Annotation of loncom/loncnew, revision 1.95
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: #
1.95 ! foxr 5: # $Id: loncnew,v 1.94 2011/01/20 11:19:37 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).
1.38 foxr 38: # - Add logging/status monitoring. (done)
39: # - Add Signal handling - HUP restarts. USR1 status report. (done)
1.7 foxr 40: # - Add Configuration file I/O (done).
1.38 foxr 41: # - Add management/status request interface. (done)
1.8 foxr 42: # - Add deferred request capability. (done)
1.38 foxr 43: # - Detect transmission timeouts. (done)
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;
1.67 albertel 63: use Fcntl qw(:flock);
1.1 foxr 64:
65:
66: # Read the httpd configuration file to get perl variables
67: # normally set in apache modules:
68:
69: my $perlvarref = LONCAPA::Configuration::read_conf('loncapa.conf');
70: my %perlvar = %{$perlvarref};
71:
72: #
73: # parent and shared variables.
74:
1.83 albertel 75: my %ChildPid; # by pid -> host.
76: my %ChildHost; # by host.
1.62 foxr 77: my %listening_to; # Socket->host table for who the parent
78: # is listening to.
79: my %parent_dispatchers; # host-> listener watcher events.
1.1 foxr 80:
1.65 foxr 81: my %parent_handlers; # Parent signal handlers...
82:
1.9 foxr 83: my $MaxConnectionCount = 10; # Will get from config later.
1.1 foxr 84: my $ClientConnection = 0; # Uniquifier for client events.
85:
1.9 foxr 86: my $DebugLevel = 0;
1.29 foxr 87: my $NextDebugLevel= 2; # So Sigint can toggle this.
1.94 foxr 88: my $IdleTimeout= 5*60; # Seconds to wait prior to pruning connections.
1.1 foxr 89:
1.39 foxr 90: my $LogTransactions = 0; # When True, all transactions/replies get logged.
1.65 foxr 91: my $executable = $0; # Get the full path to me.
1.39 foxr 92:
1.1 foxr 93: #
94: # The variables below are only used by the child processes.
95: #
96: my $RemoteHost; # Name of host child is talking to.
1.81 albertel 97: my $RemoteHostId; # default lonid of host child is talking to.
1.82 albertel 98: my @all_host_ids;
1.20 albertel 99: my $UnixSocketDir= $perlvar{'lonSockDir'};
1.1 foxr 100: my $IdleConnections = Stack->new(); # Set of idle connections
101: my %ActiveConnections; # Connections to the remote lond.
1.7 foxr 102: my %ActiveTransactions; # LondTransactions in flight.
1.1 foxr 103: my %ActiveClients; # Serial numbers of active clients by socket.
104: my $WorkQueue = Queue->new(); # Queue of pending transactions.
105: my $ConnectionCount = 0;
1.4 foxr 106: my $IdleSeconds = 0; # Number of seconds idle.
1.9 foxr 107: my $Status = ""; # Current status string.
1.14 foxr 108: my $RecentLogEntry = "";
1.72 albertel 109: my $ConnectionRetries=5; # Number of connection retries allowed.
110: my $ConnectionRetriesLeft=5; # Number of connection retries remaining.
1.40 foxr 111: my $LondVersion = "unknown"; # Version of lond we talk with.
1.49 foxr 112: my $KeyMode = ""; # e.g. ssl, local, insecure from last connect.
1.54 foxr 113: my $LondConnecting = 0; # True when a connection is being built.
1.1 foxr 114:
1.60 foxr 115:
116:
1.62 foxr 117: my $I_am_child = 0; # True if this is the child process.
1.57 foxr 118:
1.1 foxr 119: #
1.9 foxr 120: # The hash below gives the HTML format for log messages
121: # given a severity.
122: #
123: my %LogFormats;
124:
1.45 albertel 125: $LogFormats{"CRITICAL"} = "<font color='red'>CRITICAL: %s</font>";
126: $LogFormats{"SUCCESS"} = "<font color='green'>SUCCESS: %s</font>";
127: $LogFormats{"INFO"} = "<font color='yellow'>INFO: %s</font>";
128: $LogFormats{"WARNING"} = "<font color='blue'>WARNING: %s</font>";
1.9 foxr 129: $LogFormats{"DEFAULT"} = " %s ";
130:
1.10 foxr 131:
1.57 foxr 132: # UpdateStatus;
133: # Update the idle status display to show how many connections
134: # are left, retries and other stuff.
135: #
136: sub UpdateStatus {
137: if ($ConnectionRetriesLeft > 0) {
138: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
139: ." Retries remaining: ".$ConnectionRetriesLeft
140: ." ($KeyMode)");
141: } else {
142: ShowStatus(GetServerHost()." >> DEAD <<");
143: }
144: }
145:
1.10 foxr 146:
147: =pod
148:
149: =head2 LogPerm
150:
151: Makes an entry into the permanent log file.
