Annotation of loncom/loncnew, revision 1.32
1.1 foxr 1: #!/usr/bin/perl
1.2 albertel 2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
1.32 ! albertel 5: # $Id: loncnew,v 1.31 2003/10/28 10:48:10 foxr Exp $
1.2 albertel 6: #
7: # Copyright Michigan State University Board of Trustees
8: #
9: # This file is part of the LearningOnline Network with CAPA (LON-CAPA).
1.17 foxr 10: ## LON-CAPA is free software; you can redistribute it and/or modify
1.2 albertel 11: # it under the terms of the GNU General Public License as published by
12: # the Free Software Foundation; either version 2 of the License, or
13: # (at your option) any later version.
14: #
15: # LON-CAPA is distributed in the hope that it will be useful,
16: # but WITHOUT ANY WARRANTY; without even the implied warranty of
17: # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18: # GNU General Public License for more details.
19: #
20: # You should have received a copy of the GNU General Public License
21: # along with LON-CAPA; if not, write to the Free Software
22: # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23: #
24: # /home/httpd/html/adm/gpl.txt
25: #
26: # http://www.lon-capa.org/
27: #
1.1 foxr 28: #
1.15 foxr 29: # new lonc handles n request out bver m connections to londs.
1.1 foxr 30: # This module is based on the Event class.
31: # Development iterations:
32: # - Setup basic event loop. (done)
33: # - Add timer dispatch. (done)
34: # - Add ability to accept lonc UNIX domain sockets. (done)
35: # - Add ability to create/negotiate lond connections (done).
1.7 foxr 36: # - Add general logic for dispatching requests and timeouts. (done).
37: # - Add support for the lonc/lond requests. (done).
1.1 foxr 38: # - Add logging/status monitoring.
39: # - Add Signal handling - HUP restarts. USR1 status report.
1.7 foxr 40: # - Add Configuration file I/O (done).
1.1 foxr 41: # - Add management/status request interface.
1.8 foxr 42: # - Add deferred request capability. (done)
1.9 foxr 43: # - Detect transmission timeouts.
1.7 foxr 44: #
45:
1.23 foxr 46: use strict;
1.1 foxr 47: use lib "/home/httpd/lib/perl/";
48: use Event qw(:DEFAULT );
49: use POSIX qw(:signal_h);
1.12 foxr 50: use POSIX;
1.1 foxr 51: use IO::Socket;
52: use IO::Socket::INET;
53: use IO::Socket::UNIX;
1.9 foxr 54: use IO::File;
1.6 foxr 55: use IO::Handle;
1.1 foxr 56: use Socket;
57: use Crypt::IDEA;
58: use LONCAPA::Queue;
59: use LONCAPA::Stack;
60: use LONCAPA::LondConnection;
1.7 foxr 61: use LONCAPA::LondTransaction;
1.1 foxr 62: use LONCAPA::Configuration;
63: use LONCAPA::HashIterator;
64:
65:
66: #
67: # Disable all signals we might receive from outside for now.
68: #
69:
70:
71: # Read the httpd configuration file to get perl variables
72: # normally set in apache modules:
73:
74: my $perlvarref = LONCAPA::Configuration::read_conf('loncapa.conf');
75: my %perlvar = %{$perlvarref};
76:
77: #
78: # parent and shared variables.
79:
80: my %ChildHash; # by pid -> host.
1.26 foxr 81: my %HostToPid; # By host -> pid.
82: my %HostHash; # by loncapaname -> IP.
1.1 foxr 83:
84:
1.9 foxr 85: my $MaxConnectionCount = 10; # Will get from config later.
1.1 foxr 86: my $ClientConnection = 0; # Uniquifier for client events.
87:
1.9 foxr 88: my $DebugLevel = 0;
1.29 foxr 89: my $NextDebugLevel= 2; # So Sigint can toggle this.
1.1 foxr 90: my $IdleTimeout= 3600; # Wait an hour before pruning connections.
91:
92: #
93: # The variables below are only used by the child processes.
94: #
95: my $RemoteHost; # Name of host child is talking to.
1.20 albertel 96: my $UnixSocketDir= $perlvar{'lonSockDir'};
1.1 foxr 97: my $IdleConnections = Stack->new(); # Set of idle connections
98: my %ActiveConnections; # Connections to the remote lond.
1.7 foxr 99: my %ActiveTransactions; # LondTransactions in flight.
1.1 foxr 100: my %ActiveClients; # Serial numbers of active clients by socket.
101: my $WorkQueue = Queue->new(); # Queue of pending transactions.
102: my $ConnectionCount = 0;
1.4 foxr 103: my $IdleSeconds = 0; # Number of seconds idle.
1.9 foxr 104: my $Status = ""; # Current status string.
1.14 foxr 105: my $RecentLogEntry = "";
1.30 foxr 106: my $ConnectionRetries=2; # Number of connection retries allowed.
107: my $ConnectionRetriesLeft=2; # Number of connection retries remaining.
1.1 foxr 108:
109: #
1.9 foxr 110: # The hash below gives the HTML format for log messages
111: # given a severity.
112: #
113: my %LogFormats;
114:
115: $LogFormats{"CRITICAL"} = "<font color=red>CRITICAL: %s</font>";
116: $LogFormats{"SUCCESS"} = "<font color=green>SUCCESS: %s</font>";
117: $LogFormats{"INFO"} = "<font color=yellow>INFO: %s</font>";
118: $LogFormats{"WARNING"} = "<font color=blue>WARNING: %s</font>";
119: $LogFormats{"DEFAULT"} = " %s ";
120:
1.10 foxr 121:
122:
123: =pod
124:
125: =head2 LogPerm
126:
127: Makes an entry into the permanent log file.
128:
129: =cut
130: sub LogPerm {
131: my $message=shift;
132: my $execdir=$perlvar{'lonDaemons'};
133: my $now=time;
134: my $local=localtime($now);
135: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
136: print $fh "$now:$message:$local\n";
137: }
1.9 foxr 138:
139: =pod
140:
141: =head2 Log
142:
143: Logs a message to the log file.
144: Parameters:
145:
146: =item severity
147:
148: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
149: format string used to format the message. if the severity is
150: not a defined severity the Default format string is used.
151:
152: =item message
153:
154: The base message. In addtion to the format string, the message
155: will be appended to a string containing the name of our remote
156: host and the time will be formatted into the message.
157:
158: =cut
159:
160: sub Log {
161: my $severity = shift;
162: my $message = shift;
163:
164: if(!$LogFormats{$severity}) {
165: $severity = "DEFAULT";
166: }
167:
168: my $format = $LogFormats{$severity};
169:
170: # Put the window dressing in in front of the message format:
171:
172: my $now = time;
173: my $local = localtime($now);
174: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
175: my $finalformat = $finalformat.$format."\n";
176:
177: # open the file and put the result.
178:
179: my $execdir = $perlvar{'lonDaemons'};
180: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
181: my $msg = sprintf($finalformat, $message);
1.14 foxr 182: $RecentLogEntry = $msg;
1.9 foxr 183: print $fh $msg;
184:
1.10 foxr 185:
1.9 foxr 186: }
1.6 foxr 187:
1.3 albertel 188:
1.1 foxr 189: =pod
1.3 albertel 190:
191: =head2 GetPeerName
192:
193: Returns the name of the host that a socket object is connected to.
194:
1.1 foxr 195: =cut
196:
197: sub GetPeername {
198: my $connection = shift;
199: my $AdrFamily = shift;
200: my $peer = $connection->peername();
201: my $peerport;
202: my $peerip;
203: if($AdrFamily == AF_INET) {
204: ($peerport, $peerip) = sockaddr_in($peer);
1.23 foxr 205: my $peername = gethostbyaddr($peerip, $AdrFamily);
1.1 foxr 206: return $peername;
207: } elsif ($AdrFamily == AF_UNIX) {
208: my $peerfile;
209: ($peerfile) = sockaddr_un($peer);
210: return $peerfile;
211: }
212: }
213: #----------------------------- Timer management ------------------------
214: =pod
1.3 albertel 215:
1.1 foxr 216: =head2 Debug
1.3 albertel 217:
218: Invoked to issue a debug message.
