Annotation of loncom/loncnew, revision 1.36
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: #
1.36 ! albertel 5: # $Id: loncnew,v 1.35 2003/12/12 09:57:28 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: =pod
1.3 albertel 214:
1.1 foxr 215: =head2 Debug
1.3 albertel 216:
217: Invoked to issue a debug message.
218:
1.1 foxr 219: =cut
1.3 albertel 220:
1.1 foxr 221: sub Debug {
222: my $level = shift;
223: my $message = shift;
224: if ($level <= $DebugLevel) {
1.23 foxr 225: Log("INFO", "-Debug- $message host = $RemoteHost");
1.1 foxr 226: }
227: }
228:
229: sub SocketDump {
230: my $level = shift;
231: my $socket= shift;
232: if($level <= $DebugLevel) {
233: $socket->Dump();
234: }
235: }
1.3 albertel 236:
1.1 foxr 237: =pod
1.3 albertel 238:
1.5 foxr 239: =head2 ShowStatus
240:
241: Place some text as our pid status.
1.10 foxr 242: and as what we return in a SIGUSR1
1.5 foxr 243:
244: =cut
245: sub ShowStatus {
1.10 foxr 246: my $state = shift;
247: my $now = time;
248: my $local = localtime($now);
249: $Status = $local.": ".$state;
250: $0='lonc: '.$state.' '.$local;
1.5 foxr 251: }
252:
253: =pod
254:
1.15 foxr 255: =head 2 SocketTimeout
256:
257: Called when an action on the socket times out. The socket is
258: destroyed and any active transaction is failed.
259:
260:
261: =cut
262: sub SocketTimeout {
263: my $Socket = shift;
264:
1.22 foxr 265: KillSocket($Socket); # A transaction timeout also counts as
266: # a connection failure:
267: $ConnectionRetriesLeft--;
1.15 foxr 268: }
1.35 foxr 269: #----------------------------- Timer management ------------------------
1.15 foxr 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: #
1.35 foxr 304:
1.34 albertel 305: foreach my $item (keys %ActiveConnections) {
306: my $State = $ActiveConnections{$item}->data->GetState();
1.35 foxr 307: if ($State ne 'Idle') {
1.34 albertel 308: Debug(5,"Ticking Socket $State $item");
309: $ActiveConnections{$item}->data->Tick();
310: }
1.15 foxr 311: }
1.5 foxr 312: # Do we have work in the queue, but no connections to service them?
313: # If so, try to make some new connections to get things going again.
314: #
315:
316: my $Requests = $WorkQueue->Count();
1.10 foxr 317: if (($ConnectionCount == 0) && ($Requests > 0)) {
318: if ($ConnectionRetriesLeft > 0) {
319: my $Connections = ($Requests <= $MaxConnectionCount) ?
320: $Requests : $MaxConnectionCount;
1.29 foxr 321: Debug(5,"Work but no connections, start ".$Connections." of them");
1.22 foxr 322: my $successCount = 0;
1.23 foxr 323: for (my $i =0; $i < $Connections; $i++) {
1.22 foxr 324: $successCount += MakeLondConnection();
325: }
326: if($successCount == 0) { # All connections failed:
1.29 foxr 327: Debug(5,"Work in queue failed to make any connectiouns\n");
1.22 foxr 328: EmptyQueue(); # Fail pending transactions with con_lost.
1.10 foxr 329: }
330: } else {
1.22 foxr 331: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
1.29 foxr 332: Debug(5,"Work in queue, but gave up on connections..flushing\n");
1.10 foxr 333: EmptyQueue(); # Connections can't be established.
1.5 foxr 334: }
335:
336: }
1.1 foxr 337: }
338:
339: =pod
1.3 albertel 340:
1.1 foxr 341: =head2 SetupTimer
342:
1.3 albertel 343: Sets up a 1 per sec recurring timer event. The event handler is used to:
1.1 foxr 344:
1.3 albertel 345: =item
346:
347: Trigger timeouts on communications along active sockets.
348:
349: =item
350:
351: Trigger disconnections of idle sockets.
1.1 foxr 352:
353: =cut
354:
355: sub SetupTimer {
356: Debug(6, "SetupTimer");
1.33 foxr 357: Event->timer(interval => 1, cb => \&Tick );
1.1 foxr 358: }
1.3 albertel 359:
1.1 foxr 360: =pod
1.3 albertel 361:
1.1 foxr 362: =head2 ServerToIdle
1.3 albertel 363:
364: This function is called when a connection to the server is
365: ready for more work.
366:
367: If there is work in the Work queue the top element is dequeued
1.1 foxr 368: and the connection will start to work on it. If the work queue is
369: empty, the connection is pushed on the idle connection stack where
370: it will either get another work unit, or alternatively, if it sits there
371: long enough, it will be shut down and released.
372:
1.3 albertel 373: =cut
1.1 foxr 374:
375: sub ServerToIdle {
376: my $Socket = shift; # Get the socket.
1.7 foxr 377: delete($ActiveTransactions{$Socket}); # Server has no transaction
1.1 foxr 378:
1.29 foxr 379: &Debug(5, "Server to idle");
1.1 foxr 380:
381: # If there's work to do, start the transaction:
382:
1.23 foxr 383: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
1.29 foxr 384: if ($reqdata ne undef) {
385: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
1.7 foxr 386: &StartRequest($Socket, $reqdata);
1.8 foxr 387:
1.1 foxr 388: } else {
389:
390: # There's no work waiting, so push the server to idle list.
1.29 foxr 391: &Debug(5, "No new work requests, server connection going idle");
1.1 foxr 392: $IdleConnections->push($Socket);
393: }
394: }
1.3 albertel 395:
1.1 foxr 396: =pod
1.3 albertel 397:
1.1 foxr 398: =head2 ClientWritable
1.3 albertel 399:
400: Event callback for when a client socket is writable.
401:
402: This callback is established when a transaction reponse is
403: avaiable from lond. The response is forwarded to the unix socket
404: as it becomes writable in this sub.
405:
1.1 foxr 406: Parameters:
407:
1.3 albertel 408: =item Event
409:
410: The event that has been triggered. Event->w->data is
411: the data and Event->w->fd is the socket to write.
1.1 foxr 412:
413: =cut
1.3 albertel 414:
1.1 foxr 415: sub ClientWritable {
416: my $Event = shift;
417: my $Watcher = $Event->w;
418: my $Data = $Watcher->data;
419: my $Socket = $Watcher->fd;
420:
421: # Try to send the data:
422:
423: &Debug(6, "ClientWritable writing".$Data);
424: &Debug(9, "Socket is: ".$Socket);
425:
1.6 foxr 426: if($Socket->connected) {
427: my $result = $Socket->send($Data, 0);
428:
429: # $result undefined: the write failed.
430: # otherwise $result is the number of bytes written.
431: # Remove that preceding string from the data.
432: # If the resulting data is empty, destroy the watcher
433: # and set up a read event handler to accept the next
434: # request.
