 StartClientReply
623:
624: Initiates a reply to a client where the reply data is a parameter.
625:
626: =head2 parameters:
627:
628: =item Transaction
629:
630: The transaction for which we are responding to the client.
631:
632: =item data
633:
634: The data to send to apache client.
635:
636: =cut
637:
638: sub StartClientReply {
639:
640: my ($Transaction, $data) = @_;
641:
642: my $Client = $Transaction->getClient();
643:
644: &Debug(8," Reply was: ".$data);
645: my $Serial = $ActiveClients{$Client};
646: my $desc = sprintf("Connection to lonc client %d",
647: $Serial);
648: Event->io(fd => $Client,
649: poll => "w",
650: desc => $desc,
651: cb => \&ClientWritable,
652: data => $data);
653: }
654:
655: =pod
656:
657: =head2 FailTransaction
658:
659: Finishes a transaction with failure because the associated lond socket
660: disconnected. There are two possibilities:
661: - The transaction is deferred: in which case we just quietly
662: delete the transaction since there is no client connection.
663: - The transaction is 'live' in which case we initiate the sending
664: of "con_lost" to the client.
665:
666: Deleting the transaction means killing it from the %ActiveTransactions hash.
667:
668: Parameters:
669:
670: =item client
671:
672: The LondTransaction we are failing.
673:
674:
675: =cut
676:
677: sub FailTransaction {
678: my $transaction = shift;
679:
680: # If the socket is dead, that's already logged.
681:
682: if ($ConnectionRetriesLeft > 0) {
683: Log("WARNING", "Failing transaction "
684: .$transaction->getLoggableRequest());
685: }
686: Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
687: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
688: my $client = $transaction->getClient();
689: Debug(1," Replying con_lost to ".$transaction->getRequest());
690: StartClientReply($transaction, "con_lost\n");
691: }
692:
693: }
694:
695: =pod
696:
697: =head1 EmptyQueue
698:
699: Fails all items in the work queue with con_lost.
700: Note that each item in the work queue is a transaction.
701:
702: =cut
703:
704: sub EmptyQueue {
705: $ConnectionRetriesLeft--; # Counts as connection failure too.
706: while($WorkQueue->Count()) {
707: my $request = $WorkQueue->dequeue(); # This is a transaction
708: FailTransaction($request);
709: }
710: }
711:
712: =pod
713:
714: =head2 CloseAllLondConnections
715:
716: Close all connections open on lond prior to exit e.g.
717:
718: =cut
719:
720: sub CloseAllLondConnections {
721: foreach my $Socket (keys %ActiveConnections) {
722: if(exists($ActiveTransactions{$Socket})) {
723: FailTransaction($ActiveTransactions{$Socket});
724: }
725: KillSocket($Socket);
726: }
727: }
728:
729: =pod
730:
731: =head2 KillSocket
732:
733: Destroys a socket. This function can be called either when a socket
734: has died of 'natural' causes or because a socket needs to be pruned due to
735: idleness. If the socket has died naturally, if there are no longer any
736: live connections a new connection is created (in case there are transactions
737: in the queue). If the socket has been pruned, it is never re-created.
738:
739: Parameters:
740:
741: =item Socket
742:
743: The socket to kill off.
744:
745: =item restart
746:
747: non-zero if we are allowed to create a new connection.
748:
749: =cut
750:
751: sub KillSocket {
752: my $Socket = shift;
753: my $restart = shift;
754:
755: Log("WARNING", "Shutting down a socket");
756: $Socket->Shutdown();
757:
758: # If the socket came from the active connection set,
759: # delete its transaction... note that FailTransaction should
760: # already have been called!!!
761: # otherwise it came from the idle set.
762: #
763:
764: if(exists($ActiveTransactions{$Socket})) {
765: delete ($ActiveTransactions{$Socket});
766: }
767: if(exists($ActiveConnections{$Socket})) {
768: $ActiveConnections{$Socket}->cancel;
769: delete($ActiveConnections{$Socket});
770: # Decrement ConnectionCount unless we will immediately
771: # re-connect (i.e., $restart is true), because this was
772: # a connection where the SSL channel for exchange of the
773: # shared key failed, and we may use an insecure channel.
774: unless ($restart) {
775: $ConnectionCount--;
776: }
777: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
778: }
779: # If the connection count has gone to zero and there is work in the
780: # work queue, the work all gets failed with con_lost.
781: #
782:
783: if($ConnectionCount == 0) {
784: $LondConnecting = 0; # No connections so also not connecting.
785: EmptyQueue();
786: CloseAllLondConnections(); # Should all already be closed but...
787: &clear_childpid($$);
788: }
789: UpdateStatus();
790: }
791:
792: =pod
793:
794: =head2 LondReadable
795:
796: This function is called whenever a lond connection
797: is readable. The action is state dependent:
798:
799: =head3 State=Initialized
800:
801: We are waiting for the challenge, this is a no-op until the
802: state changes.
803:
804: =head3 State=Challenged
805:
806: The challenge has arrived we need to transition to Writable.
807: The connection must echo the challenge back.
808:
809: =head3 State=ChallengeReplied
810:
811: The challenge has been replied to. Then we are receiving the
812: 'ok' from the partner.
813:
814: =head3 State=ReadingVersionString
815:
816: We have requested the lond version and are reading the
817: version back. Upon completion, we'll store the version away
818: for future use(?).
819:
820: =head3 State=HostSet
821:
822: We have selected the domain name of our peer (multhomed hosts)
823: and are getting the reply (presumably ok) back.
824:
825: =head3 State=RequestingKey
826:
827: The ok has been received and we need to send the request for
828: an encryption key. Transition to writable for that.
829:
830: =head3 State=ReceivingKey
831:
832: The the key has been requested, now we are reading the new key.
833:
834: =head3 State=Idle
835:
836: The encryption key has been negotiated or we have finished
837: reading data from the a transaction. If the callback data have
838: a client as well as the socket information, then we are
839: doing a transaction and the data received are relayed to the client
840: before the socket is put on the idle list.
841:
842: =head3 State=SendingRequest
843:
844: I do not think this state can be received here, but if it is,
845: the appropriate thing to do is to transition to writable, and send
846: the request.
847:
848: =head3 State=ReceivingReply
849:
850: We finished sending the request to the server and now transition
851: to readable to receive the reply.
852:
853: The parameter to this function are:
854:
855: The event. Implicit in this is the watcher and its data. The data
856: contain at least the lond connection object and, if a
857: transaction is in progress, the socket attached to the local client.
858:
859: =cut
860:
861: sub LondReadable {
862:
863: my $Event = shift;
864: my $Watcher = $Event->w;
865: my $Socket = $Watcher->data;
866: my $client = undef;
867:
868: &Debug(6,"LondReadable called state = ".$Socket->GetState());
869:
870:
871: my $State = $Socket->GetState(); # All action depends on the state.
872:
873: SocketDump(6, $Socket);
874: my $status = $Socket->Readable();
875:
876: &Debug(2, "Socket->Readable returned: $status");
877:
878: if($status != 0) {
879: # bad return from socket read. Currently this means that
880: # The socket has become disconnected. We fail the transaction.
881:
882: Log("WARNING",
883: "Lond connection lost.");
884: my $state_on_exit = $Socket->GetState();
885: if(exists($ActiveTransactions{$Socket})) {
886: FailTransaction($ActiveTransactions{$Socket});
887: } else {
888: # Socket is connecting and failed... need to mark
889: # no longer connecting.
890: $LondConnecting = 0;
891: }
892: $Watcher->cancel();
893: if ($state_on_exit eq 'ReInitNoSSL') {
894: # SSL certificate verification failed, and insecure connection
895: # allowed. Send restart arg to KillSocket(), so EmptyQueue()
896: # is not called, as we still hope to process queued request.