152:
153: =cut
1.69 matthew 154:
1.10 foxr 155: sub LogPerm {
156: my $message=shift;
157: my $execdir=$perlvar{'lonDaemons'};
158: my $now=time;
159: my $local=localtime($now);
160: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
1.86 albertel 161: chomp($message);
1.10 foxr 162: print $fh "$now:$message:$local\n";
163: }
1.9 foxr 164:
165: =pod
166:
167: =head2 Log
168:
169: Logs a message to the log file.
170: Parameters:
171:
172: =item severity
173:
174: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
175: format string used to format the message. if the severity is
176: not a defined severity the Default format string is used.
177:
178: =item message
179:
180: The base message. In addtion to the format string, the message
181: will be appended to a string containing the name of our remote
182: host and the time will be formatted into the message.
183:
184: =cut
185:
186: sub Log {
1.47 foxr 187:
188: my ($severity, $message) = @_;
189:
1.9 foxr 190: if(!$LogFormats{$severity}) {
191: $severity = "DEFAULT";
192: }
193:
194: my $format = $LogFormats{$severity};
195:
196: # Put the window dressing in in front of the message format:
197:
198: my $now = time;
199: my $local = localtime($now);
200: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
1.76 albertel 201: $finalformat = $finalformat.$format."\n";
1.9 foxr 202:
203: # open the file and put the result.
204:
205: my $execdir = $perlvar{'lonDaemons'};
206: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
207: my $msg = sprintf($finalformat, $message);
1.14 foxr 208: $RecentLogEntry = $msg;
1.9 foxr 209: print $fh $msg;
210:
1.10 foxr 211:
1.9 foxr 212: }
1.6 foxr 213:
1.3 albertel 214:
1.1 foxr 215: =pod
1.3 albertel 216:
217: =head2 GetPeerName
218:
219: Returns the name of the host that a socket object is connected to.
220:
1.1 foxr 221: =cut
222:
223: sub GetPeername {
1.47 foxr 224:
225:
226: my ($connection, $AdrFamily) = @_;
227:
1.1 foxr 228: my $peer = $connection->peername();
229: my $peerport;
230: my $peerip;
231: if($AdrFamily == AF_INET) {
232: ($peerport, $peerip) = sockaddr_in($peer);
1.23 foxr 233: my $peername = gethostbyaddr($peerip, $AdrFamily);
1.1 foxr 234: return $peername;
235: } elsif ($AdrFamily == AF_UNIX) {
236: my $peerfile;
237: ($peerfile) = sockaddr_un($peer);
238: return $peerfile;
239: }
240: }
241: =pod
1.3 albertel 242:
1.1 foxr 243: =head2 Debug
1.3 albertel 244:
245: Invoked to issue a debug message.
246:
1.1 foxr 247: =cut
1.3 albertel 248:
1.1 foxr 249: sub Debug {
1.47 foxr 250:
251: my ($level, $message) = @_;
252:
1.1 foxr 253: if ($level <= $DebugLevel) {
1.23 foxr 254: Log("INFO", "-Debug- $message host = $RemoteHost");
1.1 foxr 255: }
256: }
257:
258: sub SocketDump {
1.47 foxr 259:
260: my ($level, $socket) = @_;
261:
1.1 foxr 262: if($level <= $DebugLevel) {
1.48 foxr 263: $socket->Dump(-1); # Ensure it will get dumped.
1.1 foxr 264: }
265: }
1.3 albertel 266:
1.1 foxr 267: =pod
1.3 albertel 268:
1.5 foxr 269: =head2 ShowStatus
270:
271: Place some text as our pid status.
1.10 foxr 272: and as what we return in a SIGUSR1
1.5 foxr 273:
274: =cut
1.69 matthew 275:
1.5 foxr 276: sub ShowStatus {
1.10 foxr 277: my $state = shift;
278: my $now = time;
279: my $local = localtime($now);
280: $Status = $local.": ".$state;
281: $0='lonc: '.$state.' '.$local;
1.5 foxr 282: }
283:
284: =pod
285:
1.69 matthew 286: =head2 SocketTimeout
1.15 foxr 287:
288: Called when an action on the socket times out. The socket is
289: destroyed and any active transaction is failed.
290:
291:
292: =cut
1.69 matthew 293:
1.15 foxr 294: sub SocketTimeout {
295: my $Socket = shift;
1.38 foxr 296: Log("WARNING", "A socket timeout was detected");
1.52 foxr 297: Debug(5, " SocketTimeout called: ");
1.48 foxr 298: $Socket->Dump(0);
1.42 foxr 299: if(exists($ActiveTransactions{$Socket})) {
1.43 albertel 300: FailTransaction($ActiveTransactions{$Socket});
1.42 foxr 301: }
1.22 foxr 302: KillSocket($Socket); # A transaction timeout also counts as
303: # a connection failure:
304: $ConnectionRetriesLeft--;
1.42 foxr 305: if($ConnectionRetriesLeft <= 0) {
1.52 foxr 306: Log("CRITICAL", "Host marked DEAD: ".GetServerHost());
1.56 foxr 307: $LondConnecting = 0;
1.42 foxr 308: }
309:
1.15 foxr 310: }
1.80 albertel 311:
1.64 foxr 312: #
313: # This function should be called by the child in all cases where it must
1.80 albertel 314: # exit. The child process must create a lock file for the AF_UNIX socket
315: # in order to prevent connection requests from lonnet in the time between
316: # process exit and the parent picking up the listen again.