435:
436: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
1.29 foxr 437: if($result ne undef) {
1.6 foxr 438: &Debug(9, "send result was defined");
439: if($result == length($Data)) { # Entire string sent.
440: &Debug(9, "ClientWritable data all written");
441: $Watcher->cancel();
442: #
443: # Set up to read next request from socket:
444:
445: my $descr = sprintf("Connection to lonc client %d",
446: $ActiveClients{$Socket});
447: Event->io(cb => \&ClientRequest,
448: poll => 'r',
449: desc => $descr,
450: data => "",
451: fd => $Socket);
452:
453: } else { # Partial string sent.
454: $Watcher->data(substr($Data, $result));
1.15 foxr 455: if($result == 0) { # client hung up on us!!
456: Log("INFO", "lonc pipe client hung up on us!");
457: $Watcher->cancel;
458: $Socket->shutdown(2);
459: $Socket->close();
460: }
1.6 foxr 461: }
462:
463: } else { # Error of some sort...
464:
465: # Some errnos are possible:
466: my $errno = $!;
467: if($errno == POSIX::EWOULDBLOCK ||
468: $errno == POSIX::EAGAIN ||
469: $errno == POSIX::EINTR) {
470: # No action taken?
471: } else { # Unanticipated errno.
472: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
473: $Watcher->cancel; # Stop the watcher.
474: $Socket->shutdown(2); # Kill connection
475: $Socket->close(); # Close the socket.
476: }
1.1 foxr 477:
478: }
1.6 foxr 479: } else {
480: $Watcher->cancel(); # A delayed request...just cancel.
1.1 foxr 481: }
482: }
483:
484: =pod
1.3 albertel 485:
1.1 foxr 486: =head2 CompleteTransaction
1.3 albertel 487:
488: Called when the reply data has been received for a lond
1.1 foxr 489: transaction. The reply data must now be sent to the
490: ultimate client on the other end of the Unix socket. This is
491: done by setting up a writable event for the socket with the
492: data the reply data.
1.3 albertel 493:
1.1 foxr 494: Parameters:
1.3 albertel 495:
496: =item Socket
497:
498: Socket on which the lond transaction occured. This is a
499: LondConnection. The data received is in the TransactionReply member.
500:
1.7 foxr 501: =item Transaction
1.3 albertel 502:
1.7 foxr 503: The transaction that is being completed.
1.1 foxr 504:
505: =cut
1.3 albertel 506:
1.1 foxr 507: sub CompleteTransaction {
1.29 foxr 508: &Debug(5,"Complete transaction");
1.1 foxr 509: my $Socket = shift;
1.7 foxr 510: my $Transaction = shift;
1.1 foxr 511:
1.7 foxr 512: if (!$Transaction->isDeferred()) { # Normal transaction
513: my $data = $Socket->GetReply(); # Data to send.
514: StartClientReply($Transaction, $data);
515: } else { # Delete deferred transaction file.
1.9 foxr 516: Log("SUCCESS", "A delayed transaction was completed");
1.23 foxr 517: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
1.7 foxr 518: unlink $Transaction->getFile();
519: }
1.6 foxr 520: }
521: =pod
522: =head1 StartClientReply
523:
524: Initiates a reply to a client where the reply data is a parameter.
525:
1.7 foxr 526: =head2 parameters:
527:
528: =item Transaction
529:
530: The transaction for which we are responding to the client.
531:
532: =item data
533:
534: The data to send to apached client.
535:
1.6 foxr 536: =cut
537: sub StartClientReply {
1.7 foxr 538: my $Transaction = shift;
1.6 foxr 539: my $data = shift;
1.1 foxr 540:
1.12 foxr 541:
1.7 foxr 542: my $Client = $Transaction->getClient();
543:
1.1 foxr 544: &Debug(8," Reply was: ".$data);
545: my $Serial = $ActiveClients{$Client};
546: my $desc = sprintf("Connection to lonc client %d",
547: $Serial);
548: Event->io(fd => $Client,
549: poll => "w",
550: desc => $desc,
551: cb => \&ClientWritable,
552: data => $data);
553: }
1.4 foxr 554: =pod
555: =head2 FailTransaction
556:
557: Finishes a transaction with failure because the associated lond socket
1.7 foxr 558: disconnected. There are two possibilities:
559: - The transaction is deferred: in which case we just quietly
560: delete the transaction since there is no client connection.
561: - The transaction is 'live' in which case we initiate the sending
562: of "con_lost" to the client.
563:
564: Deleting the transaction means killing it from the
565: %ActiveTransactions hash.
1.4 foxr 566:
567: Parameters:
568:
569: =item client
570:
1.7 foxr 571: The LondTransaction we are failing.
572:
1.4 foxr 573: =cut
574:
575: sub FailTransaction {
1.7 foxr 576: my $transaction = shift;
1.17 foxr 577: Log("WARNING", "Failing transaction ".$transaction->getRequest());
1.30 foxr 578: Debug(1, "Failing transaction: ".$transaction->getRequest());
1.10 foxr 579: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
1.11 foxr 580: my $client = $transaction->getClient();
1.30 foxr 581: Debug(1," Replying con_lost to ".$transaction->getRequest());
1.11 foxr 582: StartClientReply($transaction, "con_lost\n");
1.7 foxr 583: }
1.22 foxr 584: if($ConnectionRetriesLeft <= 0) {
585: Log("CRITICAL", "Host marked dead: ".GetServerHost());
586: }
1.4 foxr 587:
588: }
589:
590: =pod
1.6 foxr 591: =head1 EmptyQueue
1.7 foxr 592:
1.6 foxr 593: Fails all items in the work queue with con_lost.
1.7 foxr 594: Note that each item in the work queue is a transaction.
595:
1.6 foxr 596: =cut
597: sub EmptyQueue {
1.22 foxr 598: $ConnectionRetriesLeft--; # Counts as connection failure too.
1.6 foxr 599: while($WorkQueue->Count()) {
1.10 foxr 600: my $request = $WorkQueue->dequeue(); # This is a transaction
1.7 foxr 601: FailTransaction($request);
1.6 foxr 602: }
603: }
604:
605: =pod
1.4 foxr 606:
1.9 foxr 607: =head2 CloseAllLondConnections
608:
609: Close all connections open on lond prior to exit e.g.
610:
611: =cut
612: sub CloseAllLondConnections {
1.23 foxr 613: foreach my $Socket (keys %ActiveConnections) {
1.9 foxr 614: KillSocket($Socket);
615: }
616: }
617: =cut
618:
619: =pod
620:
1.4 foxr 621: =head2 KillSocket
622:
623: Destroys a socket. This function can be called either when a socket
624: has died of 'natural' causes or because a socket needs to be pruned due to
625: idleness. If the socket has died naturally, if there are no longer any
626: live connections a new connection is created (in case there are transactions
627: in the queue). If the socket has been pruned, it is never re-created.
628:
629: Parameters:
1.1 foxr 630:
1.4 foxr 631: =item Socket
632:
633: The socket to kill off.
634:
635: =item Restart
636:
637: nonzero if we are allowed to create a new connection.