897:
898: KillSocket($Socket,1);
899:
900: # Re-initiate creation of Lond Connection for use with queued
901: # request.
902:
903: ShowStatus("Connected to ".$RemoteHost);
904: Log("WARNING","No SSL channel (verification failed), will try with insecure channel");
905: &MakeLondConnection(1);
906:
907: } else {
908: KillSocket($Socket);
909: $ConnectionRetriesLeft--; # Counts as connection failure
910: }
911: return;
912: }
913: SocketDump(6,$Socket);
914:
915: $State = $Socket->GetState(); # Update in case of transition.
916: &Debug(6, "After read, state is ".$State);
917:
918: if($State eq "Initialized") {
919:
920:
921: } elsif ($State eq "ReInitNoSSL") {
922:
923: } elsif ($State eq "ChallengeReceived") {
924: # The challenge must be echoed back; The state machine
925: # in the connection takes care of setting that up. Just
926: # need to transition to writable:
927:
928: $Watcher->cb(\&LondWritable);
929: $Watcher->poll("w");
930:
931: } elsif ($State eq "ChallengeReplied") {
932:
933: } elsif ($State eq "RequestingVersion") {
934: # Need to ask for the version... that is writiability:
935:
936: $Watcher->cb(\&LondWritable);
937: $Watcher->poll("w");
938:
939: } elsif ($State eq "ReadingVersionString") {
940: # Read the rest of the version string...
941: } elsif ($State eq "SetHost") {
942: # Need to request the actual domain get set...
943:
944: $Watcher->cb(\&LondWritable);
945: $Watcher->poll("w");
946: } elsif ($State eq "HostSet") {
947: # Reading the 'ok' from the peer.
948:
949: } elsif ($State eq "RequestingKey") {
950: # The ok was received. Now we need to request the key
951: # That requires us to be writable:
952:
953: $Watcher->cb(\&LondWritable);
954: $Watcher->poll("w");
955:
956: } elsif ($State eq "ReceivingKey") {
957:
958: } elsif ($State eq "Idle") {
959:
960: if ($ConnectionCount == 1) {
961: # Write child Pid file to keep track of ssl and insecure
962: # connections
963:
964: &record_childpid($Socket);
965: }
966:
967: # This is as good a spot as any to get the peer version
968: # string:
969:
970: if($LondVersion eq "unknown") {
971: $LondVersion = $Socket->PeerVersion();
972: Log("INFO", "Connected to lond version: $LondVersion");
973: }
974: # If necessary, complete a transaction and then go into the
975: # idle queue.
976: # Note that a trasition to idle indicates a live lond
977: # on the other end so reset the connection retries.
978: #
979: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
980: $Watcher->cancel();
981: if(exists($ActiveTransactions{$Socket})) {
982: Debug(5,"Completing transaction!!");
983: CompleteTransaction($Socket,
984: $ActiveTransactions{$Socket});
985: } else {
986: my $count = $Socket->GetClientData();
987: Log("SUCCESS", "Connection ".$count." to "
988: .$RemoteHost." now ready for action");
989: }
990: ServerToIdle($Socket); # Next work unit or idle.
991:
992: #
993: $LondConnecting = 0; # Best spot I can think of for this.
994: #
995:
996: } elsif ($State eq "SendingRequest") {
997: # We need to be writable for this and probably don't belong
998: # here inthe first place.
999:
1000: Debug(6, "SendingRequest state encountered in readable");
1001: $Watcher->poll("w");
1002: $Watcher->cb(\&LondWritable);
1003:
1004: } elsif ($State eq "ReceivingReply") {
1005:
1006:
1007: } else {
1008: # Invalid state.
1009: Debug(4, "Invalid state in LondReadable");
1010: }
1011: }
1012:
1013: =pod
1014:
1015: =head2 LondWritable
1016:
1017: This function is called whenever a lond connection
1018: becomes writable while there is a writeable monitoring
1019: event. The action taken is very state dependent:
1020:
1021: =head3 State = Connected
1022:
1023: The connection is in the process of sending the 'init' hailing to the
1024: lond on the remote end. The Writable member of the connection object
1025: is called. On error, call ConnectionError to destroy the connection
1026: and remove it from the ActiveConnections hash.
1027:
1028: =head3 Initialized
1029:
1030: 'init' has been sent, writability monitoring is removed and
1031: readability monitoring is started with LondReadable as the callback.
1032:
1033: =head3 ChallengeReceived
1034:
1035: The connection has received the who are you challenge from the remote
1036: system, and is in the process of sending the challenge
1037: response. Writable is called.
1038:
1039: =head3 ChallengeReplied
1040:
1041: The connection has replied to the initial challenge The we switch to
1042: monitoring readability looking for the server to reply with 'ok'.
1043:
1044: =head3 RequestingKey
1045:
1046: The connection is in the process of requesting its encryption key.
1047: Writable is called.
1048:
1049: =head3 ReceivingKey
1050:
1051: The connection has sent the request for a key. Switch to readability
1052: monitoring to accept the key
1053:
1054: =head3 SendingRequest
1055:
1056: The connection is in the process of sending a request to the server.
1057: This request is part of a client transaction. All the states until
1058: now represent the client setup protocol. Writable is called.
1059:
1060: =head3 ReceivingReply
1061:
1062: The connection has sent a request. Now it must receive a reply.
1063: Readability monitoring is requested.
1064:
1065: This function is an event handler and therefore receives as
1066: a parameter the event that has fired. The data for the watcher
1067: of this event is a reference to a list of one or two elements,
1068: depending on state. The first (and possibly only) element is the
1069: socket. The second (present only if a request is in progress)
1070: is the socket on which to return a reply to the caller.
1071:
1072: =cut
1073:
1074: sub LondWritable {
1075: my $Event = shift;
1076: my $Watcher = $Event->w;
1077: my $Socket = $Watcher->data;
1078: my $State = $Socket->GetState();
1079:
1080: Debug(6,"LondWritable State = ".$State."\n");
1081:
1082:
1083: # Figure out what to do depending on the state of the socket:
1084:
1085:
1086:
1087:
1088: SocketDump(6,$Socket);
1089:
1090: # If the socket is writable, we must always write.
1091: # Only by writing will we undergo state transitions.
1092: # Old logic wrote in state specific code below, however
1093: # That forces us at least through another invocation of
1094: # this function after writability is possible again.
1095: # This logic also factors out common code for handling
1096: # write failures... in all cases, write failures
1097: # Kill the socket.
1098: # This logic makes the branches of the >big< if below
1099: # so that the writing states are actually NO-OPs.
1100:
1101: if ($Socket->Writable() != 0) {
1102: # The write resulted in an error.
1103: # We'll treat this as if the socket got disconnected:
1104: Log("WARNING", "Connection to ".$RemoteHost.
1105: " has been disconnected");
1106: if(exists($ActiveTransactions{$Socket})) {
1107: FailTransaction($ActiveTransactions{$Socket});
1108: } else {
1109: # In the process of conneting, so need to turn that off.
1110:
1111: $LondConnecting = 0;
1112: }
1113: $Watcher->cancel();
1114: KillSocket($Socket);
1115: return;
1116: }
1117:
1118:
1119:
1120: if ($State eq "Connected") {
1121:
1122: # "init" is being sent...
1123:
1124: } elsif ($State eq "Initialized") {
1125:
1126: # Now that init was sent, we switch
1127: # to watching for readability:
1128:
1129: $Watcher->cb(\&LondReadable);
1130: $Watcher->poll("r");
1131:
1132: } elsif ($State eq "ReInitNoSSL") {
1133:
1134: } elsif ($State eq "ChallengeReceived") {
1135: # We received the challenge, now we
1136: # are echoing it back. This is a no-op,
1137: # we're waiting for the state to change
1138:
1139: } elsif ($State eq "ChallengeReplied") {
1140: # The echo was sent back, so we switch
1141: # to watching readability.