317: #
1.64 foxr 318: # Parameters:
319: # exit_code - Exit status value, however see the next parameter.
320: # message - If this optional parameter is supplied, the exit
321: # is via a die with this message.
322: #
323: sub child_exit {
324: my ($exit_code, $message) = @_;
325:
326: # Regardless of how we exit, we may need to do the lock thing:
327:
1.80 albertel 328: #
329: # Create a lock file since there will be a time window
330: # between our exit and the parent's picking up the listen
331: # during which no listens will be done on the
332: # lonnet client socket.
333: #
334: my $lock_file = &GetLoncSocketPath().".lock";
335: open(LOCK,">$lock_file");
336: print LOCK "Contents not important";
337: close(LOCK);
1.81 albertel 338: unlink(&GetLoncSocketPath());
1.64 foxr 339:
1.80 albertel 340: if ($message) {
341: die($message);
1.64 foxr 342: } else {
343: exit($exit_code);
344: }
345: }
1.35 foxr 346: #----------------------------- Timer management ------------------------
1.15 foxr 347:
348: =pod
349:
1.1 foxr 350: =head2 Tick
1.3 albertel 351:
352: Invoked each timer tick.
353:
1.1 foxr 354: =cut
355:
1.5 foxr 356:
1.1 foxr 357: sub Tick {
1.52 foxr 358: my ($Event) = @_;
359: my $clock_watcher = $Event->w;
360:
1.1 foxr 361: my $client;
1.57 foxr 362: UpdateStatus();
363:
1.4 foxr 364: # Is it time to prune connection count:
365:
366:
367: if($IdleConnections->Count() &&
368: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
1.52 foxr 369: $IdleSeconds++;
1.4 foxr 370: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
1.23 foxr 371: my $Socket = $IdleConnections->pop();
1.6 foxr 372: KillSocket($Socket);
1.54 foxr 373: $IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
1.57 foxr 374: UpdateStatus();
1.80 albertel 375: if(($ConnectionCount == 0)) {
1.64 foxr 376: &child_exit(0);
377:
1.57 foxr 378: }
1.4 foxr 379: }
380: } else {
381: $IdleSeconds = 0; # Reset idle count if not idle.
382: }
1.15 foxr 383: #
384: # For each inflight transaction, tick down its timeout counter.
385: #
1.35 foxr 386:
1.34 albertel 387: foreach my $item (keys %ActiveConnections) {
388: my $State = $ActiveConnections{$item}->data->GetState();
1.35 foxr 389: if ($State ne 'Idle') {
1.34 albertel 390: Debug(5,"Ticking Socket $State $item");
391: $ActiveConnections{$item}->data->Tick();
392: }
1.15 foxr 393: }
1.5 foxr 394: # Do we have work in the queue, but no connections to service them?
395: # If so, try to make some new connections to get things going again.
396: #
1.57 foxr 397: # Note this code is dead now...
398: #
1.5 foxr 399: my $Requests = $WorkQueue->Count();
1.56 foxr 400: if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) {
1.10 foxr 401: if ($ConnectionRetriesLeft > 0) {
1.56 foxr 402: Debug(5,"Work but no connections, Make a new one");
403: my $success;
404: $success = &MakeLondConnection;
405: if($success == 0) { # All connections failed:
1.29 foxr 406: Debug(5,"Work in queue failed to make any connectiouns\n");
1.22 foxr 407: EmptyQueue(); # Fail pending transactions with con_lost.
1.42 foxr 408: CloseAllLondConnections(); # Should all be closed but....
1.10 foxr 409: }
410: } else {
1.56 foxr 411: $LondConnecting = 0;
1.22 foxr 412: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
1.29 foxr 413: Debug(5,"Work in queue, but gave up on connections..flushing\n");
1.10 foxr 414: EmptyQueue(); # Connections can't be established.
1.42 foxr 415: CloseAllLondConnections(); # Should all already be closed but...
1.5 foxr 416: }
417:
418: }
1.49 foxr 419: if ($ConnectionCount == 0) {
420: $KeyMode = "";
1.52 foxr 421: $clock_watcher->cancel();
1.49 foxr 422: }
1.66 albertel 423: &UpdateStatus();
1.1 foxr 424: }
425:
426: =pod
1.3 albertel 427:
1.1 foxr 428: =head2 SetupTimer
429:
1.3 albertel 430: Sets up a 1 per sec recurring timer event. The event handler is used to:
1.1 foxr 431:
1.3 albertel 432: =item
433:
434: Trigger timeouts on communications along active sockets.