219:
1.1 foxr 220: =cut
1.3 albertel 221:
1.1 foxr 222: sub Debug {
223: my $level = shift;
224: my $message = shift;
225: if ($level <= $DebugLevel) {
1.23 foxr 226: Log("INFO", "-Debug- $message host = $RemoteHost");
1.1 foxr 227: }
228: }
229:
230: sub SocketDump {
231: my $level = shift;
232: my $socket= shift;
233: if($level <= $DebugLevel) {
234: $socket->Dump();
235: }
236: }
1.3 albertel 237:
1.1 foxr 238: =pod
1.3 albertel 239:
1.5 foxr 240: =head2 ShowStatus
241:
242: Place some text as our pid status.
1.10 foxr 243: and as what we return in a SIGUSR1
1.5 foxr 244:
245: =cut
246: sub ShowStatus {
1.10 foxr 247: my $state = shift;
248: my $now = time;
249: my $local = localtime($now);
250: $Status = $local.": ".$state;
251: $0='lonc: '.$state.' '.$local;
1.5 foxr 252: }
253:
254: =pod
255:
1.15 foxr 256: =head 2 SocketTimeout
257:
258: Called when an action on the socket times out. The socket is
259: destroyed and any active transaction is failed.
260:
261:
262: =cut
263: sub SocketTimeout {
264: my $Socket = shift;
265:
1.22 foxr 266: KillSocket($Socket); # A transaction timeout also counts as
267: # a connection failure:
268: $ConnectionRetriesLeft--;
1.15 foxr 269: }
270:
271: =pod
272:
1.1 foxr 273: =head2 Tick
1.3 albertel 274:
275: Invoked each timer tick.
276:
1.1 foxr 277: =cut
278:
1.5 foxr 279:
1.1 foxr 280: sub Tick {
281: my $client;
1.22 foxr 282: if($ConnectionRetriesLeft > 0) {
283: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
284: ." Retries remaining: ".$ConnectionRetriesLeft);
285: } else {
286: ShowStatus(GetServerHost()." >> DEAD <<");
287: }
1.4 foxr 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...
1.23 foxr 295: my $Socket = $IdleConnections->pop();
1.6 foxr 296: KillSocket($Socket);
1.4 foxr 297: }
298: } else {
299: $IdleSeconds = 0; # Reset idle count if not idle.
300: }
1.15 foxr 301: #
302: # For each inflight transaction, tick down its timeout counter.
303: #
1.23 foxr 304: foreach my $item (keys %ActiveTransactions) {
1.15 foxr 305: my $Socket = $ActiveTransactions{$item}->getServer();
306: $Socket->Tick();
307: }
1.5 foxr 308: # Do we have work in the queue, but no connections to service them?
309: # If so, try to make some new connections to get things going again.
310: #
311:
312: my $Requests = $WorkQueue->Count();
1.10 foxr 313: if (($ConnectionCount == 0) && ($Requests > 0)) {
314: if ($ConnectionRetriesLeft > 0) {
315: my $Connections = ($Requests <= $MaxConnectionCount) ?
316: $Requests : $MaxConnectionCount;
1.29 foxr 317: Debug(5,"Work but no connections, start ".$Connections." of them");
1.22 foxr 318: my $successCount = 0;
1.23 foxr 319: for (my $i =0; $i < $Connections; $i++) {
1.22 foxr 320: $successCount += MakeLondConnection();
321: }
322: if($successCount == 0) { # All connections failed:
1.29 foxr 323: Debug(5,"Work in queue failed to make any connectiouns\n");
1.22 foxr 324: EmptyQueue(); # Fail pending transactions with con_lost.
1.10 foxr 325: }
326: } else {
1.22 foxr 327: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
1.29 foxr 328: Debug(5,"Work in queue, but gave up on connections..flushing\n");
1.10 foxr 329: EmptyQueue(); # Connections can't be established.
1.5 foxr 330: }
331:
332: }
1.1 foxr 333: }
334:
335: =pod
1.3 albertel 336:
1.1 foxr 337: =head2 SetupTimer
338:
1.3 albertel 339: Sets up a 1 per sec recurring timer event. The event handler is used to:
1.1 foxr 340:
1.3 albertel 341: =item
342:
343: Trigger timeouts on communications along active sockets.
344:
345: =item
346:
347: Trigger disconnections of idle sockets.
1.1 foxr 348:
349: =cut
350:
351: sub SetupTimer {
352: Debug(6, "SetupTimer");
353: Event->timer(interval => 1, debug => 1, cb => \&Tick );
354: }
1.3 albertel 355:
1.1 foxr 356: =pod
1.3 albertel 357:
1.1 foxr 358: =head2 ServerToIdle
1.3 albertel 359:
360: This function is called when a connection to the server is
361: ready for more work.
362:
363: If there is work in the Work queue the top element is dequeued
1.1 foxr 364: and the connection will start to work on it. If the work queue is
365: empty, the connection is pushed on the idle connection stack where
366: it will either get another work unit, or alternatively, if it sits there
367: long enough, it will be shut down and released.
368:
1.3 albertel 369: =cut
1.1 foxr 370:
371: sub ServerToIdle {
372: my $Socket = shift; # Get the socket.
1.7 foxr 373: delete($ActiveTransactions{$Socket}); # Server has no transaction
1.1 foxr 374:
1.29 foxr 375: &Debug(5, "Server to idle");
1.1 foxr 376:
377: # If there's work to do, start the transaction:
378:
1.23 foxr 379: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
1.29 foxr 380: if ($reqdata ne undef) {
381: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
1.7 foxr 382: &StartRequest($Socket, $reqdata);
1.8 foxr 383:
1.1 foxr 384: } else {
385:
386: # There's no work waiting, so push the server to idle list.
1.29 foxr 387: &Debug(5, "No new work requests, server connection going idle");
1.1 foxr 388: $IdleConnections->push($Socket);
389: }
390: }
1.3 albertel 391:
1.1 foxr 392: =pod
1.3 albertel 393:
1.1 foxr 394: =head2 ClientWritable
1.3 albertel 395:
396: Event callback for when a client socket is writable.
397:
398: This callback is established when a transaction reponse is
399: avaiable from lond. The response is forwarded to the unix socket
400: as it becomes writable in this sub.
401:
1.1 foxr 402: Parameters:
403:
1.3 albertel 404: =item Event
405:
406: The event that has been triggered. Event->w->data is
407: the data and Event->w->fd is the socket to write.
1.1 foxr 408:
409: =cut
1.3 albertel 410:
1.1 foxr 411: sub ClientWritable {
412: my $Event = shift;
413: my $Watcher = $Event->w;
414: my $Data = $Watcher->data;
415: my $Socket = $Watcher->fd;
416:
417: # Try to send the data:
418:
419: &Debug(6, "ClientWritable writing".$Data);
420: &Debug(9, "Socket is: ".$Socket);
421:
1.6 foxr 422: if($Socket->connected) {
423: my $result = $Socket->send($Data, 0);
424:
425: # $result undefined: the write failed.
426: # otherwise $result is the number of bytes written.
427: # Remove that preceding string from the data.
428: # If the resulting data is empty, destroy the watcher
429: # and set up a read event handler to accept the next
430: # request.
431:
432: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
1.29 foxr 433: if($result ne undef) {
1.6 foxr 434: &Debug(9, "send result was defined");
435: if($result == length($Data)) { # Entire string sent.
436: &Debug(9, "ClientWritable data all written");
437: $Watcher->cancel();
438: #
439: # Set up to read next request from socket:
440:
441: my $descr = sprintf("Connection to lonc client %d",
442: $ActiveClients{$Socket});
443: Event->io(cb => \&ClientRequest,
444: poll => 'r',
445: desc => $descr,
446: data => "",
447: fd => $Socket);
448:
449: } else { # Partial string sent.
450: $Watcher->data(substr($Data, $result));
1.15 foxr 451: if($result == 0) { # client hung up on us!!