1142:
1143: $Watcher->cb(\&LondReadable);
1144: $Watcher->poll("r");
1145: } elsif ($State eq "RequestingVersion") {
1146: # Sending the peer a version request...
1147:
1148: } elsif ($State eq "ReadingVersionString") {
1149: # Transition to read since we have sent the
1150: # version command and now just need to read the
1151: # version string from the peer:
1152:
1153: $Watcher->cb(\&LondReadable);
1154: $Watcher->poll("r");
1155:
1156: } elsif ($State eq "SetHost") {
1157: # Setting the remote domain...
1158:
1159: } elsif ($State eq "HostSet") {
1160: # Back to readable to get the ok.
1161:
1162: $Watcher->cb(\&LondReadable);
1163: $Watcher->poll("r");
1164:
1165:
1166: } elsif ($State eq "RequestingKey") {
1167: # At this time we're requesting the key.
1168: # again, this is essentially a no-op.
1169:
1170: } elsif ($State eq "ReceivingKey") {
1171: # Now we need to wait for the key
1172: # to come back from the peer:
1173:
1174: $Watcher->cb(\&LondReadable);
1175: $Watcher->poll("r");
1176:
1177: } elsif ($State eq "SendingRequest") {
1178:
1179: # At this time we are sending a request to the
1180: # peer... write the next chunk:
1181:
1182:
1183: } elsif ($State eq "ReceivingReply") {
1184: # The send has completed. Wait for the
1185: # data to come in for a reply.
1186: Debug(8,"Writable sent request/receiving reply");
1187: $Watcher->cb(\&LondReadable);
1188: $Watcher->poll("r");
1189:
1190: } else {
1191: # Control only passes here on an error:
1192: # the socket state does not match any
1193: # of the known states... so an error
1194: # must be logged.
1195:
1196: &Debug(4, "Invalid socket state ".$State."\n");
1197: }
1198:
1199: }
1200:
1201: =pod
1202:
1203: =cut
1204:
1205:
1206: sub QueueDelayed {
1207: Debug(3,"QueueDelayed called");
1208:
1209: my $path = "$perlvar{'lonSockDir'}/delayed";
1210:
1211: Debug(4, "Delayed path: ".$path);
1212: opendir(DIRHANDLE, $path);
1213:
1214: my $host_id_re = '(?:'.join('|',map {quotemeta($_)} (@all_host_ids)).')';
1215: my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
1216: closedir(DIRHANDLE);
1217: foreach my $dfname (sort(@alldelayed)) {
1218: my $reqfile = "$path/$dfname";
1219: my ($host_id) = ($dfname =~ /\.([^.]*)$/);
1220: Debug(4, "queueing ".$reqfile." for $host_id");
1221: my $Handle = IO::File->new($reqfile);
1222: my $cmd = <$Handle>;
1223: chomp $cmd; # There may or may not be a newline...
1224: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1225: my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
1226: $Transaction->SetDeferred($reqfile);
1227: QueueTransaction($Transaction);
1228: }
1229:
1230: }
1231:
1232: =pod
1233:
1234: =head2 MakeLondConnection
1235:
1236: Create a new lond connection object, and start it towards its initial
1237: idleness. Once idle, it becomes eligible to receive transactions
1238: from the work queue. If the work queue is not empty when the
1239: connection is completed and becomes idle, it will dequeue an entry and
1240: start off on it.
1241:
1242: =cut
1243:
1244: sub MakeLondConnection {
1245: my ($restart) = @_;
1246: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1247: .GetServerPort());
1248:
1249: my $Connection = LondConnection->new(&GetServerHost(),
1250: &GetServerPort(),
1251: &GetHostId());
1252:
1253: if($Connection eq undef) {
1254: Log("CRITICAL","Failed to make a connection with lond.");
1255: $ConnectionRetriesLeft--;
1256: return 0; # Failure.
1257: } else {
1258: $LondConnecting = 1; # Connection in progress.
1259: # The connection needs to have writability
1260: # monitored in order to send the init sequence
1261: # that starts the whole authentication/key
1262: # exchange underway.
1263: #
1264: my $Socket = $Connection->GetSocket();
1265: if($Socket eq undef) {
1266: &child_exit(-1, "did not get a socket from the connection");
1267: } else {
1268: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1269: }
1270:
1271: $Connection->SetTimeoutCallback(\&SocketTimeout);
1272:
1273: my $event = Event->io(fd => $Socket,
1274: poll => 'w',
1275: cb => \&LondWritable,
1276: data => $Connection,
1277: desc => 'Connection to lond server');
1278: $ActiveConnections{$Connection} = $event;
1279: if ($ConnectionCount == 0) {
1280: &SetupTimer; # Need to handle timeouts with connections...
1281: }
1282: unless ($restart) {
1283: $ConnectionCount++;
1284: }
1285: $Connection->SetClientData($ConnectionCount);
1286: Debug(4, "Connection count = ".$ConnectionCount);
1287: if($ConnectionCount == 1) { # First Connection:
1288: QueueDelayed;
1289: }
1290: Log("SUCCESS", "Created connection ".$ConnectionCount
1291: ." to host ".GetServerHost());
1292: return 1; # Return success.
1293: }
1294:
1295: }
1296:
1297: =pod
1298:
1299: =head2 StartRequest
1300:
1301: Starts a lond request going on a specified lond connection.
1302: parameters are:
1303:
1304: =item $Lond
1305:
1306: Connection to the lond that will send the transaction and receive the
1307: reply.
1308:
1309: =item $Client
1310:
1311: Connection to the client that is making this request. We got the
1312: request from this socket, and when the request has been relayed to
1313: lond and we get a reply back from lond it will get sent to this
1314: socket.
1315:
1316: =item $Request
1317:
1318: The text of the request to send.
1319:
1320: =cut
1321:
1322: sub StartRequest {
1323:
1324: my ($Lond, $Request) = @_;
1325:
1326: Debug(6, "StartRequest: ".$Request->getRequest());
1327:
1328: my $Socket = $Lond->GetSocket();
1329:
1330: $Request->Activate($Lond);
1331: $ActiveTransactions{$Lond} = $Request;
1332:
1333: $Lond->InitiateTransaction($Request->getRequest());
1334: my $event = Event->io(fd => $Socket,
1335: poll => "w",
1336: cb => \&LondWritable,
1337: data => $Lond,
1338: desc => "lond transaction connection");
1339: $ActiveConnections{$Lond} = $event;
1340: Debug(8," Start Request made watcher data with ".$event->data."\n");
1341: }
1342:
1343: =pod
1344:
1345: =head2 QueueTransaction
1346:
1347: If there is an idle lond connection, it is put to work doing this
1348: transaction. Otherwise, the transaction is placed in the work queue.
1349: If placed in the work queue and the maximum number of connections has
1350: not yet been created, a new connection will be started. Our goal is
1351: to eventually have a sufficient number of connections that the work
1352: queue will typically be empty. parameters are:
1353:
1354: =item Socket
1355:
1356: open on the lonc client.
1357:
1358: =item Request
1359:
1360: data to send to the lond.
1361:
1362: =cut
1363:
1364: sub QueueTransaction {
1365:
1366: my $requestData = shift; # This is a LondTransaction.
1367: my $cmd = $requestData->getRequest();
1368:
1369: Debug(6,"QueueTransaction: ".$cmd);
1370:
1371: my $LondSocket = $IdleConnections->pop();
1372: if(!defined $LondSocket) { # Need to queue request.
1373: Debug(5,"Must queue...");
1374: $WorkQueue->enqueue($requestData);
1375: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1376: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1377:
1378: if($ConnectionRetriesLeft > 0) {
1379: Debug(5,"Starting additional lond connection");
1380: if(&MakeLondConnection() == 0) {
1381: EmptyQueue(); # Fail transactions, can't make connection.