435:
436: =item
437:
438: Trigger disconnections of idle sockets.
1.1 foxr 439:
440: =cut
441:
442: sub SetupTimer {
1.52 foxr 443: Debug(6, "SetupTimer");
1.92 foxr 444: Event->timer(interval => 1, cb => \&Tick,
445: hard => 1);
1.1 foxr 446: }
1.3 albertel 447:
1.1 foxr 448: =pod
1.3 albertel 449:
1.1 foxr 450: =head2 ServerToIdle
1.3 albertel 451:
452: This function is called when a connection to the server is
453: ready for more work.
454:
455: If there is work in the Work queue the top element is dequeued
1.1 foxr 456: and the connection will start to work on it. If the work queue is
457: empty, the connection is pushed on the idle connection stack where
458: it will either get another work unit, or alternatively, if it sits there
459: long enough, it will be shut down and released.
460:
1.3 albertel 461: =cut
1.1 foxr 462:
463: sub ServerToIdle {
464: my $Socket = shift; # Get the socket.
1.49 foxr 465: $KeyMode = $Socket->{AuthenticationMode};
1.7 foxr 466: delete($ActiveTransactions{$Socket}); # Server has no transaction
1.1 foxr 467:
1.29 foxr 468: &Debug(5, "Server to idle");
1.1 foxr 469:
470: # If there's work to do, start the transaction:
471:
1.23 foxr 472: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
1.29 foxr 473: if ($reqdata ne undef) {
474: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
1.7 foxr 475: &StartRequest($Socket, $reqdata);
1.8 foxr 476:
1.1 foxr 477: } else {
478:
479: # There's no work waiting, so push the server to idle list.
1.29 foxr 480: &Debug(5, "No new work requests, server connection going idle");
1.1 foxr 481: $IdleConnections->push($Socket);
482: }
483: }
1.3 albertel 484:
1.1 foxr 485: =pod
1.3 albertel 486:
1.1 foxr 487: =head2 ClientWritable
1.3 albertel 488:
489: Event callback for when a client socket is writable.
490:
491: This callback is established when a transaction reponse is
492: avaiable from lond. The response is forwarded to the unix socket
493: as it becomes writable in this sub.
494:
1.1 foxr 495: Parameters:
496:
1.3 albertel 497: =item Event
498:
499: The event that has been triggered. Event->w->data is
500: the data and Event->w->fd is the socket to write.
1.1 foxr 501:
502: =cut
1.3 albertel 503:
1.1 foxr 504: sub ClientWritable {
505: my $Event = shift;
506: my $Watcher = $Event->w;
1.84 albertel 507: if (!defined($Watcher)) {
508: &child_exit(-1,'No watcher for event in ClientWritable');
509: }
1.1 foxr 510: my $Data = $Watcher->data;
511: my $Socket = $Watcher->fd;
512:
513: # Try to send the data:
514:
515: &Debug(6, "ClientWritable writing".$Data);
516: &Debug(9, "Socket is: ".$Socket);
517:
1.6 foxr 518: if($Socket->connected) {
519: my $result = $Socket->send($Data, 0);
520:
521: # $result undefined: the write failed.
522: # otherwise $result is the number of bytes written.
523: # Remove that preceding string from the data.
524: # If the resulting data is empty, destroy the watcher
525: # and set up a read event handler to accept the next
526: # request.
527:
528: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
1.29 foxr 529: if($result ne undef) {
1.6 foxr 530: &Debug(9, "send result was defined");
531: if($result == length($Data)) { # Entire string sent.
532: &Debug(9, "ClientWritable data all written");
533: $Watcher->cancel();
534: #
535: # Set up to read next request from socket:
536:
537: my $descr = sprintf("Connection to lonc client %d",
538: $ActiveClients{$Socket});
539: Event->io(cb => \&ClientRequest,
540: poll => 'r',
541: desc => $descr,
542: data => "",
543: fd => $Socket);
544:
545: } else { # Partial string sent.
546: $Watcher->data(substr($Data, $result));
1.15 foxr 547: if($result == 0) { # client hung up on us!!
1.52 foxr 548: # Log("INFO", "lonc pipe client hung up on us!");
1.15 foxr 549: $Watcher->cancel;
550: $Socket->shutdown(2);
551: $Socket->close();
552: }
1.6 foxr 553: }
554:
555: } else { # Error of some sort...
556:
557: # Some errnos are possible:
558: my $errno = $!;
559: if($errno == POSIX::EWOULDBLOCK ||
560: $errno == POSIX::EAGAIN ||
561: $errno == POSIX::EINTR) {
562: # No action taken?