638:
639:
640: =cut
641: sub KillSocket {
642: my $Socket = shift;
643:
1.17 foxr 644: Log("WARNING", "Shutting down a socket");
1.9 foxr 645: $Socket->Shutdown();
646:
1.7 foxr 647: # If the socket came from the active connection set,
648: # delete its transaction... note that FailTransaction should
649: # already have been called!!!
650: # otherwise it came from the idle set.
651: #
1.4 foxr 652:
653: if(exists($ActiveTransactions{$Socket})) {
654: delete ($ActiveTransactions{$Socket});
655: }
656: if(exists($ActiveConnections{$Socket})) {
657: delete($ActiveConnections{$Socket});
658: }
659: $ConnectionCount--;
1.36 ! albertel 660: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
1.6 foxr 661: # If the connection count has gone to zero and there is work in the
662: # work queue, the work all gets failed with con_lost.
663: #
664: if($ConnectionCount == 0) {
1.22 foxr 665: EmptyQueue();
1.4 foxr 666: }
667: }
1.1 foxr 668:
669: =pod
1.3 albertel 670:
1.1 foxr 671: =head2 LondReadable
1.3 albertel 672:
1.1 foxr 673: This function is called whenever a lond connection
674: is readable. The action is state dependent:
675:
1.3 albertel 676: =head3 State=Initialized
677:
678: We''re waiting for the challenge, this is a no-op until the
1.1 foxr 679: state changes.
1.3 albertel 680:
1.1 foxr 681: =head3 State=Challenged
1.3 albertel 682:
683: The challenge has arrived we need to transition to Writable.
1.1 foxr 684: The connection must echo the challenge back.
1.3 albertel 685:
1.1 foxr 686: =head3 State=ChallengeReplied
1.3 albertel 687:
688: The challenge has been replied to. The we are receiveing the
1.1 foxr 689: 'ok' from the partner.
1.3 albertel 690:
1.1 foxr 691: =head3 State=RequestingKey
1.3 albertel 692:
693: The ok has been received and we need to send the request for
1.1 foxr 694: an encryption key. Transition to writable for that.
1.3 albertel 695:
1.1 foxr 696: =head3 State=ReceivingKey
1.3 albertel 697:
698: The the key has been requested, now we are reading the new key.
699:
1.1 foxr 700: =head3 State=Idle
1.3 albertel 701:
702: The encryption key has been negotiated or we have finished
1.1 foxr 703: reading data from the a transaction. If the callback data has
704: a client as well as the socket iformation, then we are
705: doing a transaction and the data received is relayed to the client
706: before the socket is put on the idle list.
1.3 albertel 707:
1.1 foxr 708: =head3 State=SendingRequest
1.3 albertel 709:
710: I do not think this state can be received here, but if it is,
1.1 foxr 711: the appropriate thing to do is to transition to writable, and send
712: the request.
1.3 albertel 713:
1.1 foxr 714: =head3 State=ReceivingReply
1.3 albertel 715:
716: We finished sending the request to the server and now transition
1.1 foxr 717: to readable to receive the reply.
718:
719: The parameter to this function are:
1.3 albertel 720:
1.1 foxr 721: The event. Implicit in this is the watcher and its data. The data
722: contains at least the lond connection object and, if a
723: transaction is in progress, the socket attached to the local client.
724:
1.3 albertel 725: =cut
1.1 foxr 726:
727: sub LondReadable {
1.8 foxr 728:
1.1 foxr 729: my $Event = shift;
730: my $Watcher = $Event->w;
731: my $Socket = $Watcher->data;
732: my $client = undef;
733:
1.23 foxr 734: &Debug(6,"LondReadable called state = ".$Socket->GetState());
1.8 foxr 735:
1.1 foxr 736:
737: my $State = $Socket->GetState(); # All action depends on the state.
738:
739: SocketDump(6, $Socket);
1.12 foxr 740: my $status = $Socket->Readable();
1.17 foxr 741:
1.12 foxr 742: &Debug(2, "Socket->Readable returned: $status");
1.1 foxr 743:
1.12 foxr 744: if($status != 0) {
1.4 foxr 745: # bad return from socket read. Currently this means that
746: # The socket has become disconnected. We fail the transaction.
747:
1.17 foxr 748: Log("WARNING",
749: "Lond connection lost.");
1.4 foxr 750: if(exists($ActiveTransactions{$Socket})) {
751: FailTransaction($ActiveTransactions{$Socket});
752: }
753: $Watcher->cancel();
1.6 foxr 754: KillSocket($Socket);
1.22 foxr 755: $ConnectionRetriesLeft--; # Counts as connection failure
1.4 foxr 756: return;
1.1 foxr 757: }
758: SocketDump(6,$Socket);
759:
760: $State = $Socket->GetState(); # Update in case of transition.
761: &Debug(6, "After read, state is ".$State);
762:
763: if($State eq "Initialized") {
764:
765:
766: } elsif ($State eq "ChallengeReceived") {
767: # The challenge must be echoed back; The state machine
768: # in the connection takes care of setting that up. Just
769: # need to transition to writable:
770:
1.8 foxr 771: $Watcher->cb(\&LondWritable);
1.1 foxr 772: $Watcher->poll("w");
773:
774: } elsif ($State eq "ChallengeReplied") {
775:
776:
777: } elsif ($State eq "RequestingKey") {
778: # The ok was received. Now we need to request the key
779: # That requires us to be writable:
780:
1.8 foxr 781: $Watcher->cb(\&LondWritable);
1.1 foxr 782: $Watcher->poll("w");
783:
784: } elsif ($State eq "ReceivingKey") {
785:
786: } elsif ($State eq "Idle") {
787: # If necessary, complete a transaction and then go into the
788: # idle queue.
1.22 foxr 789: # Note that a trasition to idle indicates a live lond
790: # on the other end so reset the connection retries.
791: #
792: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
1.8 foxr 793: $Watcher->cancel();
1.1 foxr 794: if(exists($ActiveTransactions{$Socket})) {
1.29 foxr 795: Debug(5,"Completing transaction!!");
1.1 foxr 796: CompleteTransaction($Socket,
797: $ActiveTransactions{$Socket});
1.9 foxr 798: } else {
799: Log("SUCCESS", "Connection ".$ConnectionCount." to "
800: .$RemoteHost." now ready for action");
1.1 foxr 801: }
802: ServerToIdle($Socket); # Next work unit or idle.
1.6 foxr 803:
1.1 foxr 804: } elsif ($State eq "SendingRequest") {
805: # We need to be writable for this and probably don't belong
806: # here inthe first place.
807:
808: Deubg(6, "SendingRequest state encountered in readable");
809: $Watcher->poll("w");
810: $Watcher->cb(\&LondWritable);
811:
812: } elsif ($State eq "ReceivingReply") {
813:
814:
815: } else {
816: # Invalid state.