1382: CloseAllLondConnections; # Should all be closed but...
1383: }
1384: } else {
1385: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1386: $LondConnecting = 0;
1387: EmptyQueue(); # It's worse than that ... he's dead Jim.
1388: CloseAllLondConnections; # Should all be closed but..
1389: }
1390: }
1391: } else { # Can start the request:
1392: Debug(8,"Can start...");
1393: StartRequest($LondSocket, $requestData);
1394: }
1395: }
1396:
1397: #-------------------------- Lonc UNIX socket handling -------------------
1398: =pod
1399:
1400: =head2 ClientRequest
1401:
1402: Callback that is called when data can be read from the UNIX domain
1403: socket connecting us with an apache server process.
1404:
1405: =cut
1406:
1407: sub ClientRequest {
1408: Debug(6, "ClientRequest");
1409: my $event = shift;
1410: my $watcher = $event->w;
1411: my $socket = $watcher->fd;
1412: my $data = $watcher->data;
1413: my $thisread;
1414:
1415: Debug(9, " Watcher named: ".$watcher->desc);
1416:
1417: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1418: Debug(8, "rcv: data length = ".length($thisread)
1419: ." read =".$thisread);
1420: unless (defined $rv && length($thisread)) {
1421: # Likely eof on socket.
1422: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1423: close($socket);
1424: $watcher->cancel();
1425: delete($ActiveClients{$socket});
1426: return;
1427: }
1428: Debug(8,"Data: ".$data." this read: ".$thisread);
1429: $data = $data.$thisread; # Append new data.
1430: $watcher->data($data);
1431: if($data =~ /\n$/) { # Request entirely read.
1432: if ($data eq "close_connection_exit\n") {
1433: Log("CRITICAL",
1434: "Request Close Connection ... exiting");
1435: CloseAllLondConnections();
1436: exit;
1437: } elsif ($data eq "reset_retries\n") {
1438: Log("INFO", "Resetting Connection Retries.");
1439: $ConnectionRetriesLeft = $ConnectionRetries;
1440: &UpdateStatus();
1441: my $Transaction = LondTransaction->new($data);
1442: $Transaction->SetClient($socket);
1443: StartClientReply($Transaction, "ok\n");
1444: $watcher->cancel();
1445: return;
1446: }
1447: Debug(8, "Complete transaction received: ".$data);
1448: if ($LogTransactions) {
1449: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1450: }
1451: my $Transaction = LondTransaction->new($data);
1452: $Transaction->SetClient($socket);
1453: QueueTransaction($Transaction);
1454: $watcher->cancel(); # Done looking for input data.
1455: }
1456:
1457: }
1458:
1459: #
1460: # Accept a connection request for a client (lonc child) and
1461: # start up an event watcher to keep an eye on input from that
1462: # Event. This can be called both from NewClient and from
1463: # ChildProcess.
1464: # Parameters:
1465: # $socket - The listener socket.
1466: # Returns:
1467: # NONE
1468: # Side Effects:
1469: # An event is made to watch the accepted connection.
1470: # Active clients hash is updated to reflect the new connection.
1471: # The client connection count is incremented.
1472: #
1473: sub accept_client {
1474: my ($socket) = @_;
1475:
1476: Debug(8, "Entering accept for lonc UNIX socket\n");
1477: my $connection = $socket->accept(); # Accept the client connection.
1478: Debug(8,"Connection request accepted from "
1479: .GetPeername($connection, AF_UNIX));
1480:
1481:
1482: my $description = sprintf("Connection to lonc client %d",
1483: $ClientConnection);
1484: Debug(9, "Creating event named: ".$description);
1485: Event->io(cb => \&ClientRequest,
1486: poll => 'r',
1487: desc => $description,
1488: data => "",
1489: fd => $connection);
1490: $ActiveClients{$connection} = $ClientConnection;
1491: $ClientConnection++;
1492: }
1493:
1494: =pod
1495:
1496: =head2 NewClient
1497:
1498: Callback that is called when a connection is received on the unix
1499: socket for a new client of lonc. The callback is parameterized by the
1500: event.. which is a-priori assumed to be an io event, and therefore has
1501: an fd member that is the Listener socket. We accept the connection
1502: and register a new event on the readability of that socket:
1503:
1504: =cut
1505:
1506: sub NewClient {
1507: Debug(6, "NewClient");
1508: my $event = shift; # Get the event parameters.
1509: my $watcher = $event->w;
1510: my $socket = $watcher->fd; # Get the event' socket.
1511:
1512: &accept_client($socket);
1513: }
1514:
1515: =pod
1516:
1517: =head2 GetLoncSocketPath
1518:
1519: Returns the name of the UNIX socket on which to listen for client
1520: connections.
1521:
1522: =head2 Parameters:
1523:
1524: host (optional) - Name of the host socket to return.. defaults to
1525: the return from GetServerHost().
1526:
1527: =cut
1528:
1529: sub GetLoncSocketPath {
1530:
1531: my $host = GetServerHost(); # Default host.
1532: if (@_) {
1533: ($host) = @_; # Override if supplied.
1534: }
1535: return $UnixSocketDir."/".$host;
1536: }
1537:
1538: =pod
1539:
1540: =head2 GetServerHost
1541:
1542: Returns the host whose lond we talk with.
1543:
1544: =cut
1545:
1546: sub GetServerHost {
1547: return $RemoteHost; # Setup by the fork.
1548: }
1549:
1550: =pod
1551:
1552: =head2 GetServerId
1553:
1554: Returns the hostid whose lond we talk with.
1555:
1556: =cut
1557:
1558: sub GetHostId {
1559: return $RemoteHostId; # Setup by the fork.
1560: }
1561:
1562: =pod
1563:
1564: =head2 GetServerPort
1565:
1566: Returns the lond port number.
1567:
1568: =cut
1569:
1570: sub GetServerPort {
1571: return $perlvar{londPort};
1572: }
1573:
1574: =pod
1575:
1576: =head2 SetupLoncListener
1577:
1578: Setup a lonc listener event. The event is called when the socket
1579: becomes readable.. that corresponds to the receipt of a new
1580: connection. The event handler established will accept the connection
1581: (creating a communications channel), that in turn will establish
1582: another event handler to subess requests.
1583:
1584: =head2 Parameters:
1585:
1586: host (optional) Name of the host to set up a unix socket to.
1587:
1588: =cut
1589:
1590: sub SetupLoncListener {
1591: my ($host,$SocketName) = @_;
1592: if (!$host) { $host = &GetServerHost(); }
1593: if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }
1594:
1595:
1596: unlink($SocketName);
1597:
1598: my $socket;
1599: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1600: Listen => 250,
1601: Type => SOCK_STREAM)) {
1602: if($I_am_child) {
1603: &child_exit(-1, "Failed to create a lonc listener socket");
1604: } else {
1605: die "Failed to create a lonc listner socket";
1606: }
1607: }
1608: return $socket;
1609: }
1610:
1611: #
1612: # Toggle transaction logging.
1613: # Implicit inputs:
1614: # LogTransactions
1615: # Implicit Outputs:
1616: # LogTransactions
1617: sub ToggleTransactionLogging {
1618: print STDERR "Toggle transaction logging...\n";
1619: if(!$LogTransactions) {
1620: $LogTransactions = 1;
1621: } else {
1622: $LogTransactions = 0;
1623: }
1624:
1625:
1626: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1627: }
1628:
1629: =pod
1630:
1631: =head2 ChildStatus
1632:
1633: Child USR1 signal handler to report the most recent status
1634: into the status file.
1635:
1636: We also use this to reset the retries count in order to allow the
1637: client to retry connections with a previously dead server.