452: Log("INFO", "lonc pipe client hung up on us!");
453: $Watcher->cancel;
454: $Socket->shutdown(2);
455: $Socket->close();
456: }
1.6 foxr 457: }
458:
459: } else { # Error of some sort...
460:
461: # Some errnos are possible:
462: my $errno = $!;
463: if($errno == POSIX::EWOULDBLOCK ||
464: $errno == POSIX::EAGAIN ||
465: $errno == POSIX::EINTR) {
466: # No action taken?
467: } else { # Unanticipated errno.
468: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
469: $Watcher->cancel; # Stop the watcher.
470: $Socket->shutdown(2); # Kill connection
471: $Socket->close(); # Close the socket.
472: }
1.1 foxr 473:
474: }
1.6 foxr 475: } else {
476: $Watcher->cancel(); # A delayed request...just cancel.
1.1 foxr 477: }
478: }
479:
480: =pod
1.3 albertel 481:
1.1 foxr 482: =head2 CompleteTransaction
1.3 albertel 483:
484: Called when the reply data has been received for a lond
1.1 foxr 485: transaction. The reply data must now be sent to the
486: ultimate client on the other end of the Unix socket. This is
487: done by setting up a writable event for the socket with the
488: data the reply data.
1.3 albertel 489:
1.1 foxr 490: Parameters:
1.3 albertel 491:
492: =item Socket
493:
494: Socket on which the lond transaction occured. This is a
495: LondConnection. The data received is in the TransactionReply member.
496:
1.7 foxr 497: =item Transaction
1.3 albertel 498:
1.7 foxr 499: The transaction that is being completed.
1.1 foxr 500:
501: =cut
1.3 albertel 502:
1.1 foxr 503: sub CompleteTransaction {
1.29 foxr 504: &Debug(5,"Complete transaction");
1.1 foxr 505: my $Socket = shift;
1.7 foxr 506: my $Transaction = shift;
1.1 foxr 507:
1.7 foxr 508: if (!$Transaction->isDeferred()) { # Normal transaction
509: my $data = $Socket->GetReply(); # Data to send.
510: StartClientReply($Transaction, $data);
511: } else { # Delete deferred transaction file.
1.9 foxr 512: Log("SUCCESS", "A delayed transaction was completed");
1.23 foxr 513: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
1.7 foxr 514: unlink $Transaction->getFile();
515: }
1.6 foxr 516: }
517: =pod
518: =head1 StartClientReply
519:
520: Initiates a reply to a client where the reply data is a parameter.
521:
1.7 foxr 522: =head2 parameters:
523:
524: =item Transaction
525:
526: The transaction for which we are responding to the client.
527:
528: =item data
529:
530: The data to send to apached client.
531:
1.6 foxr 532: =cut
533: sub StartClientReply {
1.7 foxr 534: my $Transaction = shift;
1.6 foxr 535: my $data = shift;
1.1 foxr 536:
1.12 foxr 537:
1.7 foxr 538: my $Client = $Transaction->getClient();
539:
1.1 foxr 540: &Debug(8," Reply was: ".$data);
541: my $Serial = $ActiveClients{$Client};
542: my $desc = sprintf("Connection to lonc client %d",
543: $Serial);
544: Event->io(fd => $Client,
545: poll => "w",
546: desc => $desc,
547: cb => \&ClientWritable,
548: data => $data);
549: }
1.4 foxr 550: =pod
551: =head2 FailTransaction
552:
553: Finishes a transaction with failure because the associated lond socket
1.7 foxr 554: disconnected. There are two possibilities:
555: - The transaction is deferred: in which case we just quietly
556: delete the transaction since there is no client connection.
557: - The transaction is 'live' in which case we initiate the sending
558: of "con_lost" to the client.
559:
560: Deleting the transaction means killing it from the
561: %ActiveTransactions hash.
1.4 foxr 562:
563: Parameters:
564:
565: =item client
566:
1.7 foxr 567: The LondTransaction we are failing.
568:
1.4 foxr 569: =cut
570:
571: sub FailTransaction {
1.7 foxr 572: my $transaction = shift;
1.17 foxr 573: Log("WARNING", "Failing transaction ".$transaction->getRequest());
1.30 foxr 574: Debug(1, "Failing transaction: ".$transaction->getRequest());
1.10 foxr 575: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
1.11 foxr 576: my $client = $transaction->getClient();
1.30 foxr 577: Debug(1," Replying con_lost to ".$transaction->getRequest());
1.11 foxr 578: StartClientReply($transaction, "con_lost\n");
1.7 foxr 579: }
1.22 foxr 580: if($ConnectionRetriesLeft <= 0) {
581: Log("CRITICAL", "Host marked dead: ".GetServerHost());
582: }
1.4 foxr 583:
584: }
585:
586: =pod
1.6 foxr 587: =head1 EmptyQueue
1.7 foxr 588:
1.6 foxr 589: Fails all items in the work queue with con_lost.
1.7 foxr 590: Note that each item in the work queue is a transaction.
591:
1.6 foxr 592: =cut
593: sub EmptyQueue {
1.22 foxr 594: $ConnectionRetriesLeft--; # Counts as connection failure too.
1.6 foxr 595: while($WorkQueue->Count()) {
1.10 foxr 596: my $request = $WorkQueue->dequeue(); # This is a transaction
1.7 foxr 597: FailTransaction($request);
1.6 foxr 598: }
599: }
600:
601: =pod
1.4 foxr 602:
1.9 foxr 603: =head2 CloseAllLondConnections
604:
605: Close all connections open on lond prior to exit e.g.
606:
607: =cut
608: sub CloseAllLondConnections {
1.23 foxr 609: foreach my $Socket (keys %ActiveConnections) {
1.9 foxr 610: KillSocket($Socket);
611: }
612: }
613: =cut
614:
615: =pod
616:
1.4 foxr 617: =head2 KillSocket
618:
619: Destroys a socket. This function can be called either when a socket
620: has died of 'natural' causes or because a socket needs to be pruned due to
621: idleness. If the socket has died naturally, if there are no longer any
622: live connections a new connection is created (in case there are transactions
623: in the queue). If the socket has been pruned, it is never re-created.
624:
625: Parameters:
1.1 foxr 626:
1.4 foxr 627: =item Socket
628:
629: The socket to kill off.
630:
631: =item Restart
632:
633: nonzero if we are allowed to create a new connection.
634:
635:
636: =cut
637: sub KillSocket {
638: my $Socket = shift;
639:
1.17 foxr 640: Log("WARNING", "Shutting down a socket");
1.9 foxr 641: $Socket->Shutdown();
642:
1.7 foxr 643: # If the socket came from the active connection set,
644: # delete its transaction... note that FailTransaction should
645: # already have been called!!!
646: # otherwise it came from the idle set.
647: #
1.4 foxr 648:
649: if(exists($ActiveTransactions{$Socket})) {
650: delete ($ActiveTransactions{$Socket});
651: }
652: if(exists($ActiveConnections{$Socket})) {
653: delete($ActiveConnections{$Socket});
654: }
655: $ConnectionCount--;
1.6 foxr 656:
657: # If the connection count has gone to zero and there is work in the
658: # work queue, the work all gets failed with con_lost.
659: #
660: if($ConnectionCount == 0) {
1.22 foxr 661: EmptyQueue();
1.4 foxr 662: }
663: }
1.1 foxr 664:
665: =pod
1.3 albertel 666:
1.1 foxr 667: =head2 LondReadable
1.3 albertel 668:
1.1 foxr 669: This function is called whenever a lond connection
670: is readable. The action is state dependent:
671:
1.3 albertel 672: =head3 State=Initialized
673:
674: We''re waiting for the challenge, this is a no-op until the
1.1 foxr 675: state changes.
1.3 albertel 676:
1.1 foxr 677: =head3 State=Challenged
1.3 albertel 678:
679: The challenge has arrived we need to transition to Writable.
1.1 foxr 680: The connection must echo the challenge back.
1.3 albertel 681:
1.1 foxr 682: =head3 State=ChallengeReplied
1.3 albertel 683:
684: The challenge has been replied to. The we are receiveing the
1.1 foxr 685: 'ok' from the partner.