817: Debug(4, "Invalid state in LondReadable");
818: }
819: }
1.3 albertel 820:
1.1 foxr 821: =pod
1.3 albertel 822:
1.1 foxr 823: =head2 LondWritable
1.3 albertel 824:
1.1 foxr 825: This function is called whenever a lond connection
826: becomes writable while there is a writeable monitoring
827: event. The action taken is very state dependent:
1.3 albertel 828:
1.1 foxr 829: =head3 State = Connected
1.3 albertel 830:
831: The connection is in the process of sending the 'init' hailing to the
832: lond on the remote end. The connection object''s Writable member is
833: called. On error, ConnectionError is called to destroy the connection
834: and remove it from the ActiveConnections hash
835:
1.1 foxr 836: =head3 Initialized
1.3 albertel 837:
838: 'init' has been sent, writability monitoring is removed and
839: readability monitoring is started with LondReadable as the callback.
840:
1.1 foxr 841: =head3 ChallengeReceived
1.3 albertel 842:
843: The connection has received the who are you challenge from the remote
844: system, and is in the process of sending the challenge
845: response. Writable is called.
846:
1.1 foxr 847: =head3 ChallengeReplied
1.3 albertel 848:
849: The connection has replied to the initial challenge The we switch to
850: monitoring readability looking for the server to reply with 'ok'.
851:
1.1 foxr 852: =head3 RequestingKey
1.3 albertel 853:
854: The connection is in the process of requesting its encryption key.
855: Writable is called.
856:
1.1 foxr 857: =head3 ReceivingKey
1.3 albertel 858:
859: The connection has sent the request for a key. Switch to readability
860: monitoring to accept the key
861:
1.1 foxr 862: =head3 SendingRequest
1.3 albertel 863:
864: The connection is in the process of sending a request to the server.
865: This request is part of a client transaction. All the states until
866: now represent the client setup protocol. Writable is called.
867:
1.1 foxr 868: =head3 ReceivingReply
869:
1.3 albertel 870: The connection has sent a request. Now it must receive a reply.
871: Readability monitoring is requested.
872:
873: This function is an event handler and therefore receives as
1.1 foxr 874: a parameter the event that has fired. The data for the watcher
875: of this event is a reference to a list of one or two elements,
876: depending on state. The first (and possibly only) element is the
877: socket. The second (present only if a request is in progress)
878: is the socket on which to return a reply to the caller.
879:
880: =cut
1.3 albertel 881:
1.1 foxr 882: sub LondWritable {
883: my $Event = shift;
884: my $Watcher = $Event->w;
1.8 foxr 885: my $Socket = $Watcher->data;
886: my $State = $Socket->GetState();
1.1 foxr 887:
1.8 foxr 888: Debug(6,"LondWritable State = ".$State."\n");
1.1 foxr 889:
1.8 foxr 890:
1.1 foxr 891: # Figure out what to do depending on the state of the socket:
892:
893:
894:
895:
896: SocketDump(6,$Socket);
897:
898: if ($State eq "Connected") {
899:
900: if ($Socket->Writable() != 0) {
901: # The write resulted in an error.
1.4 foxr 902: # We'll treat this as if the socket got disconnected:
1.9 foxr 903: Log("WARNING", "Connection to ".$RemoteHost.
904: " has been disconnected");
1.18 foxr 905: FailTransaction($ActiveTransactions{$Socket});
1.4 foxr 906: $Watcher->cancel();
1.6 foxr 907: KillSocket($Socket);
1.4 foxr 908: return;
1.1 foxr 909: }
1.4 foxr 910: # "init" is being sent...
911:
1.1 foxr 912:
913: } elsif ($State eq "Initialized") {
914:
915: # Now that init was sent, we switch
916: # to watching for readability:
917:
1.8 foxr 918: $Watcher->cb(\&LondReadable);
1.1 foxr 919: $Watcher->poll("r");
920:
921: } elsif ($State eq "ChallengeReceived") {
922: # We received the challenge, now we
923: # are echoing it back. This is a no-op,
924: # we're waiting for the state to change
925:
926: if($Socket->Writable() != 0) {
1.5 foxr 927:
928: $Watcher->cancel();
1.6 foxr 929: KillSocket($Socket);
1.5 foxr 930: return;
1.1 foxr 931: }
932:
933: } elsif ($State eq "ChallengeReplied") {
934: # The echo was sent back, so we switch
935: # to watching readability.
936:
1.8 foxr 937: $Watcher->cb(\&LondReadable);
1.1 foxr 938: $Watcher->poll("r");
939:
940: } elsif ($State eq "RequestingKey") {
941: # At this time we're requesting the key.
942: # again, this is essentially a no-op.
943: # we'll write the next chunk until the
944: # state changes.
945:
946: if($Socket->Writable() != 0) {
947: # Write resulted in an error.
1.5 foxr 948:
949: $Watcher->cancel();
1.6 foxr 950: KillSocket($Socket);
1.5 foxr 951: return;
952:
1.1 foxr 953: }
954: } elsif ($State eq "ReceivingKey") {
955: # Now we need to wait for the key
956: # to come back from the peer:
957:
1.8 foxr 958: $Watcher->cb(\&LondReadable);
1.1 foxr 959: $Watcher->poll("r");
960:
961: } elsif ($State eq "SendingRequest") {
962: # At this time we are sending a request to the
963: # peer... write the next chunk:
964:
965: if($Socket->Writable() != 0) {
966:
1.5 foxr 967: if(exists($ActiveTransactions{$Socket})) {
968: Debug(3, "Lond connection lost, failing transactions");
969: FailTransaction($ActiveTransactions{$Socket});
970: }
971: $Watcher->cancel();
1.6 foxr 972: KillSocket($Socket);
1.5 foxr 973: return;
974:
1.1 foxr 975: }
976:
977: } elsif ($State eq "ReceivingReply") {
978: # The send has completed. Wait for the
979: # data to come in for a reply.
980: Debug(8,"Writable sent request/receiving reply");
1.8 foxr 981: $Watcher->cb(\&LondReadable);
1.1 foxr 982: $Watcher->poll("r");
983:
984: } else {
985: # Control only passes here on an error:
986: # the socket state does not match any
987: # of the known states... so an error
988: # must be logged.
989:
990: &Debug(4, "Invalid socket state ".$State."\n");
991: }
992:
993: }
1.6 foxr 994: =pod
995:
996: =cut
997: sub QueueDelayed {
1.8 foxr 998: Debug(3,"QueueDelayed called");
999:
1.6 foxr 1000: my $path = "$perlvar{'lonSockDir'}/delayed";
1.8 foxr 1001:
1002: Debug(4, "Delayed path: ".$path);
1.6 foxr 1003: opendir(DIRHANDLE, $path);
1.8 foxr 1004:
1.23 foxr 1005: my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
1.6 foxr 1006: closedir(DIRHANDLE);
1007: my $dfname;
1.8 foxr 1008: my $reqfile;
1009: foreach $dfname (sort @alldelayed) {
1010: $reqfile = "$path/$dfname";
1011: Debug(4, "queueing ".$reqfile);
1.6 foxr 1012: my $Handle = IO::File->new($reqfile);
1013: my $cmd = <$Handle>;
1.8 foxr 1014: chomp $cmd; # There may or may not be a newline...