1638:
1639: =cut
1640:
1641: sub ChildStatus {
1642: my $event = shift;
1643: my $watcher = $event->w;
1644:
1645: Debug(2, "Reporting child status because : ".$watcher->data);
1646: my $docdir = $perlvar{'lonDocRoot'};
1647:
1648: open(LOG,">>$docdir/lon-status/loncstatus.txt");
1649: flock(LOG,LOCK_EX);
1650: print LOG $$."\t".$RemoteHost."\t".$Status."\t".
1651: $RecentLogEntry."\n";
1652: #
1653: # Write out information about each of the connections:
1654: #
1655: if ($DebugLevel > 2) {
1656: print LOG "Active connection statuses: \n";
1657: my $i = 1;
1658: print STDERR "================================= Socket Status Dump:\n";
1659: foreach my $item (keys %ActiveConnections) {
1660: my $Socket = $ActiveConnections{$item}->data;
1661: my $state = $Socket->GetState();
1662: print LOG "Connection $i State: $state\n";
1663: print STDERR "---------------------- Connection $i \n";
1664: $Socket->Dump(-1); # Ensure it gets dumped..
1665: $i++;
1666: }
1667: }
1668: flock(LOG,LOCK_UN);
1669: close(LOG);
1670: $ConnectionRetriesLeft = $ConnectionRetries;
1671: UpdateStatus();
1672: }
1673:
1674: =pod
1675:
1676: =head2 SignalledToDeath
1677:
1678: Called in response to a signal that causes a chid process to die.
1679:
1680: =cut
1681:
1682:
1683: sub SignalledToDeath {
1684: my $event = shift;
1685: my $watcher= $event->w;
1686:
1687: Debug(2,"Signalled to death! via ".$watcher->data);
1688: my ($signal) = $watcher->data;
1689: chomp($signal);
1690: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1691: ."died through "."\"$signal\"");
1692: #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1693: # ."\"$signal\"");
1694: &clear_childpid($$);
1695: exit 0;
1696:
1697: }
1698:
1699: =pod
1700:
1701: =head2 ToggleDebug
1702:
1703: This sub toggles trace debugging on and off.
1704:
1705: =cut
1706:
1707: sub ToggleDebug {
1708: my $Current = $DebugLevel;
1709: $DebugLevel = $NextDebugLevel;
1710: $NextDebugLevel = $Current;
1711:
1712: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1713:
1714: }
1715:
1716: =pod
1717:
1718: =head2 ChildProcess
1719:
1720: This sub implements a child process for a single lonc daemon.
1721: Optional parameter:
1722: $socket - if provided, this is a socket already open for listening
1723: on the client socket. Otherwise, a new listener is set up.
1724:
1725: =cut
1726:
1727: sub ChildProcess {
1728: # We've inherited all the
1729: # events of our parent and those have to be cancelled or else
1730: # all holy bloody chaos will result.. trust me, I already made
1731: # >that< mistake.
1732:
1733: my $host = GetServerHost();
1734: foreach my $listener (keys %parent_dispatchers) {
1735: my $watcher = $parent_dispatchers{$listener};
1736: my $s = $watcher->fd;
1737: if ($listener ne $host) { # Close everyone but me.
1738: Debug(5, "Closing listen socket for $listener");
1739: $s->close();
1740: }
1741: Debug(5, "Killing watcher for $listener");
1742:
1743: $watcher->cancel();
1744: delete($parent_dispatchers{$listener});
1745:
1746: }
1747:
1748: # kill off the parent's signal handlers too!
1749: #
1750:
1751: for my $handler (keys %parent_handlers) {
1752: my $watcher = $parent_handlers{$handler};
1753: $watcher->cancel();
1754: delete($parent_handlers{$handler});
1755: }
1756:
1757: $I_am_child = 1; # Seems like in spite of it all I may still getting
1758: # parent event dispatches.. flag I'm a child.
1759:
1760:
1761: #
1762: # Signals must be handled by the Event framework...
1763: #
1764:
1765: Event->signal(signal => "QUIT",
1766: cb => \&SignalledToDeath,
1767: data => "QUIT");
1768: Event->signal(signal => "HUP",
1769: cb => \&ChildStatus,
1770: data => "HUP");
1771: Event->signal(signal => "USR1",
1772: cb => \&ChildStatus,
1773: data => "USR1");
1774: Event->signal(signal => "USR2",
1775: cb => \&ToggleTransactionLogging);
1776: Event->signal(signal => "INT",
1777: cb => \&ToggleDebug,
1778: data => "INT");
1779:
1780: # Block the pipe signal we'll get when the socket disconnects. We detect
1781: # socket disconnection via send/receive failures. On disconnect, the
1782: # socket becomes readable .. which will force the disconnect detection.
1783:
1784: my $set = POSIX::SigSet->new(SIGPIPE);
1785: sigprocmask(SIG_BLOCK, $set);
1786:
1787: # Figure out if we got passed a socket or need to open one to listen for
1788: # client requests.
1789:
1790: my ($socket) = @_;
1791: if (!$socket) {
1792:
1793: $socket = SetupLoncListener();
1794: }
1795: # Establish an event to listen for client connection requests.
1796:
1797:
1798: Event->io(cb => \&NewClient,
1799: poll => 'r',
1800: desc => 'Lonc Listener Unix Socket',
1801: fd => $socket);
1802:
1803: $Event::DebugLevel = $DebugLevel;
1804:
1805: Debug(9, "Making initial lond connection for ".$RemoteHost);
1806:
1807: # Setup the initial server connection:
1808:
1809: # &MakeLondConnection(); // let first work request do it.
1810:
1811: # need to accept the connection since the event may not fire.
1812:
1813: &accept_client($socket);
1814:
1815: Debug(9,"Entering event loop");
1816: my $ret = Event::loop(); # Start the main event loop.
1817:
1818:
1819: &child_exit (-1,"Main event loop exited!!!");
1820: }
1821:
1822: # Create a new child for host passed in:
1823:
1824: sub CreateChild {
1825: my ($host, $hostid) = @_;
1826:
1827: my $sigset = POSIX::SigSet->new(SIGINT);
1828: sigprocmask(SIG_BLOCK, $sigset);
1829: $RemoteHost = $host;
1830: ShowStatus('Parent keeping the flock'); # Update time in status message.
1831: Log("CRITICAL", "Forking server for ".$host);
1832: my $pid = fork;
1833: if($pid) { # Parent
1834: $RemoteHost = "Parent";
1835: $ChildPid{$pid} = $host;
1836: sigprocmask(SIG_UNBLOCK, $sigset);
1837: undef(@all_host_ids);
1838: } else { # child.
1839: $RemoteHostId = $hostid;
1840: ShowStatus("Connected to ".$RemoteHost);
1841: $SIG{INT} = 'DEFAULT';
1842: sigprocmask(SIG_UNBLOCK, $sigset);
1843: &ChildProcess(); # Does not return.
1844: }
1845: }
1846:
1847: # parent_client_connection:
1848: # Event handler that processes client connections for the parent process.
1849: # This sub is called when the parent is listening on a socket and
1850: # a connection request arrives. We must:
1851: # Start a child process to accept the connection request.
1852: # Kill our listen on the socket.
1853: # Parameter:
1854: # event - The event object that was created to monitor this socket.
1855: # event->w->fd is the socket.
1856: # Returns:
1857: # NONE
1858: #
1859: sub parent_client_connection {
1860: if ($I_am_child) {
1861: # Should not get here, but seem to anyway:
1862: &Debug(5," Child caught parent client connection event!!");
1863: my ($event) = @_;
1864: my $watcher = $event->w;
1865: $watcher->cancel(); # Try to kill it off again!!
1866: } else {
1867: &Debug(9, "parent_client_connection");
1868: my ($event) = @_;
1869: my $watcher = $event->w;
1870: my $socket = $watcher->fd;
1871: my $connection = $socket->accept(); # Accept the client connection.