1.3 albertel 686:
1.1 foxr 687: =head3 State=RequestingKey
1.3 albertel 688:
689: The ok has been received and we need to send the request for
1.1 foxr 690: an encryption key. Transition to writable for that.
1.3 albertel 691:
1.1 foxr 692: =head3 State=ReceivingKey
1.3 albertel 693:
694: The the key has been requested, now we are reading the new key.
695:
1.1 foxr 696: =head3 State=Idle
1.3 albertel 697:
698: The encryption key has been negotiated or we have finished
1.1 foxr 699: reading data from the a transaction. If the callback data has
700: a client as well as the socket iformation, then we are
701: doing a transaction and the data received is relayed to the client
702: before the socket is put on the idle list.
1.3 albertel 703:
1.1 foxr 704: =head3 State=SendingRequest
1.3 albertel 705:
706: I do not think this state can be received here, but if it is,
1.1 foxr 707: the appropriate thing to do is to transition to writable, and send
708: the request.
1.3 albertel 709:
1.1 foxr 710: =head3 State=ReceivingReply
1.3 albertel 711:
712: We finished sending the request to the server and now transition
1.1 foxr 713: to readable to receive the reply.
714:
715: The parameter to this function are:
1.3 albertel 716:
1.1 foxr 717: The event. Implicit in this is the watcher and its data. The data
718: contains at least the lond connection object and, if a
719: transaction is in progress, the socket attached to the local client.
720:
1.3 albertel 721: =cut
1.1 foxr 722:
723: sub LondReadable {
1.8 foxr 724:
1.1 foxr 725: my $Event = shift;
726: my $Watcher = $Event->w;
727: my $Socket = $Watcher->data;
728: my $client = undef;
729:
1.23 foxr 730: &Debug(6,"LondReadable called state = ".$Socket->GetState());
1.8 foxr 731:
1.1 foxr 732:
733: my $State = $Socket->GetState(); # All action depends on the state.
734:
735: SocketDump(6, $Socket);
1.12 foxr 736: my $status = $Socket->Readable();
1.17 foxr 737:
1.12 foxr 738: &Debug(2, "Socket->Readable returned: $status");
1.1 foxr 739:
1.12 foxr 740: if($status != 0) {
1.4 foxr 741: # bad return from socket read. Currently this means that
742: # The socket has become disconnected. We fail the transaction.
743:
1.17 foxr 744: Log("WARNING",
745: "Lond connection lost.");
1.4 foxr 746: if(exists($ActiveTransactions{$Socket})) {
747: FailTransaction($ActiveTransactions{$Socket});
748: }
749: $Watcher->cancel();
1.6 foxr 750: KillSocket($Socket);
1.22 foxr 751: $ConnectionRetriesLeft--; # Counts as connection failure
1.4 foxr 752: return;
1.1 foxr 753: }
754: SocketDump(6,$Socket);
755:
756: $State = $Socket->GetState(); # Update in case of transition.
757: &Debug(6, "After read, state is ".$State);
758:
759: if($State eq "Initialized") {
760:
761:
762: } elsif ($State eq "ChallengeReceived") {
763: # The challenge must be echoed back; The state machine
764: # in the connection takes care of setting that up. Just
765: # need to transition to writable:
766:
1.8 foxr 767: $Watcher->cb(\&LondWritable);
1.1 foxr 768: $Watcher->poll("w");
769:
770: } elsif ($State eq "ChallengeReplied") {
771:
772:
773: } elsif ($State eq "RequestingKey") {
774: # The ok was received. Now we need to request the key
775: # That requires us to be writable:
776:
1.8 foxr 777: $Watcher->cb(\&LondWritable);
1.1 foxr 778: $Watcher->poll("w");
779:
780: } elsif ($State eq "ReceivingKey") {
781:
782: } elsif ($State eq "Idle") {
783: # If necessary, complete a transaction and then go into the
784: # idle queue.
1.22 foxr 785: # Note that a trasition to idle indicates a live lond
786: # on the other end so reset the connection retries.
787: #
788: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
1.8 foxr 789: $Watcher->cancel();
1.1 foxr 790: if(exists($ActiveTransactions{$Socket})) {
1.29 foxr 791: Debug(5,"Completing transaction!!");
1.1 foxr 792: CompleteTransaction($Socket,
793: $ActiveTransactions{$Socket});
1.9 foxr 794: } else {
795: Log("SUCCESS", "Connection ".$ConnectionCount." to "
796: .$RemoteHost." now ready for action");
1.1 foxr 797: }
798: ServerToIdle($Socket); # Next work unit or idle.
1.6 foxr 799:
1.1 foxr 800: } elsif ($State eq "SendingRequest") {
801: # We need to be writable for this and probably don't belong
802: # here inthe first place.
803:
804: Deubg(6, "SendingRequest state encountered in readable");
805: $Watcher->poll("w");
806: $Watcher->cb(\&LondWritable);
807:
808: } elsif ($State eq "ReceivingReply") {
809:
810:
811: } else {
812: # Invalid state.
813: Debug(4, "Invalid state in LondReadable");
814: }
815: }
1.3 albertel 816:
1.1 foxr 817: =pod
1.3 albertel 818:
1.1 foxr 819: =head2 LondWritable
1.3 albertel 820:
1.1 foxr 821: This function is called whenever a lond connection
822: becomes writable while there is a writeable monitoring
823: event. The action taken is very state dependent:
1.3 albertel 824:
1.1 foxr 825: =head3 State = Connected
1.3 albertel 826:
827: The connection is in the process of sending the 'init' hailing to the
828: lond on the remote end. The connection object''s Writable member is
829: called. On error, ConnectionError is called to destroy the connection
830: and remove it from the ActiveConnections hash
831:
1.1 foxr 832: =head3 Initialized
1.3 albertel 833:
834: 'init' has been sent, writability monitoring is removed and
835: readability monitoring is started with LondReadable as the callback.
836:
1.1 foxr 837: =head3 ChallengeReceived
1.3 albertel 838:
839: The connection has received the who are you challenge from the remote
840: system, and is in the process of sending the challenge
841: response. Writable is called.
842:
1.1 foxr 843: =head3 ChallengeReplied
1.3 albertel 844:
845: The connection has replied to the initial challenge The we switch to
846: monitoring readability looking for the server to reply with 'ok'.
847:
1.1 foxr 848: =head3 RequestingKey
1.3 albertel 849:
850: The connection is in the process of requesting its encryption key.
851: Writable is called.
852:
1.1 foxr 853: =head3 ReceivingKey
1.3 albertel 854:
855: The connection has sent the request for a key. Switch to readability
856: monitoring to accept the key
857:
1.1 foxr 858: =head3 SendingRequest
1.3 albertel 859:
860: The connection is in the process of sending a request to the server.
861: This request is part of a client transaction. All the states until
862: now represent the client setup protocol. Writable is called.
863:
1.1 foxr 864: =head3 ReceivingReply
865:
1.3 albertel 866: The connection has sent a request. Now it must receive a reply.
867: Readability monitoring is requested.
868:
869: This function is an event handler and therefore receives as
1.1 foxr 870: a parameter the event that has fired. The data for the watcher
871: of this event is a reference to a list of one or two elements,
872: depending on state. The first (and possibly only) element is the
873: socket. The second (present only if a request is in progress)
874: is the socket on which to return a reply to the caller.
875:
876: =cut
1.3 albertel 877:
1.1 foxr 878: sub LondWritable {
879: my $Event = shift;
880: my $Watcher = $Event->w;
1.8 foxr 881: my $Socket = $Watcher->data;
882: my $State = $Socket->GetState();
1.1 foxr 883:
1.8 foxr 884: Debug(6,"LondWritable State = ".$State."\n");
1.1 foxr 885:
1.8 foxr 886:
1.1 foxr 887: # Figure out what to do depending on the state of the socket:
888:
889:
890:
891:
892: SocketDump(6,$Socket);
893:
894: if ($State eq "Connected") {
895:
896: if ($Socket->Writable() != 0) {
897: # The write resulted in an error.