1.12 foxr 1015: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1.7 foxr 1016: my $Transaction = LondTransaction->new($cmd);
1017: $Transaction->SetDeferred($reqfile);
1018: QueueTransaction($Transaction);
1.6 foxr 1019: }
1020:
1021: }
1.1 foxr 1022:
1023: =pod
1.3 albertel 1024:
1.1 foxr 1025: =head2 MakeLondConnection
1.3 albertel 1026:
1027: Create a new lond connection object, and start it towards its initial
1028: idleness. Once idle, it becomes elligible to receive transactions
1029: from the work queue. If the work queue is not empty when the
1030: connection is completed and becomes idle, it will dequeue an entry and
1031: start off on it.
1032:
1.1 foxr 1033: =cut
1.3 albertel 1034:
1.1 foxr 1035: sub MakeLondConnection {
1036: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1037: .GetServerPort());
1038:
1039: my $Connection = LondConnection->new(&GetServerHost(),
1040: &GetServerPort());
1041:
1.30 foxr 1042: if($Connection eq undef) { # Needs to be more robust later.
1.9 foxr 1043: Log("CRITICAL","Failed to make a connection with lond.");
1.10 foxr 1044: $ConnectionRetriesLeft--;
1045: return 0; # Failure.
1.5 foxr 1046: } else {
1.22 foxr 1047:
1.5 foxr 1048: # The connection needs to have writability
1049: # monitored in order to send the init sequence
1050: # that starts the whole authentication/key
1051: # exchange underway.
1052: #
1053: my $Socket = $Connection->GetSocket();
1.30 foxr 1054: if($Socket eq undef) {
1.5 foxr 1055: die "did not get a socket from the connection";
1056: } else {
1057: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1058: }
1.1 foxr 1059:
1.21 foxr 1060: $Connection->SetTimeoutCallback(\&SocketTimeout);
1061:
1.23 foxr 1062: my $event = Event->io(fd => $Socket,
1.5 foxr 1063: poll => 'w',
1064: cb => \&LondWritable,
1.8 foxr 1065: data => $Connection,
1.5 foxr 1066: desc => 'Connection to lond server');
1067: $ActiveConnections{$Connection} = $event;
1068:
1069: $ConnectionCount++;
1.8 foxr 1070: Debug(4, "Connection count = ".$ConnectionCount);
1.6 foxr 1071: if($ConnectionCount == 1) { # First Connection:
1072: QueueDelayed;
1073: }
1.9 foxr 1074: Log("SUCESS", "Created connection ".$ConnectionCount
1075: ." to host ".GetServerHost());
1.10 foxr 1076: return 1; # Return success.
1.1 foxr 1077: }
1078:
1079: }
1.3 albertel 1080:
1.1 foxr 1081: =pod
1.3 albertel 1082:
1.1 foxr 1083: =head2 StartRequest
1.3 albertel 1084:
1085: Starts a lond request going on a specified lond connection.
1086: parameters are:
1087:
1088: =item $Lond
1089:
1090: Connection to the lond that will send the transaction and receive the
1091: reply.
1092:
1093: =item $Client
1094:
1095: Connection to the client that is making this request We got the
1096: request from this socket, and when the request has been relayed to
1097: lond and we get a reply back from lond it will get sent to this
1098: socket.
1099:
1100: =item $Request
1101:
1102: The text of the request to send.
1103:
1.1 foxr 1104: =cut
1105:
1106: sub StartRequest {
1107: my $Lond = shift;
1.7 foxr 1108: my $Request = shift; # This is a LondTransaction.
1.1 foxr 1109:
1.7 foxr 1110: Debug(6, "StartRequest: ".$Request->getRequest());
1.1 foxr 1111:
1112: my $Socket = $Lond->GetSocket();
1113:
1.7 foxr 1114: $Request->Activate($Lond);
1115: $ActiveTransactions{$Lond} = $Request;
1.1 foxr 1116:
1.7 foxr 1117: $Lond->InitiateTransaction($Request->getRequest());
1.23 foxr 1118: my $event = Event->io(fd => $Socket,
1.1 foxr 1119: poll => "w",
1120: cb => \&LondWritable,
1121: data => $Lond,
1122: desc => "lond transaction connection");
1123: $ActiveConnections{$Lond} = $event;
1124: Debug(8," Start Request made watcher data with ".$event->data."\n");
1125: }
1126:
1127: =pod
1.3 albertel 1128:
1.1 foxr 1129: =head2 QueueTransaction
1.3 albertel 1130:
1131: If there is an idle lond connection, it is put to work doing this
1132: transaction. Otherwise, the transaction is placed in the work queue.
1133: If placed in the work queue and the maximum number of connections has
1134: not yet been created, a new connection will be started. Our goal is
1135: to eventually have a sufficient number of connections that the work
1136: queue will typically be empty. parameters are:
1137:
1138: =item Socket
1139:
1140: open on the lonc client.
1141:
1142: =item Request
1143:
1144: data to send to the lond.
1.1 foxr 1145:
1146: =cut
1.3 albertel 1147:
1.1 foxr 1148: sub QueueTransaction {
1149:
1.7 foxr 1150: my $requestData = shift; # This is a LondTransaction.
1151: my $cmd = $requestData->getRequest();
1152:
1153: Debug(6,"QueueTransaction: ".$cmd);
1.1 foxr 1154:
1155: my $LondSocket = $IdleConnections->pop();
1156: if(!defined $LondSocket) { # Need to queue request.
1.29 foxr 1157: Debug(5,"Must queue...");
1.1 foxr 1158: $WorkQueue->enqueue($requestData);
1159: if($ConnectionCount < $MaxConnectionCount) {
1.22 foxr 1160: if($ConnectionRetriesLeft > 0) {
1.29 foxr 1161: Debug(5,"Starting additional lond connection");
1.22 foxr 1162: if(MakeLondConnection() == 0) {
1163: EmptyQueue(); # Fail transactions, can't make connection.
1164: }
1165: } else {
1166: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1167: EmptyQueue(); # It's worse than that ... he's dead Jim.
1.17 foxr 1168: }
1.1 foxr 1169: }
1170: } else { # Can start the request:
1171: Debug(8,"Can start...");
1.7 foxr 1172: StartRequest($LondSocket, $requestData);
1.1 foxr 1173: }
1174: }
1175:
1176: #-------------------------- Lonc UNIX socket handling ---------------------
1.3 albertel 1177:
1.1 foxr 1178: =pod
1.3 albertel 1179:
1.1 foxr 1180: =head2 ClientRequest
1.3 albertel 1181: Callback that is called when data can be read from the UNIX domain
1182: socket connecting us with an apache server process.
1.1 foxr 1183:
1184: =cut
1185:
1186: sub ClientRequest {
1187: Debug(6, "ClientRequest");
1188: my $event = shift;
1189: my $watcher = $event->w;
1190: my $socket = $watcher->fd;
1191: my $data = $watcher->data;
1192: my $thisread;
1193:
1194: Debug(9, " Watcher named: ".$watcher->desc);
1195:
1196: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1197: Debug(8, "rcv: data length = ".length($thisread)
1198: ." read =".$thisread);
1.29 foxr 1199: unless (defined $rv && length($thisread)) {
1.1 foxr 1200: # Likely eof on socket.