563: } else { # Unanticipated errno.
564: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
565: $Watcher->cancel; # Stop the watcher.
566: $Socket->shutdown(2); # Kill connection
567: $Socket->close(); # Close the socket.
568: }
1.1 foxr 569:
570: }
1.6 foxr 571: } else {
572: $Watcher->cancel(); # A delayed request...just cancel.
1.84 albertel 573: return;
1.1 foxr 574: }
575: }
576:
577: =pod
1.3 albertel 578:
1.1 foxr 579: =head2 CompleteTransaction
1.3 albertel 580:
581: Called when the reply data has been received for a lond
1.1 foxr 582: transaction. The reply data must now be sent to the
583: ultimate client on the other end of the Unix socket. This is
584: done by setting up a writable event for the socket with the
585: data the reply data.
1.3 albertel 586:
1.1 foxr 587: Parameters:
1.3 albertel 588:
589: =item Socket
590:
591: Socket on which the lond transaction occured. This is a
592: LondConnection. The data received is in the TransactionReply member.
593:
1.7 foxr 594: =item Transaction
1.3 albertel 595:
1.7 foxr 596: The transaction that is being completed.
1.1 foxr 597:
598: =cut
1.3 albertel 599:
1.1 foxr 600: sub CompleteTransaction {
1.29 foxr 601: &Debug(5,"Complete transaction");
1.47 foxr 602:
603: my ($Socket, $Transaction) = @_;
1.1 foxr 604:
1.7 foxr 605: if (!$Transaction->isDeferred()) { # Normal transaction
606: my $data = $Socket->GetReply(); # Data to send.
1.39 foxr 607: if($LogTransactions) {
608: Log("SUCCESS", "Reply from lond: '$data'");
609: }
1.7 foxr 610: StartClientReply($Transaction, $data);
611: } else { # Delete deferred transaction file.
1.9 foxr 612: Log("SUCCESS", "A delayed transaction was completed");
1.86 albertel 613: LogPerm("S:".$Transaction->getClient().":".$Transaction->getRequest());
614: unlink($Transaction->getFile());
1.7 foxr 615: }
1.6 foxr 616: }
1.42 foxr 617:
1.6 foxr 618: =pod
1.42 foxr 619:
1.6 foxr 620: =head1 StartClientReply
621:
622: Initiates a reply to a client where the reply data is a parameter.
623:
1.7 foxr 624: =head2 parameters:
625:
626: =item Transaction
627:
628: The transaction for which we are responding to the client.
629:
630: =item data
631:
632: The data to send to apached client.
633:
1.6 foxr 634: =cut
1.42 foxr 635:
1.6 foxr 636: sub StartClientReply {
1.1 foxr 637:
1.47 foxr 638: my ($Transaction, $data) = @_;
1.12 foxr 639:
1.7 foxr 640: my $Client = $Transaction->getClient();
641:
1.1 foxr 642: &Debug(8," Reply was: ".$data);
643: my $Serial = $ActiveClients{$Client};
644: my $desc = sprintf("Connection to lonc client %d",
645: $Serial);
646: Event->io(fd => $Client,
647: poll => "w",
648: desc => $desc,
649: cb => \&ClientWritable,
650: data => $data);
651: }
1.42 foxr 652:
1.4 foxr 653: =pod
1.42 foxr 654:
1.4 foxr 655: =head2 FailTransaction
656:
657: Finishes a transaction with failure because the associated lond socket
1.7 foxr 658: disconnected. There are two possibilities:
659: - The transaction is deferred: in which case we just quietly
660: delete the transaction since there is no client connection.
661: - The transaction is 'live' in which case we initiate the sending
662: of "con_lost" to the client.
663:
1.42 foxr 664: Deleting the transaction means killing it from the %ActiveTransactions hash.
1.4 foxr 665:
666: Parameters:
667:
668: =item client
669:
1.7 foxr 670: The LondTransaction we are failing.
671:
1.42 foxr 672:
1.4 foxr 673: =cut
674:
675: sub FailTransaction {
1.7 foxr 676: my $transaction = shift;
1.52 foxr 677:
678: # If the socket is dead, that's already logged.
679:
680: if ($ConnectionRetriesLeft > 0) {
681: Log("WARNING", "Failing transaction "
1.71 albertel 682: .$transaction->getLoggableRequest());
1.52 foxr 683: }
1.71 albertel 684: Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
1.10 foxr 685: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
1.11 foxr 686: my $client = $transaction->getClient();
1.30 foxr 687: Debug(1," Replying con_lost to ".$transaction->getRequest());
1.11 foxr 688: StartClientReply($transaction, "con_lost\n");
1.7 foxr 689: }
1.4 foxr 690:
691: }
692:
693: =pod
1.69 matthew 694:
1.6 foxr 695: =head1 EmptyQueue
1.7 foxr 696:
1.6 foxr 697: Fails all items in the work queue with con_lost.