1872: Event->io(cb => \&get_remote_hostname,
1873: poll => 'r',
1874: data => "",
1875: fd => $connection);
1876: }
1877: }
1878:
1879: sub get_remote_hostname {
1880: my ($event) = @_;
1881: my $watcher = $event->w;
1882: my $socket = $watcher->fd;
1883:
1884: my $thisread;
1885: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1886: Debug(8, "rcv: data length = ".length($thisread)." read =".$thisread);
1887: if (!defined($rv) || length($thisread) == 0) {
1888: # Likely eof on socket.
1889: Debug(5,"Client Socket closed on lonc for p_c_c");
1890: close($socket);
1891: $watcher->cancel();
1892: return;
1893: }
1894:
1895: my $data = $watcher->data().$thisread;
1896: $watcher->data($data);
1897: if($data =~ /\n$/) { # Request entirely read.
1898: chomp($data);
1899: } else {
1900: return;
1901: }
1902:
1903: &Debug(5,"Creating child for $data (parent_client_connection)");
1904: (my $hostname,my $lonid,@all_host_ids) = split(':',$data);
1905: $ChildHost{$hostname}++;
1906: if ($ChildHost{$hostname} == 1) {
1907: &CreateChild($hostname,$lonid);
1908: } else {
1909: &Log('WARNING',"Request for a second child on $hostname");
1910: }
1911: # Clean up the listen since now the child takes over until it exits.
1912: $watcher->cancel(); # Nolonger listening to this event
1913: $socket->send("done\n");
1914: $socket->close();
1915: }
1916:
1917: # parent_listen:
1918: # Opens a socket and starts a listen for the parent process on a client UNIX
1919: # domain socket.
1920: #
1921: # This involves:
1922: # Creating a socket for listen.
1923: # Removing any socket lock file
1924: # Adding an event handler for this socket becoming readable
1925: # To the parent's event dispatcher.
1926: # Parameters:
1927: # loncapa_host - LonCAPA cluster name of the host represented by the client
1928: # socket.
1929: # Returns:
1930: # NONE
1931: #
1932: sub parent_listen {
1933: my ($loncapa_host) = @_;
1934: Debug(5, "parent_listen: $loncapa_host");
1935:
1936: my ($socket,$file);
1937: if (!$loncapa_host) {
1938: $loncapa_host = 'common_parent';
1939: $file = $perlvar{'lonSockCreate'};
1940: } else {
1941: $file = &GetLoncSocketPath($loncapa_host);
1942: }
1943: $socket = &SetupLoncListener($loncapa_host,$file);
1944:
1945: $listening_to{$socket} = $loncapa_host;
1946: if (!$socket) {
1947: die "Unable to create a listen socket for $loncapa_host";
1948: }
1949:
1950: my $lock_file = $file.".lock";
1951: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1952:
1953: my $watcher =
1954: Event->io(cb => \&parent_client_connection,
1955: poll => 'r',
1956: desc => "Parent listener unix socket ($loncapa_host)",
1957: data => "",
1958: fd => $socket);
1959: $parent_dispatchers{$loncapa_host} = $watcher;
1960:
1961: }
1962:
1963: sub parent_clean_up {
1964: my ($loncapa_host) = @_;
1965: Debug(1, "parent_clean_up: $loncapa_host");
1966:
1967: my $socket_file = &GetLoncSocketPath($loncapa_host);
1968: unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
1969: my $lock_file = $socket_file.".lock";
1970: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1971: }
1972:
1973:
1974:
1975: # This sub initiates a listen on the common unix domain lonc client socket.
1976: # loncnew starts up with no children, and only spawns off children when a
1977: # connection request occurs on the common client unix socket. The spawned
1978: # child continues to run until it has been idle a while at which point it
1979: # eventually exits and once more the parent picks up the listen.
1980: #
1981: # Parameters:
1982: # NONE
1983: # Implicit Inputs:
1984: # The configuration file that has been read in by LondConnection.
1985: # Returns:
1986: # NONE
1987: #
1988: sub listen_on_common_socket {
1989: Debug(5, "listen_on_common_socket");
1990: &parent_listen();
1991: }
1992:
1993: # server_died is called whenever a child process exits.
1994: # Since this is dispatched via a signal, we must process all
1995: # dead children until there are no more left. The action
1996: # is to:
1997: # - Remove the child from the bookeeping hashes
1998: # - Re-establish a listen on the unix domain socket associated
1999: # with that host.
2000: # Parameters:
2001: # The event, but we don't actually care about it.
2002: sub server_died {
2003: &Debug(9, "server_died called...");
2004:
2005: while(1) { # Loop until waitpid nowait fails.
2006: my $pid = waitpid(-1, WNOHANG);
2007: if($pid <= 0) {
2008: return; # Nothing left to wait for.
2009: }
2010: # need the host to restart:
2011:
2012: my $host = $ChildPid{$pid};
2013: if($host) { # It's for real...
2014: &Debug(9, "Caught sigchild for $host");
2015: &clear_childpid($pid);
2016: delete($ChildPid{$pid});
2017: delete($ChildHost{$host});
2018: &parent_clean_up($host);
2019:
2020: } else {
2021: &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
2022: }
2023: }
2024:
2025: }
2026:
2027: #
2028: # Parent process logic pass 1:
2029: # For each entry in the hosts table, we will
2030: # fork off an instance of ChildProcess to service the transactions
2031: # to that host. Each pid will be entered in a global hash
2032: # with the value of the key, the host.
2033: # The parent will then enter a loop to wait for process exits.
2034: # Each exit gets logged and the child gets restarted.
2035: #
2036:
2037: #
2038: # Fork and start in new session so hang-up isn't going to
2039: # happen without intent.
2040: #
2041:
2042:
2043:
2044:
2045:
2046:
2047: ShowStatus("Forming new session");
2048: my $childpid = fork;
2049: if ($childpid != 0) {
2050: sleep 4; # Give child a chacne to break to
2051: exit 0; # a new sesion.
2052: }
2053: #
2054: # Write my pid into the pid file so I can be located
2055: #
2056:
2057: ShowStatus("Parent writing pid file:");
2058: my $execdir = $perlvar{'lonDaemons'};
2059: open (PIDSAVE, ">$execdir/logs/lonc.pid");
2060: print PIDSAVE "$$\n";
2061: close(PIDSAVE);
2062:
2063:
2064:
2065: if (POSIX::setsid() < 0) {
2066: print "Could not create new session\n";
2067: exit -1;
2068: }
2069:
2070: ShowStatus("Forking node servers");
2071:
2072: Log("CRITICAL", "--------------- Starting children ---------------");
2073:
2074: LondConnection::ReadConfig; # Read standard config files.
2075:
2076: $RemoteHost = "[parent]";
2077: &listen_on_common_socket();
2078:
2079: $RemoteHost = "Parent Server";
2080:
2081: # Maintain the population:
2082:
2083: ShowStatus("Parent keeping the flock");
2084:
2085:
2086: # We need to setup a SIGChild event to handle the exit (natural or otherwise)
2087: # of the children.
2088:
2089: Event->signal(cb => \&server_died,
2090: desc => "Child exit handler",
2091: signal => "CHLD");
2092:
2093:
2094: # Set up all the other signals we set up.
2095:
2096: $parent_handlers{INT} = Event->signal(cb => \&Terminate,
2097: desc => "Parent INT handler",
2098: signal => "INT");
2099: $parent_handlers{TERM} = Event->signal(cb => \&Terminate,
2100: desc => "Parent TERM handler",
2101: signal => "TERM");
2102: $parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
2103: desc => "Parent HUP handler.",
2104: signal => "HUP");
2105: $parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
2106: desc => "Parent USR1 handler",
2107: signal => "USR1");
2108: $parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
2109: desc => "Parent USR2 handler.",
2110: signal => "USR2");
2111:
2112: # Start procdesing events.