1.4 foxr 898: # We'll treat this as if the socket got disconnected:
1.9 foxr 899: Log("WARNING", "Connection to ".$RemoteHost.
900: " has been disconnected");
1.18 foxr 901: FailTransaction($ActiveTransactions{$Socket});
1.4 foxr 902: $Watcher->cancel();
1.6 foxr 903: KillSocket($Socket);
1.4 foxr 904: return;
1.1 foxr 905: }
1.4 foxr 906: # "init" is being sent...
907:
1.1 foxr 908:
909: } elsif ($State eq "Initialized") {
910:
911: # Now that init was sent, we switch
912: # to watching for readability:
913:
1.8 foxr 914: $Watcher->cb(\&LondReadable);
1.1 foxr 915: $Watcher->poll("r");
916:
917: } elsif ($State eq "ChallengeReceived") {
918: # We received the challenge, now we
919: # are echoing it back. This is a no-op,
920: # we're waiting for the state to change
921:
922: if($Socket->Writable() != 0) {
1.5 foxr 923:
924: $Watcher->cancel();
1.6 foxr 925: KillSocket($Socket);
1.5 foxr 926: return;
1.1 foxr 927: }
928:
929: } elsif ($State eq "ChallengeReplied") {
930: # The echo was sent back, so we switch
931: # to watching readability.
932:
1.8 foxr 933: $Watcher->cb(\&LondReadable);
1.1 foxr 934: $Watcher->poll("r");
935:
936: } elsif ($State eq "RequestingKey") {
937: # At this time we're requesting the key.
938: # again, this is essentially a no-op.
939: # we'll write the next chunk until the
940: # state changes.
941:
942: if($Socket->Writable() != 0) {
943: # Write resulted in an error.
1.5 foxr 944:
945: $Watcher->cancel();
1.6 foxr 946: KillSocket($Socket);
1.5 foxr 947: return;
948:
1.1 foxr 949: }
950: } elsif ($State eq "ReceivingKey") {
951: # Now we need to wait for the key
952: # to come back from the peer:
953:
1.8 foxr 954: $Watcher->cb(\&LondReadable);
1.1 foxr 955: $Watcher->poll("r");
956:
957: } elsif ($State eq "SendingRequest") {
958: # At this time we are sending a request to the
959: # peer... write the next chunk:
960:
961: if($Socket->Writable() != 0) {
962:
1.5 foxr 963: if(exists($ActiveTransactions{$Socket})) {
964: Debug(3, "Lond connection lost, failing transactions");
965: FailTransaction($ActiveTransactions{$Socket});
966: }
967: $Watcher->cancel();
1.6 foxr 968: KillSocket($Socket);
1.5 foxr 969: return;
970:
1.1 foxr 971: }
972:
973: } elsif ($State eq "ReceivingReply") {
974: # The send has completed. Wait for the
975: # data to come in for a reply.
976: Debug(8,"Writable sent request/receiving reply");
1.8 foxr 977: $Watcher->cb(\&LondReadable);
1.1 foxr 978: $Watcher->poll("r");
979:
980: } else {
981: # Control only passes here on an error:
982: # the socket state does not match any
983: # of the known states... so an error
984: # must be logged.
985:
986: &Debug(4, "Invalid socket state ".$State."\n");
987: }
988:
989: }
1.6 foxr 990: =pod
991:
992: =cut
993: sub QueueDelayed {
1.8 foxr 994: Debug(3,"QueueDelayed called");
995:
1.6 foxr 996: my $path = "$perlvar{'lonSockDir'}/delayed";
1.8 foxr 997:
998: Debug(4, "Delayed path: ".$path);
1.6 foxr 999: opendir(DIRHANDLE, $path);
1.8 foxr 1000:
1.23 foxr 1001: my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
1.6 foxr 1002: closedir(DIRHANDLE);
1003: my $dfname;
1.8 foxr 1004: my $reqfile;
1005: foreach $dfname (sort @alldelayed) {
1006: $reqfile = "$path/$dfname";
1007: Debug(4, "queueing ".$reqfile);
1.6 foxr 1008: my $Handle = IO::File->new($reqfile);
1009: my $cmd = <$Handle>;
1.8 foxr 1010: chomp $cmd; # There may or may not be a newline...
1.12 foxr 1011: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1.7 foxr 1012: my $Transaction = LondTransaction->new($cmd);
1013: $Transaction->SetDeferred($reqfile);
1014: QueueTransaction($Transaction);
1.6 foxr 1015: }
1016:
1017: }
1.1 foxr 1018:
1019: =pod
1.3 albertel 1020:
1.1 foxr 1021: =head2 MakeLondConnection
1.3 albertel 1022:
1023: Create a new lond connection object, and start it towards its initial
1024: idleness. Once idle, it becomes elligible to receive transactions
1025: from the work queue. If the work queue is not empty when the
1026: connection is completed and becomes idle, it will dequeue an entry and
1027: start off on it.
1028:
1.1 foxr 1029: =cut
1.3 albertel 1030:
1.1 foxr 1031: sub MakeLondConnection {
1032: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1033: .GetServerPort());
1034:
1035: my $Connection = LondConnection->new(&GetServerHost(),
1036: &GetServerPort());
1037:
1.30 foxr 1038: if($Connection eq undef) { # Needs to be more robust later.
1.9 foxr 1039: Log("CRITICAL","Failed to make a connection with lond.");
1.10 foxr 1040: $ConnectionRetriesLeft--;
1041: return 0; # Failure.
1.5 foxr 1042: } else {
1.22 foxr 1043:
1.5 foxr 1044: # The connection needs to have writability
1045: # monitored in order to send the init sequence
1046: # that starts the whole authentication/key
1047: # exchange underway.
1048: #
1049: my $Socket = $Connection->GetSocket();
1.30 foxr 1050: if($Socket eq undef) {
1.5 foxr 1051: die "did not get a socket from the connection";
1052: } else {
1053: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1054: }
1.1 foxr 1055:
1.21 foxr 1056: $Connection->SetTimeoutCallback(\&SocketTimeout);
1057:
1.23 foxr 1058: my $event = Event->io(fd => $Socket,
1.5 foxr 1059: poll => 'w',
1060: cb => \&LondWritable,
1.8 foxr 1061: data => $Connection,
1.5 foxr 1062: desc => 'Connection to lond server');
1063: $ActiveConnections{$Connection} = $event;
1064:
1065: $ConnectionCount++;
1.8 foxr 1066: Debug(4, "Connection count = ".$ConnectionCount);
1.6 foxr 1067: if($ConnectionCount == 1) { # First Connection:
1068: QueueDelayed;
1069: }
1.9 foxr 1070: Log("SUCESS", "Created connection ".$ConnectionCount
1071: ." to host ".GetServerHost());
1.10 foxr 1072: return 1; # Return success.
1.1 foxr 1073: }
1074:
1075: }
1.3 albertel 1076:
1.1 foxr 1077: =pod
1.3 albertel 1078:
1.1 foxr 1079: =head2 StartRequest
1.3 albertel 1080:
1081: Starts a lond request going on a specified lond connection.
1082: parameters are:
1083:
1084: =item $Lond
1085:
1086: Connection to the lond that will send the transaction and receive the
1087: reply.
1088:
1089: =item $Client
1090:
1091: Connection to the client that is making this request We got the
1092: request from this socket, and when the request has been relayed to
1093: lond and we get a reply back from lond it will get sent to this
1094: socket.
1095:
1096: =item $Request
1097:
1098: The text of the request to send.
1099:
1.1 foxr 1100: =cut
1101:
1102: sub StartRequest {
1103: my $Lond = shift;
1.7 foxr 1104: my $Request = shift; # This is a LondTransaction.