1201: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1202: close($socket);
1203: $watcher->cancel();
1204: delete($ActiveClients{$socket});
1.10 foxr 1205: return;
1.1 foxr 1206: }
1207: Debug(8,"Data: ".$data." this read: ".$thisread);
1208: $data = $data.$thisread; # Append new data.
1209: $watcher->data($data);
1210: if($data =~ /(.*\n)/) { # Request entirely read.
1.10 foxr 1211: if($data eq "close_connection_exit\n") {
1.9 foxr 1212: Log("CRITICAL",
1213: "Request Close Connection ... exiting");
1214: CloseAllLondConnections();
1215: exit;
1216: }
1.1 foxr 1217: Debug(8, "Complete transaction received: ".$data);
1.8 foxr 1218: my $Transaction = LondTransaction->new($data);
1.7 foxr 1219: $Transaction->SetClient($socket);
1220: QueueTransaction($Transaction);
1.1 foxr 1221: $watcher->cancel(); # Done looking for input data.
1222: }
1223:
1224: }
1225:
1226:
1227: =pod
1.3 albertel 1228:
1.1 foxr 1229: =head2 NewClient
1.3 albertel 1230:
1231: Callback that is called when a connection is received on the unix
1232: socket for a new client of lonc. The callback is parameterized by the
1233: event.. which is a-priori assumed to be an io event, and therefore has
1234: an fd member that is the Listener socket. We Accept the connection
1235: and register a new event on the readability of that socket:
1236:
1.1 foxr 1237: =cut
1.3 albertel 1238:
1.1 foxr 1239: sub NewClient {
1240: Debug(6, "NewClient");
1241: my $event = shift; # Get the event parameters.
1242: my $watcher = $event->w;
1243: my $socket = $watcher->fd; # Get the event' socket.
1244: my $connection = $socket->accept(); # Accept the client connection.
1245: Debug(8,"Connection request accepted from "
1246: .GetPeername($connection, AF_UNIX));
1247:
1248:
1249: my $description = sprintf("Connection to lonc client %d",
1250: $ClientConnection);
1251: Debug(9, "Creating event named: ".$description);
1252: Event->io(cb => \&ClientRequest,
1253: poll => 'r',
1254: desc => $description,
1255: data => "",
1256: fd => $connection);
1257: $ActiveClients{$connection} = $ClientConnection;
1258: $ClientConnection++;
1259: }
1.3 albertel 1260:
1261: =pod
1262:
1263: =head2 GetLoncSocketPath
1264:
1265: Returns the name of the UNIX socket on which to listen for client
1266: connections.
1.1 foxr 1267:
1268: =cut
1.3 albertel 1269:
1.1 foxr 1270: sub GetLoncSocketPath {
1271: return $UnixSocketDir."/".GetServerHost();
1272: }
1273:
1.3 albertel 1274: =pod
1275:
1276: =head2 GetServerHost
1277:
1278: Returns the host whose lond we talk with.
1279:
1.1 foxr 1280: =cut
1.3 albertel 1281:
1.7 foxr 1282: sub GetServerHost {
1.1 foxr 1283: return $RemoteHost; # Setup by the fork.
1284: }
1.3 albertel 1285:
1286: =pod
1287:
1288: =head2 GetServerPort
1289:
1290: Returns the lond port number.
1291:
1.1 foxr 1292: =cut
1.3 albertel 1293:
1.7 foxr 1294: sub GetServerPort {
1.1 foxr 1295: return $perlvar{londPort};
1296: }
1.3 albertel 1297:
1298: =pod
1299:
1300: =head2 SetupLoncListener
1301:
1302: Setup a lonc listener event. The event is called when the socket
1303: becomes readable.. that corresponds to the receipt of a new
1304: connection. The event handler established will accept the connection
1305: (creating a communcations channel), that int turn will establish
1306: another event handler to subess requests.
1.1 foxr 1307:
1308: =cut
1.3 albertel 1309:
1.1 foxr 1310: sub SetupLoncListener {
1311:
1312: my $socket;
1313: my $SocketName = GetLoncSocketPath();
1314: unlink($SocketName);
1.7 foxr 1315: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1.1 foxr 1316: Listen => 10,
1317: Type => SOCK_STREAM)) {
1318: die "Failed to create a lonc listner socket";
1319: }
1320: Event->io(cb => \&NewClient,
1321: poll => 'r',
1322: desc => 'Lonc listener Unix Socket',
1323: fd => $socket);
1324: }
1325:
1.14 foxr 1326: =pod
1327:
1328: =head2 ChildStatus
1329:
1330: Child USR1 signal handler to report the most recent status
1331: into the status file.
1332:
1.22 foxr 1333: We also use this to reset the retries count in order to allow the
1334: client to retry connections with a previously dead server.
1.14 foxr 1335: =cut
1336: sub ChildStatus {
1337: my $event = shift;
1338: my $watcher = $event->w;
1339:
1340: Debug(2, "Reporting child status because : ".$watcher->data);
1341: my $docdir = $perlvar{'lonDocRoot'};
1342: my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
1343: print $fh $$."\t".$RemoteHost."\t".$Status."\t".
1344: $RecentLogEntry."\n";
1.22 foxr 1345: $ConnectionRetriesLeft = $ConnectionRetries;
1.14 foxr 1346: }
1347:
1.1 foxr 1348: =pod
1.3 albertel 1349:
1.10 foxr 1350: =head2 SignalledToDeath
1351:
1352: Called in response to a signal that causes a chid process to die.
1353:
1354: =cut
1355:
1356:
1357: sub SignalledToDeath {
1.14 foxr 1358: my $event = shift;
1359: my $watcher= $event->w;
1360:
1361: Debug(2,"Signalled to death! via ".$watcher->data);
1.17 foxr 1362: my ($signal) = $watcher->data;
1.10 foxr 1363: chomp($signal);
1364: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1365: ."died through "."\"$signal\"");
1366: LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1367: ."\"$signal\"");
1.12 foxr 1368: exit 0;
1.10 foxr 1369:
1370: }
1.16 foxr 1371:
1372: =head2 ToggleDebug
1373:
1374: This sub toggles trace debugging on and off.
1375:
1376: =cut
1377:
1378: sub ToggleDebug {
1379: my $Current = $DebugLevel;
1380: $DebugLevel = $NextDebugLevel;
1381: $NextDebugLevel = $Current;
1382:
1383: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1384:
1385: }
1386:
1.1 foxr 1387: =head2 ChildProcess
1388:
1389: This sub implements a child process for a single lonc daemon.
1390:
1391: =cut
1392:
1393: sub ChildProcess {
1394:
1395:
1.14 foxr 1396: #
1397: # Signals must be handled by the Event framework...