1.7 foxr 698: Note that each item in the work queue is a transaction.
699:
1.6 foxr 700: =cut
1.69 matthew 701:
1.6 foxr 702: sub EmptyQueue {
1.22 foxr 703: $ConnectionRetriesLeft--; # Counts as connection failure too.
1.6 foxr 704: while($WorkQueue->Count()) {
1.10 foxr 705: my $request = $WorkQueue->dequeue(); # This is a transaction
1.7 foxr 706: FailTransaction($request);
1.6 foxr 707: }
708: }
709:
710: =pod
1.4 foxr 711:
1.9 foxr 712: =head2 CloseAllLondConnections
713:
714: Close all connections open on lond prior to exit e.g.
715:
716: =cut
1.69 matthew 717:
1.9 foxr 718: sub CloseAllLondConnections {
1.23 foxr 719: foreach my $Socket (keys %ActiveConnections) {
1.42 foxr 720: if(exists($ActiveTransactions{$Socket})) {
721: FailTransaction($ActiveTransactions{$Socket});
722: }
723: KillSocket($Socket);
1.9 foxr 724: }
725: }
726:
727: =pod
728:
1.4 foxr 729: =head2 KillSocket
730:
731: Destroys a socket. This function can be called either when a socket
732: has died of 'natural' causes or because a socket needs to be pruned due to
733: idleness. If the socket has died naturally, if there are no longer any
734: live connections a new connection is created (in case there are transactions
735: in the queue). If the socket has been pruned, it is never re-created.
736:
737: Parameters:
1.1 foxr 738:
1.4 foxr 739: =item Socket
740:
741: The socket to kill off.
742:
743: =item Restart
744:
745: nonzero if we are allowed to create a new connection.
746:
1.69 matthew 747: =cut
1.4 foxr 748:
749: sub KillSocket {
750: my $Socket = shift;
751:
1.17 foxr 752: Log("WARNING", "Shutting down a socket");
1.9 foxr 753: $Socket->Shutdown();
754:
1.7 foxr 755: # If the socket came from the active connection set,
756: # delete its transaction... note that FailTransaction should
757: # already have been called!!!
758: # otherwise it came from the idle set.
759: #
1.4 foxr 760:
761: if(exists($ActiveTransactions{$Socket})) {
762: delete ($ActiveTransactions{$Socket});
763: }
764: if(exists($ActiveConnections{$Socket})) {
1.90 foxr 765: $ActiveConnections{$Socket}->cancel;
1.4 foxr 766: delete($ActiveConnections{$Socket});
1.37 albertel 767: $ConnectionCount--;
768: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
1.4 foxr 769: }
1.6 foxr 770: # If the connection count has gone to zero and there is work in the
771: # work queue, the work all gets failed with con_lost.
772: #
773: if($ConnectionCount == 0) {
1.22 foxr 774: EmptyQueue();
1.42 foxr 775: CloseAllLondConnections; # Should all already be closed but...
1.4 foxr 776: }
1.90 foxr 777: UpdateStatus();
1.4 foxr 778: }
1.1 foxr 779:
780: =pod
1.3 albertel 781:
1.1 foxr 782: =head2 LondReadable
1.3 albertel 783:
1.1 foxr 784: This function is called whenever a lond connection
785: is readable. The action is state dependent:
786:
1.3 albertel 787: =head3 State=Initialized
788:
789: We''re waiting for the challenge, this is a no-op until the
1.1 foxr 790: state changes.
1.3 albertel 791:
1.1 foxr 792: =head3 State=Challenged
1.3 albertel 793:
794: The challenge has arrived we need to transition to Writable.
1.1 foxr 795: The connection must echo the challenge back.
1.3 albertel 796:
1.1 foxr 797: =head3 State=ChallengeReplied
1.3 albertel 798:
799: The challenge has been replied to. The we are receiveing the
1.1 foxr 800: 'ok' from the partner.
1.3 albertel 801:
1.40 foxr 802: =head3 State=ReadingVersionString
803:
804: We have requested the lond version and are reading the
805: version back. Upon completion, we'll store the version away
806: for future use(?).
807:
808: =head3 State=HostSet
809:
810: We have selected the domain name of our peer (multhomed hosts)
811: and are getting the reply (presumably ok) back.
812:
1.1 foxr 813: =head3 State=RequestingKey
1.3 albertel 814:
815: The ok has been received and we need to send the request for
1.1 foxr 816: an encryption key. Transition to writable for that.
1.3 albertel 817:
1.1 foxr 818: =head3 State=ReceivingKey
1.3 albertel 819:
820: The the key has been requested, now we are reading the new key.
821:
1.1 foxr 822: =head3 State=Idle
1.3 albertel 823:
824: The encryption key has been negotiated or we have finished
1.1 foxr 825: reading data from the a transaction. If the callback data has
826: a client as well as the socket iformation, then we are
827: doing a transaction and the data received is relayed to the client
828: before the socket is put on the idle list.