1.1 foxr 1105:
1.7 foxr 1106: Debug(6, "StartRequest: ".$Request->getRequest());
1.1 foxr 1107:
1108: my $Socket = $Lond->GetSocket();
1109:
1.7 foxr 1110: $Request->Activate($Lond);
1111: $ActiveTransactions{$Lond} = $Request;
1.1 foxr 1112:
1.7 foxr 1113: $Lond->InitiateTransaction($Request->getRequest());
1.23 foxr 1114: my $event = Event->io(fd => $Socket,
1.1 foxr 1115: poll => "w",
1116: cb => \&LondWritable,
1117: data => $Lond,
1118: desc => "lond transaction connection");
1119: $ActiveConnections{$Lond} = $event;
1120: Debug(8," Start Request made watcher data with ".$event->data."\n");
1121: }
1122:
1123: =pod
1.3 albertel 1124:
1.1 foxr 1125: =head2 QueueTransaction
1.3 albertel 1126:
1127: If there is an idle lond connection, it is put to work doing this
1128: transaction. Otherwise, the transaction is placed in the work queue.
1129: If placed in the work queue and the maximum number of connections has
1130: not yet been created, a new connection will be started. Our goal is
1131: to eventually have a sufficient number of connections that the work
1132: queue will typically be empty. parameters are:
1133:
1134: =item Socket
1135:
1136: open on the lonc client.
1137:
1138: =item Request
1139:
1140: data to send to the lond.
1.1 foxr 1141:
1142: =cut
1.3 albertel 1143:
1.1 foxr 1144: sub QueueTransaction {
1145:
1.7 foxr 1146: my $requestData = shift; # This is a LondTransaction.
1147: my $cmd = $requestData->getRequest();
1148:
1149: Debug(6,"QueueTransaction: ".$cmd);
1.1 foxr 1150:
1151: my $LondSocket = $IdleConnections->pop();
1152: if(!defined $LondSocket) { # Need to queue request.
1.29 foxr 1153: Debug(5,"Must queue...");
1.1 foxr 1154: $WorkQueue->enqueue($requestData);
1155: if($ConnectionCount < $MaxConnectionCount) {
1.22 foxr 1156: if($ConnectionRetriesLeft > 0) {
1.29 foxr 1157: Debug(5,"Starting additional lond connection");
1.22 foxr 1158: if(MakeLondConnection() == 0) {
1159: EmptyQueue(); # Fail transactions, can't make connection.
1160: }
1161: } else {
1162: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1163: EmptyQueue(); # It's worse than that ... he's dead Jim.
1.17 foxr 1164: }
1.1 foxr 1165: }
1166: } else { # Can start the request:
1167: Debug(8,"Can start...");
1.7 foxr 1168: StartRequest($LondSocket, $requestData);
1.1 foxr 1169: }
1170: }
1171:
1172: #-------------------------- Lonc UNIX socket handling ---------------------
1.3 albertel 1173:
1.1 foxr 1174: =pod
1.3 albertel 1175:
1.1 foxr 1176: =head2 ClientRequest
1.3 albertel 1177: Callback that is called when data can be read from the UNIX domain
1178: socket connecting us with an apache server process.
1.1 foxr 1179:
1180: =cut
1181:
1182: sub ClientRequest {
1183: Debug(6, "ClientRequest");
1184: my $event = shift;
1185: my $watcher = $event->w;
1186: my $socket = $watcher->fd;
1187: my $data = $watcher->data;
1188: my $thisread;
1189:
1190: Debug(9, " Watcher named: ".$watcher->desc);
1191:
1192: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1193: Debug(8, "rcv: data length = ".length($thisread)
1194: ." read =".$thisread);
1.29 foxr 1195: unless (defined $rv && length($thisread)) {
1.1 foxr 1196: # Likely eof on socket.
1197: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1198: close($socket);
1199: $watcher->cancel();
1200: delete($ActiveClients{$socket});
1.10 foxr 1201: return;
1.1 foxr 1202: }
1203: Debug(8,"Data: ".$data." this read: ".$thisread);
1204: $data = $data.$thisread; # Append new data.
1205: $watcher->data($data);
1206: if($data =~ /(.*\n)/) { # Request entirely read.
1.10 foxr 1207: if($data eq "close_connection_exit\n") {
1.9 foxr 1208: Log("CRITICAL",
1209: "Request Close Connection ... exiting");
1210: CloseAllLondConnections();
1211: exit;
1212: }
1.1 foxr 1213: Debug(8, "Complete transaction received: ".$data);
1.8 foxr 1214: my $Transaction = LondTransaction->new($data);
1.7 foxr 1215: $Transaction->SetClient($socket);
1216: QueueTransaction($Transaction);
1.1 foxr 1217: $watcher->cancel(); # Done looking for input data.
1218: }
1219:
1220: }
1221:
1222:
1223: =pod
1.3 albertel 1224:
1.1 foxr 1225: =head2 NewClient
1.3 albertel 1226:
1227: Callback that is called when a connection is received on the unix
1228: socket for a new client of lonc. The callback is parameterized by the
1229: event.. which is a-priori assumed to be an io event, and therefore has
1230: an fd member that is the Listener socket. We Accept the connection
1231: and register a new event on the readability of that socket:
1232:
1.1 foxr 1233: =cut
1.3 albertel 1234:
1.1 foxr 1235: sub NewClient {
1236: Debug(6, "NewClient");
1237: my $event = shift; # Get the event parameters.
1238: my $watcher = $event->w;
1239: my $socket = $watcher->fd; # Get the event' socket.
1240: my $connection = $socket->accept(); # Accept the client connection.
1241: Debug(8,"Connection request accepted from "
1242: .GetPeername($connection, AF_UNIX));
1243:
1244:
1245: my $description = sprintf("Connection to lonc client %d",
1246: $ClientConnection);
1247: Debug(9, "Creating event named: ".$description);
1248: Event->io(cb => \&ClientRequest,
1249: poll => 'r',
1250: desc => $description,
1251: data => "",
1252: fd => $connection);
1253: $ActiveClients{$connection} = $ClientConnection;
1254: $ClientConnection++;
1255: }
1.3 albertel 1256:
1257: =pod
1258:
1259: =head2 GetLoncSocketPath
1260:
1261: Returns the name of the UNIX socket on which to listen for client
1262: connections.
1.1 foxr 1263:
1264: =cut
1.3 albertel 1265:
1.1 foxr 1266: sub GetLoncSocketPath {
1267: return $UnixSocketDir."/".GetServerHost();
1268: }
1269:
1.3 albertel 1270: =pod
1271:
1272: =head2 GetServerHost
1273:
1274: Returns the host whose lond we talk with.
1275:
1.1 foxr 1276: =cut
1.3 albertel 1277:
1.7 foxr 1278: sub GetServerHost {
1.1 foxr 1279: return $RemoteHost; # Setup by the fork.
1280: }
1.3 albertel 1281:
1282: =pod
1283:
1284: =head2 GetServerPort
1285:
1286: Returns the lond port number.
1287:
1.1 foxr 1288: =cut
1.3 albertel 1289:
1.7 foxr 1290: sub GetServerPort {
1.1 foxr 1291: return $perlvar{londPort};
1292: }
1.3 albertel 1293:
1294: =pod
1295:
1296: =head2 SetupLoncListener
1297:
1298: Setup a lonc listener event. The event is called when the socket
1299: becomes readable.. that corresponds to the receipt of a new
1300: connection. The event handler established will accept the connection
1301: (creating a communcations channel), that int turn will establish
1302: another event handler to subess requests.
1.1 foxr 1303:
1304: =cut
1.3 albertel 1305:
1.1 foxr 1306: sub SetupLoncListener {
1307:
1308: my $socket;
1309: my $SocketName = GetLoncSocketPath();
1310: unlink($SocketName);
1.7 foxr 1311: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1.1 foxr 1312: Listen => 10,
1313: Type => SOCK_STREAM)) {
1314: die "Failed to create a lonc listner socket";
1315: }
1316: Event->io(cb => \&NewClient,
1317: poll => 'r',
1318: desc => 'Lonc listener Unix Socket',
1319: fd => $socket);
1320: }
1321:
1.14 foxr 1322: =pod
1323:
1324: =head2 ChildStatus
1325:
1326: Child USR1 signal handler to report the most recent status
1327: into the status file.
1328:
1.22 foxr 1329: We also use this to reset the retries count in order to allow the
1330: client to retry connections with a previously dead server.