1398: #
1399:
1400: Event->signal(signal => "QUIT",
1401: cb => \&SignalledToDeath,
1402: data => "QUIT");
1403: Event->signal(signal => "HUP",
1404: cb => \&ChildStatus,
1405: data => "HUP");
1406: Event->signal(signal => "USR1",
1407: cb => \&ChildStatus,
1408: data => "USR1");
1.16 foxr 1409: Event->signal(signal => "INT",
1410: cb => \&ToggleDebug,
1411: data => "INT");
1.1 foxr 1412:
1413: SetupTimer();
1414:
1415: SetupLoncListener();
1416:
1417: $Event::Debuglevel = $DebugLevel;
1418:
1419: Debug(9, "Making initial lond connection for ".$RemoteHost);
1420:
1421: # Setup the initial server connection:
1422:
1.14 foxr 1423: # &MakeLondConnection(); // let first work requirest do it.
1.10 foxr 1424:
1.5 foxr 1425:
1.1 foxr 1426: Debug(9,"Entering event loop");
1427: my $ret = Event::loop(); # Start the main event loop.
1428:
1429:
1430: die "Main event loop exited!!!";
1431: }
1432:
1433: # Create a new child for host passed in:
1434:
1435: sub CreateChild {
1.12 foxr 1436: my $sigset = POSIX::SigSet->new(SIGINT);
1437: sigprocmask(SIG_BLOCK, $sigset);
1.1 foxr 1438: my $host = shift;
1439: $RemoteHost = $host;
1.9 foxr 1440: Log("CRITICAL", "Forking server for ".$host);
1.23 foxr 1441: my $pid = fork;
1.1 foxr 1442: if($pid) { # Parent
1.17 foxr 1443: $RemoteHost = "Parent";
1.27 foxr 1444: $ChildHash{$pid} = $host;
1.26 foxr 1445: $HostToPid{$host}= $pid;
1.12 foxr 1446: sigprocmask(SIG_UNBLOCK, $sigset);
1447:
1.1 foxr 1448: } else { # child.
1.5 foxr 1449: ShowStatus("Connected to ".$RemoteHost);
1.23 foxr 1450: $SIG{INT} = 'DEFAULT';
1.12 foxr 1451: sigprocmask(SIG_UNBLOCK, $sigset);
1452: ChildProcess; # Does not return.
1.1 foxr 1453: }
1454:
1455: }
1456: #
1457: # Parent process logic pass 1:
1458: # For each entry in the hosts table, we will
1459: # fork off an instance of ChildProcess to service the transactions
1460: # to that host. Each pid will be entered in a global hash
1461: # with the value of the key, the host.
1462: # The parent will then enter a loop to wait for process exits.
1463: # Each exit gets logged and the child gets restarted.
1464: #
1465:
1.5 foxr 1466: #
1467: # Fork and start in new session so hang-up isn't going to
1468: # happen without intent.
1469: #
1470:
1471:
1.6 foxr 1472:
1473:
1.8 foxr 1474:
1.6 foxr 1475:
1476: ShowStatus("Forming new session");
1477: my $childpid = fork;
1478: if ($childpid != 0) {
1479: sleep 4; # Give child a chacne to break to
1480: exit 0; # a new sesion.
1481: }
1.8 foxr 1482: #
1483: # Write my pid into the pid file so I can be located
1484: #
1485:
1486: ShowStatus("Parent writing pid file:");
1.23 foxr 1487: my $execdir = $perlvar{'lonDaemons'};
1.8 foxr 1488: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1489: print PIDSAVE "$$\n";
1490: close(PIDSAVE);
1.6 foxr 1491:
1.17 foxr 1492:
1493:
1.6 foxr 1494: if (POSIX::setsid() < 0) {
1495: print "Could not create new session\n";
1496: exit -1;
1497: }
1.5 foxr 1498:
1499: ShowStatus("Forking node servers");
1500:
1.9 foxr 1501: Log("CRITICAL", "--------------- Starting children ---------------");
1502:
1.31 foxr 1503: LondConnection::ReadConfig; # Read standard config files.
1.1 foxr 1504: my $HostIterator = LondConnection::GetHostIterator;
1505: while (! $HostIterator->end()) {
1506:
1.23 foxr 1507: my $hostentryref = $HostIterator->get();
1.1 foxr 1508: CreateChild($hostentryref->[0]);
1.26 foxr 1509: $HostHash{$hostentryref->[0]} = $hostentryref->[4];
1.1 foxr 1510: $HostIterator->next();
1511: }
1.12 foxr 1512: $RemoteHost = "Parent Server";
1.1 foxr 1513:
1514: # Maintain the population:
1.5 foxr 1515:
1516: ShowStatus("Parent keeping the flock");
1.1 foxr 1517:
1.10 foxr 1518: #
1519: # Set up parent signals:
1520: #
1.12 foxr 1521:
1.14 foxr 1522: $SIG{INT} = \&Terminate;
1523: $SIG{TERM} = \&Terminate;
1.13 foxr 1524: $SIG{HUP} = \&Restart;
1.14 foxr 1525: $SIG{USR1} = \&CheckKids;
1.24 foxr 1526: $SIG{USR2} = \&UpdateKids; # LonManage update request.
1.10 foxr 1527:
1.1 foxr 1528: while(1) {
1.23 foxr 1529: my $deadchild = wait();
1.1 foxr 1530: if(exists $ChildHash{$deadchild}) { # need to restart.
1.23 foxr 1531: my $deadhost = $ChildHash{$deadchild};
1.26 foxr 1532: delete($HostToPid{$deadhost});
1.1 foxr 1533: delete($ChildHash{$deadchild});
1.9 foxr 1534: Log("WARNING","Lost child pid= ".$deadchild.
1.1 foxr 1535: "Connected to host ".$deadhost);
1.9 foxr 1536: Log("INFO", "Restarting child procesing ".$deadhost);
1.1 foxr 1537: CreateChild($deadhost);
1538: }
1.13 foxr 1539: }
1540:
1.14 foxr 1541:
1542:
1543: =pod
1544:
1545: =head1 CheckKids
1546:
1547: Since kids do not die as easily in this implementation
1548: as the previous one, there is no need to restart the
1549: dead ones (all dead kids get restarted when they die!!)
1550: The only thing this function does is to pass USR1 to the
1551: kids so that they report their status.
1552:
1553: =cut
1554:
1555: sub CheckKids {
1556: Debug(2, "Checking status of children");
1557: my $docdir = $perlvar{'lonDocRoot'};
1558: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
1559: my $now=time;
1560: my $local=localtime($now);
1561: print $fh "LONC status $local - parent $$ \n\n";
1.23 foxr 1562: foreach my $pid (keys %ChildHash) {
1.14 foxr 1563: Debug(2, "Sending USR1 -> $pid");
1564: kill 'USR1' => $pid; # Tell Child to report status.
1565: sleep 1; # Wait so file doesn't intermix.
1566: }
1567: }
1.24 foxr 1568:
1569: =pod
1570:
1571: =head1 UpdateKids
1572:
1.25 foxr 1573: parent's SIGUSR2 handler. This handler:
1.24 foxr 1574:
1575: =item
1576:
1577: Rereads the hosts file.