1.3 albertel 829:
1.1 foxr 830: =head3 State=SendingRequest
1.3 albertel 831:
832: I do not think this state can be received here, but if it is,
1.1 foxr 833: the appropriate thing to do is to transition to writable, and send
834: the request.
1.3 albertel 835:
1.1 foxr 836: =head3 State=ReceivingReply
1.3 albertel 837:
838: We finished sending the request to the server and now transition
1.1 foxr 839: to readable to receive the reply.
840:
841: The parameter to this function are:
1.3 albertel 842:
1.1 foxr 843: The event. Implicit in this is the watcher and its data. The data
844: contains at least the lond connection object and, if a
845: transaction is in progress, the socket attached to the local client.
846:
1.3 albertel 847: =cut
1.1 foxr 848:
849: sub LondReadable {
1.8 foxr 850:
1.41 albertel 851: my $Event = shift;
852: my $Watcher = $Event->w;
853: my $Socket = $Watcher->data;
854: my $client = undef;
1.40 foxr 855:
1.41 albertel 856: &Debug(6,"LondReadable called state = ".$Socket->GetState());
1.40 foxr 857:
858:
1.41 albertel 859: my $State = $Socket->GetState(); # All action depends on the state.
1.40 foxr 860:
1.41 albertel 861: SocketDump(6, $Socket);
862: my $status = $Socket->Readable();
1.40 foxr 863:
1.41 albertel 864: &Debug(2, "Socket->Readable returned: $status");
1.40 foxr 865:
1.41 albertel 866: if($status != 0) {
867: # bad return from socket read. Currently this means that
868: # The socket has become disconnected. We fail the transaction.
1.40 foxr 869:
1.41 albertel 870: Log("WARNING",
871: "Lond connection lost.");
872: if(exists($ActiveTransactions{$Socket})) {
873: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 874: } else {
875: # Socket is connecting and failed... need to mark
876: # no longer connecting.
877:
878: $LondConnecting = 0;
1.41 albertel 879: }
880: $Watcher->cancel();
881: KillSocket($Socket);
882: $ConnectionRetriesLeft--; # Counts as connection failure
883: return;
884: }
885: SocketDump(6,$Socket);
1.17 foxr 886:
1.41 albertel 887: $State = $Socket->GetState(); # Update in case of transition.
888: &Debug(6, "After read, state is ".$State);
1.1 foxr 889:
1.41 albertel 890: if($State eq "Initialized") {
1.1 foxr 891:
892:
1.41 albertel 893: } elsif ($State eq "ChallengeReceived") {
1.1 foxr 894: # The challenge must be echoed back; The state machine
895: # in the connection takes care of setting that up. Just
896: # need to transition to writable:
1.41 albertel 897:
898: $Watcher->cb(\&LondWritable);
899: $Watcher->poll("w");
1.1 foxr 900:
1.41 albertel 901: } elsif ($State eq "ChallengeReplied") {
1.1 foxr 902:
1.41 albertel 903: } elsif ($State eq "RequestingVersion") {
904: # Need to ask for the version... that is writiability:
1.1 foxr 905:
1.41 albertel 906: $Watcher->cb(\&LondWritable);
907: $Watcher->poll("w");
908:
909: } elsif ($State eq "ReadingVersionString") {
910: # Read the rest of the version string...
911: } elsif ($State eq "SetHost") {
912: # Need to request the actual domain get set...
913:
914: $Watcher->cb(\&LondWritable);
915: $Watcher->poll("w");
916: } elsif ($State eq "HostSet") {
917: # Reading the 'ok' from the peer.
918:
919: } elsif ($State eq "RequestingKey") {
1.1 foxr 920: # The ok was received. Now we need to request the key
921: # That requires us to be writable:
922:
1.41 albertel 923: $Watcher->cb(\&LondWritable);
924: $Watcher->poll("w");
1.1 foxr 925:
1.41 albertel 926: } elsif ($State eq "ReceivingKey") {
1.1 foxr 927:
1.41 albertel 928: } elsif ($State eq "Idle") {
1.40 foxr 929:
1.41 albertel 930: # This is as good a spot as any to get the peer version
931: # string:
1.40 foxr 932:
1.41 albertel 933: if($LondVersion eq "unknown") {
934: $LondVersion = $Socket->PeerVersion();
935: Log("INFO", "Connected to lond version: $LondVersion");
936: }
1.1 foxr 937: # If necessary, complete a transaction and then go into the
938: # idle queue.
1.22 foxr 939: # Note that a trasition to idle indicates a live lond
940: # on the other end so reset the connection retries.