1.14 foxr 1331: =cut
1332: sub ChildStatus {
1333: my $event = shift;
1334: my $watcher = $event->w;
1335:
1336: Debug(2, "Reporting child status because : ".$watcher->data);
1337: my $docdir = $perlvar{'lonDocRoot'};
1338: my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
1339: print $fh $$."\t".$RemoteHost."\t".$Status."\t".
1340: $RecentLogEntry."\n";
1.22 foxr 1341: $ConnectionRetriesLeft = $ConnectionRetries;
1.14 foxr 1342: }
1343:
1.1 foxr 1344: =pod
1.3 albertel 1345:
1.10 foxr 1346: =head2 SignalledToDeath
1347:
1348: Called in response to a signal that causes a chid process to die.
1349:
1350: =cut
1351:
1352:
1353: sub SignalledToDeath {
1.14 foxr 1354: my $event = shift;
1355: my $watcher= $event->w;
1356:
1357: Debug(2,"Signalled to death! via ".$watcher->data);
1.17 foxr 1358: my ($signal) = $watcher->data;
1.10 foxr 1359: chomp($signal);
1360: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1361: ."died through "."\"$signal\"");
1362: LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1363: ."\"$signal\"");
1.12 foxr 1364: exit 0;
1.10 foxr 1365:
1366: }
1.16 foxr 1367:
1368: =head2 ToggleDebug
1369:
1370: This sub toggles trace debugging on and off.
1371:
1372: =cut
1373:
1374: sub ToggleDebug {
1375: my $Current = $DebugLevel;
1376: $DebugLevel = $NextDebugLevel;
1377: $NextDebugLevel = $Current;
1378:
1379: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1380:
1381: }
1382:
1.1 foxr 1383: =head2 ChildProcess
1384:
1385: This sub implements a child process for a single lonc daemon.
1386:
1387: =cut
1388:
1389: sub ChildProcess {
1390:
1391:
1.14 foxr 1392: #
1393: # Signals must be handled by the Event framework...
1394: #
1395:
1396: Event->signal(signal => "QUIT",
1397: cb => \&SignalledToDeath,
1398: data => "QUIT");
1399: Event->signal(signal => "HUP",
1400: cb => \&ChildStatus,
1401: data => "HUP");
1402: Event->signal(signal => "USR1",
1403: cb => \&ChildStatus,
1404: data => "USR1");
1.16 foxr 1405: Event->signal(signal => "INT",
1406: cb => \&ToggleDebug,
1407: data => "INT");
1.1 foxr 1408:
1409: SetupTimer();
1410:
1411: SetupLoncListener();
1412:
1413: $Event::Debuglevel = $DebugLevel;
1414:
1415: Debug(9, "Making initial lond connection for ".$RemoteHost);
1416:
1417: # Setup the initial server connection:
1418:
1.14 foxr 1419: # &MakeLondConnection(); // let first work requirest do it.
1.10 foxr 1420:
1.5 foxr 1421:
1.1 foxr 1422: Debug(9,"Entering event loop");
1423: my $ret = Event::loop(); # Start the main event loop.
1424:
1425:
1426: die "Main event loop exited!!!";
1427: }
1428:
1429: # Create a new child for host passed in:
1430:
1431: sub CreateChild {
1.12 foxr 1432: my $sigset = POSIX::SigSet->new(SIGINT);
1433: sigprocmask(SIG_BLOCK, $sigset);
1.1 foxr 1434: my $host = shift;
1435: $RemoteHost = $host;
1.9 foxr 1436: Log("CRITICAL", "Forking server for ".$host);
1.23 foxr 1437: my $pid = fork;
1.1 foxr 1438: if($pid) { # Parent
1.17 foxr 1439: $RemoteHost = "Parent";
1.27 foxr 1440: $ChildHash{$pid} = $host;
1.26 foxr 1441: $HostToPid{$host}= $pid;
1.12 foxr 1442: sigprocmask(SIG_UNBLOCK, $sigset);
1443:
1.1 foxr 1444: } else { # child.
1.5 foxr 1445: ShowStatus("Connected to ".$RemoteHost);
1.23 foxr 1446: $SIG{INT} = 'DEFAULT';
1.12 foxr 1447: sigprocmask(SIG_UNBLOCK, $sigset);
1448: ChildProcess; # Does not return.
1.1 foxr 1449: }
1450:
1451: }
1452: #
1453: # Parent process logic pass 1:
1454: # For each entry in the hosts table, we will
1455: # fork off an instance of ChildProcess to service the transactions
1456: # to that host. Each pid will be entered in a global hash
1457: # with the value of the key, the host.
1458: # The parent will then enter a loop to wait for process exits.
1459: # Each exit gets logged and the child gets restarted.
1460: #
1461:
1.5 foxr 1462: #
1463: # Fork and start in new session so hang-up isn't going to
1464: # happen without intent.
1465: #
1466:
1467:
1.6 foxr 1468:
1469:
1.8 foxr 1470:
1.6 foxr 1471:
1472: ShowStatus("Forming new session");
1473: my $childpid = fork;
1474: if ($childpid != 0) {
1475: sleep 4; # Give child a chacne to break to
1476: exit 0; # a new sesion.
1477: }
1.8 foxr 1478: #
1479: # Write my pid into the pid file so I can be located
1480: #
1481:
1482: ShowStatus("Parent writing pid file:");
1.23 foxr 1483: my $execdir = $perlvar{'lonDaemons'};
1.8 foxr 1484: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1485: print PIDSAVE "$$\n";
1486: close(PIDSAVE);
1.6 foxr 1487:
1.17 foxr 1488:
1489:
1.6 foxr 1490: if (POSIX::setsid() < 0) {
1491: print "Could not create new session\n";
1492: exit -1;
1493: }
1.5 foxr 1494:
1495: ShowStatus("Forking node servers");
1496:
1.9 foxr 1497: Log("CRITICAL", "--------------- Starting children ---------------");
1498:
1.31 foxr 1499: LondConnection::ReadConfig; # Read standard config files.
1.1 foxr 1500: my $HostIterator = LondConnection::GetHostIterator;
1501: while (! $HostIterator->end()) {
1502:
1.23 foxr 1503: my $hostentryref = $HostIterator->get();
1.1 foxr 1504: CreateChild($hostentryref->[0]);
1.26 foxr 1505: $HostHash{$hostentryref->[0]} = $hostentryref->[4];
1.1 foxr 1506: $HostIterator->next();
1507: }
1.12 foxr 1508: $RemoteHost = "Parent Server";
1.1 foxr 1509:
1510: # Maintain the population:
1.5 foxr 1511:
1512: ShowStatus("Parent keeping the flock");
1.1 foxr 1513:
1.10 foxr 1514: #
1515: # Set up parent signals:
1516: #
1.12 foxr 1517:
1.14 foxr 1518: $SIG{INT} = \&Terminate;
1519: $SIG{TERM} = \&Terminate;
1.13 foxr 1520: $SIG{HUP} = \&Restart;
1.14 foxr 1521: $SIG{USR1} = \&CheckKids;
1.24 foxr 1522: $SIG{USR2} = \&UpdateKids; # LonManage update request.
1.10 foxr 1523:
1.1 foxr 1524: while(1) {
1.23 foxr 1525: my $deadchild = wait();
1.1 foxr 1526: if(exists $ChildHash{$deadchild}) { # need to restart.
1.23 foxr 1527: my $deadhost = $ChildHash{$deadchild};
1.26 foxr 1528: delete($HostToPid{$deadhost});
1.1 foxr 1529: delete($ChildHash{$deadchild});
1.9 foxr 1530: Log("WARNING","Lost child pid= ".$deadchild.
1.1 foxr 1531: "Connected to host ".$deadhost);
1.9 foxr 1532: Log("INFO", "Restarting child procesing ".$deadhost);
1.1 foxr 1533: CreateChild($deadhost);
1534: }
1.13 foxr 1535: }
1536:
1.14 foxr 1537:
1538:
1539: =pod
1540:
1541: =head1 CheckKids
1542:
1543: Since kids do not die as easily in this implementation
1544: as the previous one, there is no need to restart the
1545: dead ones (all dead kids get restarted when they die!!)