2113:
2114: $Event::DebugLevel = $DebugLevel;
2115: Debug(9, "Parent entering event loop");
2116: my $ret = Event::loop();
2117: die "Main Event loop exited: $ret";
2118:
2119: =pod
2120:
2121: =head1 CheckKids
2122:
2123: Since kids do not die as easily in this implementation
2124: as the previous one, there is no need to restart the
2125: dead ones (all dead kids get restarted when they die!!)
2126: The only thing this function does is to pass USR1 to the
2127: kids so that they report their status.
2128:
2129: =cut
2130:
2131: sub CheckKids {
2132: Debug(2, "Checking status of children");
2133: my $docdir = $perlvar{'lonDocRoot'};
2134: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
2135: my $now=time;
2136: my $local=localtime($now);
2137: print $fh "LONC status $local - parent $$ \n\n";
2138: foreach my $host (keys %parent_dispatchers) {
2139: print $fh "LONC Parent process listening for $host\n";
2140: }
2141: foreach my $pid (keys %ChildPid) {
2142: Debug(2, "Sending USR1 -> $pid");
2143: kill 'USR1' => $pid; # Tell Child to report status.
2144: }
2145:
2146: }
2147:
2148: =pod
2149:
2150: =head1 UpdateKids
2151:
2152: parent's SIGUSR2 handler. This handler:
2153:
2154: =item
2155:
2156: Rereads the hosts file.
2157:
2158: =item
2159:
2160: Kills off (via sigint) children for hosts that have disappeared.
2161:
2162: =item
2163:
2164: QUITs children for hosts that already exist (this just forces a status display
2165: and resets the connection retry count for that host.
2166:
2167: =item
2168:
2169: Starts new children for hosts that have been added to the hosts.tab file since
2170: the start of the master program and maintains them.
2171:
2172: =cut
2173:
2174: sub UpdateKids {
2175:
2176: Log("INFO", "Updating connections via SIGUSR2");
2177:
2178: # I'm not sure what I was thinking in the first implementation.
2179: # someone will have to work hard to convince me the effect is any
2180: # different than Restart, especially now that we don't start up
2181: # per host servers automatically, may as well just restart.
2182: # The down side is transactions that are in flight will get timed out
2183: # (lost unless they are critical).
2184:
2185: &KillThemAll();
2186: LondConnection->ResetReadConfig();
2187: ShowStatus('Parent keeping the flock');
2188: }
2189:
2190:
2191: =pod
2192:
2193: =head1 Restart
2194:
2195: Signal handler for HUP... all children are killed and
2196: we self restart. This is an el-cheapo way to re-read
2197: the config file.
2198:
2199: =cut
2200:
2201: sub Restart {
2202: &KillThemAll; # First kill all the children.
2203: LondConnection->ResetReadConfig();
2204: Log("CRITICAL", "Restarting");
2205: my $execdir = $perlvar{'lonDaemons'};
2206: unlink("$execdir/logs/lonc.pid");
2207: exec("$executable");
2208: }
2209:
2210: =pod
2211:
2212: =head1 KillThemAll
2213:
2214: Signal handler that kills all children by sending them a
2215: SIGHUP. Responds to sigint and sigterm.
2216:
2217: =cut
2218:
2219: sub KillThemAll {
2220: Debug(2, "Kill them all!!");
2221:
2222: #local($SIG{CHLD}) = 'IGNORE';
2223: # Our children >will< die.
2224: # but we need to catch their death and cleanup after them in case this is
2225: # a restart set of kills
2226: my $execdir = $perlvar{'lonDaemons'};
2227: my @allpids = keys(%ChildPid);
2228: foreach my $pid (@allpids) {
2229: my $serving = $ChildPid{$pid};
2230: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2231: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
2232: kill 'QUIT' => $pid;
2233: &clear_childpid($pid);
2234: }
2235: ShowStatus("Finished killing child processes off.");
2236: }
2237:
2238:
2239: #
2240: # Kill all children via KILL. Just in case the
2241: # first shot didn't get them.
2242:
2243: sub really_kill_them_all_dammit
2244: {
2245: Debug(2, "Kill them all Dammit");
2246: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
2247: foreach my $pid (keys %ChildPid) {
2248: my $serving = $ChildPid{$pid};
2249: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2250: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2251: kill 'KILL' => $pid;
2252: delete($ChildPid{$pid});
2253: delete($ChildKeyMode{$pid});
2254: my $execdir = $perlvar{'lonDaemons'};
2255: unlink("$execdir/logs/lonc.pid");
2256: }
2257: }
2258:
2259: =pod
2260:
2261: =head1 Terminate
2262:
2263: Terminate the system.
2264:
2265: =cut
2266:
2267: sub Terminate {
2268: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2269: &KillThemAll;
2270: #
2271: # By now they really should all be dead.. but just in case
2272: # send them all SIGKILL's after a bit of waiting:
2273:
2274: sleep(4);
2275: &Log("CRITICAL", "Now kill children nasty");
2276: &really_kill_them_all_dammit;
2277: Log("CRITICAL","Master process exiting");
2278: exit 0;
2279:
2280: }
2281:
2282: =pod
2283:
2284: =cut
2285:
2286: sub my_hostname {
2287: use Sys::Hostname::FQDN();
2288: my $name = Sys::Hostname::FQDN::fqdn();
2289: &Debug(9,"Name is $name");
2290: return $name;
2291: }
2292:
2293: sub record_childpid {
2294: my ($Socket) = @_;
2295: my $docdir = $perlvar{'lonDocRoot'};
2296: my $authmode = $Socket->GetKeyMode();
2297: my $peer = $Socket->PeerLoncapaHim();
2298: if (($authmode eq 'ssl') || ($authmode eq 'insecure')) {
2299: my $childpid = $$;
2300: if ($childpid) {
2301: unless (exists($ChildKeyMode{$childpid})) {
2302: $ChildKeyMode{$childpid} = $authmode;
2303: }
2304: if (-d "$docdir/lon-status/loncchld") {
2305: unless (-e "$docdir/lon-status/loncchld/$childpid") {
2306: if (open (my $pidfh,'>',"$docdir/lon-status/loncchld/$childpid")) {
2307: print $pidfh "$peer:$authmode\n";
2308: close($pidfh);
2309: }
2310: }
2311: }
2312: }
2313: }
2314: return;
2315: }
2316:
2317: sub clear_childpid {
2318: my ($childpid) = @_;
2319: my $docdir = $perlvar{'lonDocRoot'};
2320: if (-d "$docdir/lon-status/loncchld") {
2321: if ($childpid =~ /^\d+$/) {
2322: if (($ChildKeyMode{$childpid} eq 'insecure') ||
2323: ($ChildKeyMode{$childpid} eq 'ssl')) {
2324: if (-e "$docdir/lon-status/loncchld/$childpid") {
2325: unlink("$docdir/lon-status/loncchld/$childpid");
2326: }
2327: }
2328: }
2329: }
2330: if (exists($ChildKeyMode{$childpid})) {
2331: delete($ChildKeyMode{$childpid});
2332: }
2333: return;
2334: }
2335:
2336: =pod
2337:
2338: =head1 Theory
2339:
2340: The event class is used to build this as a single process with an
2341: event driven model. The following events are handled:
2342:
2343: =item UNIX Socket connection Received
2344:
2345: =item Request data arrives on UNIX data transfer socket.
2346:
2347: =item lond connection becomes writable.
2348:
2349: =item timer fires at 1 second intervals.
2350:
2351: All sockets are run in non-blocking mode. Timeouts managed by the timer
2352: handler prevents hung connections.
2353:
2354: Key data structures:
2355:
2356: =item RequestQueue
2357:
2358: A queue of requests received from UNIX sockets that are
2359: waiting for a chance to be forwarded on a lond connection socket.