1578:
1579: =item
1580:
1581: Kills off (via sigint) children for hosts that have disappeared.
1582:
1583: =item
1584:
1.27 foxr 1585: QUITs children for hosts that already exist (this just forces a status display
1.24 foxr 1586: and resets the connection retry count for that host.
1587:
1588: =item
1589:
1590: Starts new children for hosts that have been added to the hosts.tab file since
1591: the start of the master program and maintains them.
1592:
1593: =cut
1594:
1595: sub UpdateKids {
1.27 foxr 1596:
1.25 foxr 1597: Log("INFO", "Updating connections via SIGUSR2");
1.27 foxr 1598:
1599: # Just in case we need to kill our own lonc, we wait a few seconds to
1600: # give it a chance to receive and relay lond's response to the
1601: # re-init command.
1602: #
1603:
1604: sleep(2); # Wait a couple of seconds.
1605:
1606: my %hosts; # Indexed by loncapa hostname, value=ip.
1607:
1608: # Need to re-read the host table:
1609:
1610:
1611: LondConnection::ReadConfig();
1612: my $I = LondConnection::GetHostIterator;
1613: while (! $I->end()) {
1614: my $item = $I->get();
1615: $hosts{$item->[0]} = $item->[4];
1616: $I->next();
1617: }
1618:
1619: # The logic below is written for clarity not for efficiency.
1620: # Since I anticipate that this function is only rarely called, that's
1621: # appropriate. There are certainly ways to combine the loops below,
1622: # and anyone wishing to obscure the logic is welcome to go for it.
1623: # Note that we don't re-direct sigchild. Instead we do what's needed
1624: # to the data structures that keep track of children to ensure that
1625: # when sigchild is honored, no new child is born.
1626: #
1627:
1628: # For each existing child; if it's host doesn't exist, kill the child.
1629:
1630: foreach my $child (keys %ChildHash) {
1631: my $oldhost = $ChildHash{$child};
1632: if (!(exists $hosts{$oldhost})) {
1633: Log("CRITICAL", "Killing child for $oldhost host no longer exists");
1634: delete $ChildHash{$child};
1635: delete $HostToPid{$oldhost};
1636: kill 'QUIT' => $child;
1637: }
1638: }
1639: # For each remaining existing child; if it's host's ip has changed,
1640: # Restart the child on the new IP.
1641:
1642: foreach my $child (keys %ChildHash) {
1643: my $oldhost = $ChildHash{$child};
1644: my $oldip = $HostHash{$oldhost};
1645: if ($hosts{$oldhost} ne $oldip) {
1646:
1647: # kill the old child.
1648:
1649: Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
1650: delete $ChildHash{$child};
1651: delete $HostToPid{$oldhost};
1652: kill 'QUIT' => $child;
1653:
1654: # Do the book-keeping needed to start a new child on the
1655: # new ip.
1656:
1657: $HostHash{$oldhost} = $hosts{$oldhost};
1658: CreateChild($oldhost);
1659: }
1660: }
1661: # Finally, for each new host, not in the host hash, create a
1662: # enter the host and create a new child.
1663: # Force a status display of any existing process.
1664:
1665: foreach my $host (keys %hosts) {
1666: if(!(exists $HostHash{$host})) {
1667: Log("INFO", "New host $host discovered in hosts.tab...");
1668: $HostHash{$host} = $hosts{$host};
1669: CreateChild($host);
1670: } else {
1671: kill 'HUP' => $HostToPid{$host}; # status display.
1672: }
1673: }
1.24 foxr 1674: }
1675:
1.14 foxr 1676:
1.13 foxr 1677: =pod
1678:
1679: =head1 Restart
1680:
1681: Signal handler for HUP... all children are killed and
1682: we self restart. This is an el-cheapo way to re read
1683: the config file.
1684:
1685: =cut
1686:
1687: sub Restart {
1.23 foxr 1688: &KillThemAll; # First kill all the children.
1.13 foxr 1689: Log("CRITICAL", "Restarting");
1690: my $execdir = $perlvar{'lonDaemons'};
1691: unlink("$execdir/logs/lonc.pid");
1.28 albertel 1692: exec("$execdir/loncnew");
1.10 foxr 1693: }
1.12 foxr 1694:
1695: =pod
1696:
1697: =head1 KillThemAll
1698:
1699: Signal handler that kills all children by sending them a
1.17 foxr 1700: SIGHUP. Responds to sigint and sigterm.
1.12 foxr 1701:
1702: =cut
1703:
1.10 foxr 1704: sub KillThemAll {
1.12 foxr 1705: Debug(2, "Kill them all!!");
1706: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
1.23 foxr 1707: foreach my $pid (keys %ChildHash) {
1.12 foxr 1708: my $serving = $ChildHash{$pid};
1709: Debug(2, "Killing lonc for $serving pid = $pid");
1710: ShowStatus("Killing lonc for $serving pid = $pid");
1711: Log("CRITICAL", "Killing lonc for $serving pid = $pid");
1.17 foxr 1712: kill 'QUIT' => $pid;
1713: delete($ChildHash{$pid});
1.12 foxr 1714: }
1.14 foxr 1715: my $execdir = $perlvar{'lonDaemons'};
1716: unlink("$execdir/logs/lonc.pid");
1.17 foxr 1717:
1.1 foxr 1718: }
1.12 foxr 1719:
1.14 foxr 1720: =pod
1721:
1722: =head1 Terminate
1723:
1724: Terminate the system.
1725:
1726: =cut
1727:
1728: sub Terminate {
1729: KillThemAll;
1.17 foxr 1730: Log("CRITICAL","Master process exiting");
1731: exit 0;
1.14 foxr 1732:
1733: }
1.12 foxr 1734: =pod
1.1 foxr 1735:
1736: =head1 Theory
1.3 albertel 1737:
1738: The event class is used to build this as a single process with an
1739: event driven model. The following events are handled:
1.1 foxr 1740:
1741: =item UNIX Socket connection Received
1742:
1743: =item Request data arrives on UNIX data transfer socket.
1744:
1745: =item lond connection becomes writable.
1746:
1747: =item timer fires at 1 second intervals.
1748:
1749: All sockets are run in non-blocking mode. Timeouts managed by the timer
1750: handler prevents hung connections.
1751:
1752: Key data structures:
1753:
1.3 albertel 1754: =item RequestQueue
1755:
1756: A queue of requests received from UNIX sockets that are
1757: waiting for a chance to be forwarded on a lond connection socket.
1758:
1759: =item ActiveConnections
1760:
1761: A hash of lond connections that have transactions in process that are
1762: available to be timed out.
1763:
1764: =item ActiveTransactions
1765:
1766: A hash indexed by lond connections that contain the client reply
1767: socket for each connection that has an active transaction on it.
1768:
1769: =item IdleConnections
1770:
1771: A hash of lond connections that have no work to do. These connections
1772: can be closed if they are idle for a long enough time.
1.1 foxr 1773:
1774: =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