1546: The only thing this function does is to pass USR1 to the
1547: kids so that they report their status.
1548:
1549: =cut
1550:
1551: sub CheckKids {
1552: Debug(2, "Checking status of children");
1553: my $docdir = $perlvar{'lonDocRoot'};
1554: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
1555: my $now=time;
1556: my $local=localtime($now);
1557: print $fh "LONC status $local - parent $$ \n\n";
1.23 foxr 1558: foreach my $pid (keys %ChildHash) {
1.14 foxr 1559: Debug(2, "Sending USR1 -> $pid");
1560: kill 'USR1' => $pid; # Tell Child to report status.
1561: sleep 1; # Wait so file doesn't intermix.
1562: }
1563: }
1.24 foxr 1564:
1565: =pod
1566:
1567: =head1 UpdateKids
1568:
1.25 foxr 1569: parent's SIGUSR2 handler. This handler:
1.24 foxr 1570:
1571: =item
1572:
1573: Rereads the hosts file.
1574:
1575: =item
1576:
1577: Kills off (via sigint) children for hosts that have disappeared.
1578:
1579: =item
1580:
1.27 foxr 1581: QUITs children for hosts that already exist (this just forces a status display
1.24 foxr 1582: and resets the connection retry count for that host.
1583:
1584: =item
1585:
1586: Starts new children for hosts that have been added to the hosts.tab file since
1587: the start of the master program and maintains them.
1588:
1589: =cut
1590:
1591: sub UpdateKids {
1.27 foxr 1592:
1.25 foxr 1593: Log("INFO", "Updating connections via SIGUSR2");
1.27 foxr 1594:
1595: # Just in case we need to kill our own lonc, we wait a few seconds to
1596: # give it a chance to receive and relay lond's response to the
1597: # re-init command.
1598: #
1599:
1600: sleep(2); # Wait a couple of seconds.
1601:
1602: my %hosts; # Indexed by loncapa hostname, value=ip.
1603:
1604: # Need to re-read the host table:
1605:
1606:
1607: LondConnection::ReadConfig();
1608: my $I = LondConnection::GetHostIterator;
1609: while (! $I->end()) {
1610: my $item = $I->get();
1611: $hosts{$item->[0]} = $item->[4];
1612: $I->next();
1613: }
1614:
1615: # The logic below is written for clarity not for efficiency.
1616: # Since I anticipate that this function is only rarely called, that's
1617: # appropriate. There are certainly ways to combine the loops below,
1618: # and anyone wishing to obscure the logic is welcome to go for it.
1619: # Note that we don't re-direct sigchild. Instead we do what's needed
1620: # to the data structures that keep track of children to ensure that
1621: # when sigchild is honored, no new child is born.
1622: #
1623:
1624: # For each existing child; if it's host doesn't exist, kill the child.
1625:
1626: foreach my $child (keys %ChildHash) {
1627: my $oldhost = $ChildHash{$child};
1628: if (!(exists $hosts{$oldhost})) {
1629: Log("CRITICAL", "Killing child for $oldhost host no longer exists");
1630: delete $ChildHash{$child};
1631: delete $HostToPid{$oldhost};
1632: kill 'QUIT' => $child;
1633: }
1634: }
1635: # For each remaining existing child; if it's host's ip has changed,
1636: # Restart the child on the new IP.
1637:
1638: foreach my $child (keys %ChildHash) {
1639: my $oldhost = $ChildHash{$child};
1640: my $oldip = $HostHash{$oldhost};
1641: if ($hosts{$oldhost} ne $oldip) {
1642:
1643: # kill the old child.
1644:
1645: Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
1646: delete $ChildHash{$child};
1647: delete $HostToPid{$oldhost};
1648: kill 'QUIT' => $child;
1649:
1650: # Do the book-keeping needed to start a new child on the
1651: # new ip.
1652:
1653: $HostHash{$oldhost} = $hosts{$oldhost};
1654: CreateChild($oldhost);
1655: }
1656: }
1657: # Finally, for each new host, not in the host hash, create a
1658: # enter the host and create a new child.
1659: # Force a status display of any existing process.
1660:
1661: foreach my $host (keys %hosts) {
1662: if(!(exists $HostHash{$host})) {
1663: Log("INFO", "New host $host discovered in hosts.tab...");
1664: $HostHash{$host} = $hosts{$host};
1665: CreateChild($host);
1666: } else {
1667: kill 'HUP' => $HostToPid{$host}; # status display.
1668: }
1669: }
1.24 foxr 1670: }
1671:
1.14 foxr 1672:
1.13 foxr 1673: =pod
1674:
1675: =head1 Restart
1676:
1677: Signal handler for HUP... all children are killed and
1678: we self restart. This is an el-cheapo way to re read
1679: the config file.
1680:
1681: =cut
1682:
1683: sub Restart {
1.23 foxr 1684: &KillThemAll; # First kill all the children.
1.13 foxr 1685: Log("CRITICAL", "Restarting");
1686: my $execdir = $perlvar{'lonDaemons'};
1687: unlink("$execdir/logs/lonc.pid");
1.28 albertel 1688: exec("$execdir/loncnew");
1.10 foxr 1689: }
1.12 foxr 1690:
1691: =pod
1692:
1693: =head1 KillThemAll
1694:
1695: Signal handler that kills all children by sending them a
1.17 foxr 1696: SIGHUP. Responds to sigint and sigterm.
1.12 foxr 1697:
1698: =cut
1699:
1.10 foxr 1700: sub KillThemAll {
1.12 foxr 1701: Debug(2, "Kill them all!!");
1702: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
1.23 foxr 1703: foreach my $pid (keys %ChildHash) {
1.12 foxr 1704: my $serving = $ChildHash{$pid};
1705: Debug(2, "Killing lonc for $serving pid = $pid");
1706: ShowStatus("Killing lonc for $serving pid = $pid");
1707: Log("CRITICAL", "Killing lonc for $serving pid = $pid");
1.17 foxr 1708: kill 'QUIT' => $pid;
1709: delete($ChildHash{$pid});
1.12 foxr 1710: }
1.14 foxr 1711: my $execdir = $perlvar{'lonDaemons'};
1712: unlink("$execdir/logs/lonc.pid");
1.17 foxr 1713:
1.1 foxr 1714: }
1.12 foxr 1715:
1.14 foxr 1716: =pod
1717:
1718: =head1 Terminate
1719:
1720: Terminate the system.
1721:
1722: =cut
1723:
1724: sub Terminate {
1725: KillThemAll;
1.17 foxr 1726: Log("CRITICAL","Master process exiting");
1727: exit 0;
1.14 foxr 1728:
1729: }
1.12 foxr 1730: =pod
1.1 foxr 1731:
1732: =head1 Theory
1.3 albertel 1733:
1734: The event class is used to build this as a single process with an
1735: event driven model. The following events are handled:
1.1 foxr 1736:
1737: =item UNIX Socket connection Received
1738:
1739: =item Request data arrives on UNIX data transfer socket.
1740:
1741: =item lond connection becomes writable.
1742:
1743: =item timer fires at 1 second intervals.
1744:
1745: All sockets are run in non-blocking mode. Timeouts managed by the timer
1746: handler prevents hung connections.
1747:
1748: Key data structures:
1749:
1.3 albertel 1750: =item RequestQueue
1751:
1752: A queue of requests received from UNIX sockets that are
1753: waiting for a chance to be forwarded on a lond connection socket.
1754:
1755: =item ActiveConnections
1756:
1757: A hash of lond connections that have transactions in process that are
1758: available to be timed out.
1759:
1760: =item ActiveTransactions
1761:
1762: A hash indexed by lond connections that contain the client reply
1763: socket for each connection that has an active transaction on it.
1764:
1765: =item IdleConnections
1766:
1767: A hash of lond connections that have no work to do. These connections
1768: can be closed if they are idle for a long enough time.
1.1 foxr 1769:
1770: =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