2360:
2361: =item ActiveConnections
2362:
2363: A hash of lond connections that have transactions in process that are
2364: available to be timed out.
2365:
2366: =item ActiveTransactions
2367:
2368: A hash indexed by lond connections that contain the client reply
2369: socket for each connection that has an active transaction on it.
2370:
2371: =item IdleConnections
2372:
2373: A hash of lond connections that have no work to do. These connections
2374: can be closed if they are idle for a long enough time.
2375:
2376: =cut
2377:
2378: =pod
2379:
2380: =head1 Log messages
2381:
2382: The following is a list of log messages that can appear in the
2383: lonc.log file. Each log file has a severity and a message.
2384:
2385: =over 2
2386:
2387: =item Warning A socket timeout was detected
2388:
2389: If there are pending transactions in the socket's queue,
2390: they are failed (saved if critical). If the connection
2391: retry count gets exceeded by this, the
2392: remote host is marked as dead.
2393: Called when timeouts occurred during the connection and
2394: connection dialog with a remote host.
2395:
2396: =item Critical Host makred DEAD <hostname>
2397:
2398: The numer of retry counts for contacting a host was
2399: exceeded. The host is marked dead an no
2400: further attempts will be made by that child.
2401:
2402: =item Info lonc pipe client hung up on us
2403:
2404: Write to the client pipe indicated no data transferred
2405: Socket to remote host is shut down. Reply to the client
2406: is discarded. Note: This is commented out in &ClientWriteable
2407:
2408: =item Success Reply from lond: <data>
2409:
2410: Can be enabled for debugging by setting LogTransactions to nonzero.
2411: Indicates a successful transaction with lond, <data> is the data received
2412: from the remote lond.
2413:
2414: =item Success A delayed transaction was completed
2415:
2416: A transaction that must be reliable was executed and completed
2417: as lonc restarted. This is followed by a mesage of the form
2418:
2419: S: client-name : request
2420:
2421: =item WARNING Failing transaction <cmd>:<subcmd>
2422:
2423: Transaction failed on a socket, but the failure retry count for the remote
2424: node has not yet been exhausted (the node is not yet marked dead).
2425: cmd is the command, subcmd is the subcommand. This results from a con_lost
2426: when communicating with lond.
2427:
2428: =item WARNING Shutting down a socket
2429:
2430: Called when a socket is being closed to lond. This is emitted both when
2431: idle pruning is being done and when the socket has been disconnected by the remote.
2432:
2433: =item WARNING Lond connection lost.
2434:
2435: Called when a read from lond's socket failed indicating lond has closed the
2436: connection or died. This should be followed by one or more
2437:
2438: "WARNING Failing transaction..." msgs for each in-flight or queued transaction.
2439:
2440: =item WARNING No SSL channel (verification failed), will try with insecure channel.
2441:
2442: Called when promotion of a socket to SSL failed because SSL certificate verification failed.
2443: Domain configuration must also permit insecure channel use for key exchange. Connection
2444: negotiation will start again from the beginning, but with Authentication Mode not set to ssl.
2445:
2446: =item INFO Connected to lond version: <version>
2447:
2448: When connection negotiation is complete, the lond version is requested and logged here.
2449:
2450: =item SUCCESS Connection n to host now ready for action
2451:
2452: Emitted when connection has been completed with lond. n is then number of
2453: concurrent connections and host, the host to which the connection has just
2454: been established.
2455:
2456: =item WARNING Connection to host has been disconnected
2457:
2458: Write to a lond resulted in failure status. Connection to lond is dropped.
2459:
2460: =item SUCCESS Created connection n to host host
2461:
2462: Initial connection request to host..(before negotiation).
2463:
2464: =item CRITICAL Request Close Connection ... exiting
2465:
2466: Client has sent "close_connection_exit" The loncnew server is exiting.
2467:
2468: =item INFO Resetting Connection Retries
2469:
2470: Client has sent "reset_retries" The lond connection retries are reset to zero for the
2471: corresponding lond.
2472:
2473: =item SUCCESS Transaction <data>
2474:
2475: Only emitted if the global variable $LogTransactions was set to true.
2476: A client has requested a lond transaction <data> is the contents of the request.
2477:
2478: =item SUCCESS Toggled transaction logging <LogTransactions>
2479:
2480: The state of the $LogTransactions global has been toggled, and its current value
2481: (after being toggled) is displayed. When non zero additional logging of transactions
2482: is enabled for debugging purposes. Transaction logging is toggled on receipt of a USR2
2483: signal.
2484:
2485: =item CRITICAL Abnormal exit. Child <pid> for <host> died thorugh signal.
2486:
2487: QUIT signal received. lonc child process is exiting.
2488:
2489: =item SUCCESS New debugging level for <RemoteHost> now <DebugLevel>
2490:
2491: Debugging toggled for the host loncnew is talking with.
2492: Currently debugging is a level based scheme with higher number
2493: conveying more information. The daemon starts out at
2494: DebugLevel 0 and can toggle back and forth between that and
2495: DebugLevel 2 These are controlled by
2496: the global variables $DebugLevel and $NextDebugLevel
2497: The debug level can go up to 9.
2498: SIGINT toggles the debug level. The higher the debug level the
2499: more debugging information is spewed. See the Debug
2500: sub in loncnew.
2501:
2502: =item CRITICAL Forking server for host
2503:
2504: A child is being created to service requests for the specified host.
2505:
2506:
2507: =item WARNING Request for a second child on hostname
2508:
2509: Somehow loncnew was asked to start a second child on a host that already had a child
2510: servicing it. This request is not honored, but themessage is emitted. This could happen
2511: due to a race condition. When a client attempts to contact loncnew for a new host, a child
2512: is forked off to handle the requests for that server. The parent then backs off the Unix
2513: domain socket leaving it for the child to service all requests. If in the time between
2514: creating the child, and backing off, a new connection request comes in to the unix domain
2515: socket, this could trigger (unlikely but remotely possible),.
2516:
2517: =item CRITICAL ------ Starting Children ----
2518:
2519: This message should probably be changed to "Entering event loop" as the loncnew only starts
2520: children as needed. This message is emitted as new events are established and
2521: the event processing loop is entered.
2522:
2523: =item INFO Updating connections via SIGUSR2
2524:
2525: SIGUSR2 received. The original code would kill all clients, re-read the host file,
2526: then restart children for each host. Now that children are started on demand, this
2527: just kills all child processes and lets requests start them as needed again.
2528:
2529:
2530: =item CRITICAL Restarting
2531:
2532: SigHUP received. all the children are killed and the script exec's itself to start again.
2533:
2534: =item CRITICAL Nicely killing lonc for host pid = <pid>
2535:
2536: Attempting to kill the child that is serving the specified host (pid given) cleanly via
2537: SIGQUIT. The child should handle that, clean up nicely and exit.
2538:
2539: =item CRITICAL Nastily killing lonc for host pid = <pid>
2540:
2541: The child specified did not die when requested via SIGQUIT. Therefore it is killed
2542: via SIGKILL.
2543:
2544: =item CRITICAL Asked to kill children.. first be nice..
2545:
2546: In the parent's INT handler. INT kills the child processes. This inidicate loncnew
2547: is about to attempt to kill all known children via SIGQUIT. This message should be followed
2548: by one "Nicely killing" message for each extant child.
2549:
2550: =item CRITICAL Now kill children nasty
2551:
2552: In the parent's INT handler. remaining children are about to be killed via
2553: SIGKILL. Should be followed by a Nastily killing... for each lonc child that
2554: refused to die.
2555:
2556: =item CRITICAL Master process exiting
2557:
2558: In the parent's INT handler. just prior to the exit 0 call.
2559:
2560: =back
2561:
2562: =cut
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>