941: #
1.41 albertel 942: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
943: $Watcher->cancel();
944: if(exists($ActiveTransactions{$Socket})) {
945: Debug(5,"Completing transaction!!");
946: CompleteTransaction($Socket,
947: $ActiveTransactions{$Socket});
948: } else {
1.95 ! foxr 949: my $count = $Socket->GetClientData();
! 950: Log("SUCCESS", "Connection ".$count." to "
1.41 albertel 951: .$RemoteHost." now ready for action");
952: }
953: ServerToIdle($Socket); # Next work unit or idle.
1.54 foxr 954:
955: #
956: $LondConnecting = 0; # Best spot I can think of for this.
957: #
1.6 foxr 958:
1.41 albertel 959: } elsif ($State eq "SendingRequest") {
1.1 foxr 960: # We need to be writable for this and probably don't belong
961: # here inthe first place.
962:
1.73 albertel 963: Debug(6, "SendingRequest state encountered in readable");
1.41 albertel 964: $Watcher->poll("w");
965: $Watcher->cb(\&LondWritable);
1.1 foxr 966:
1.41 albertel 967: } elsif ($State eq "ReceivingReply") {
1.1 foxr 968:
969:
1.41 albertel 970: } else {
971: # Invalid state.
972: Debug(4, "Invalid state in LondReadable");
973: }
1.1 foxr 974: }
1.3 albertel 975:
1.1 foxr 976: =pod
1.3 albertel 977:
1.1 foxr 978: =head2 LondWritable
1.3 albertel 979:
1.1 foxr 980: This function is called whenever a lond connection
981: becomes writable while there is a writeable monitoring
982: event. The action taken is very state dependent:
1.3 albertel 983:
1.1 foxr 984: =head3 State = Connected
1.3 albertel 985:
986: The connection is in the process of sending the 'init' hailing to the
987: lond on the remote end. The connection object''s Writable member is
988: called. On error, ConnectionError is called to destroy the connection
989: and remove it from the ActiveConnections hash
990:
1.1 foxr 991: =head3 Initialized
1.3 albertel 992:
993: 'init' has been sent, writability monitoring is removed and
994: readability monitoring is started with LondReadable as the callback.
995:
1.1 foxr 996: =head3 ChallengeReceived
1.3 albertel 997:
998: The connection has received the who are you challenge from the remote
999: system, and is in the process of sending the challenge
1000: response. Writable is called.
1001:
1.1 foxr 1002: =head3 ChallengeReplied
1.3 albertel 1003:
1004: The connection has replied to the initial challenge The we switch to
1005: monitoring readability looking for the server to reply with 'ok'.
1006:
1.1 foxr 1007: =head3 RequestingKey
1.3 albertel 1008:
1009: The connection is in the process of requesting its encryption key.
1010: Writable is called.
1011:
1.1 foxr 1012: =head3 ReceivingKey
1.3 albertel 1013:
1014: The connection has sent the request for a key. Switch to readability
1015: monitoring to accept the key
1016:
1.1 foxr 1017: =head3 SendingRequest
1.3 albertel 1018:
1019: The connection is in the process of sending a request to the server.
1020: This request is part of a client transaction. All the states until
1021: now represent the client setup protocol. Writable is called.
1022:
1.1 foxr 1023: =head3 ReceivingReply
1024:
1.3 albertel 1025: The connection has sent a request. Now it must receive a reply.
1026: Readability monitoring is requested.
1027:
1028: This function is an event handler and therefore receives as
1.1 foxr 1029: a parameter the event that has fired. The data for the watcher
1030: of this event is a reference to a list of one or two elements,
1031: depending on state. The first (and possibly only) element is the
1032: socket. The second (present only if a request is in progress)
1033: is the socket on which to return a reply to the caller.
1034:
1035: =cut
1.3 albertel 1036:
1.1 foxr 1037: sub LondWritable {
1038: my $Event = shift;
1039: my $Watcher = $Event->w;
1.8 foxr 1040: my $Socket = $Watcher->data;
1041: my $State = $Socket->GetState();
1.1 foxr 1042:
1.8 foxr 1043: Debug(6,"LondWritable State = ".$State."\n");
1.1 foxr 1044:
1.8 foxr 1045:
1.1 foxr 1046: # Figure out what to do depending on the state of the socket:
1047:
1048:
1049:
1050:
1051: SocketDump(6,$Socket);
1052:
1.42 foxr 1053: # If the socket is writable, we must always write.
1054: # Only by writing will we undergo state transitions.
1055: # Old logic wrote in state specific code below, however
1056: # That forces us at least through another invocation of
1057: # this function after writability is possible again.
1058: # This logic also factors out common code for handling
1059: # write failures... in all cases, write failures
1060: # Kill the socket.
1061: # This logic makes the branches of the >big< if below
1062: # so that the writing states are actually NO-OPs.
1063:
1064: if ($Socket->Writable() != 0) {
1.43 albertel 1065: # The write resulted in an error.
1066: # We'll treat this as if the socket got disconnected:
1067: Log("WARNING", "Connection to ".$RemoteHost.
1068: " has been disconnected");
1069: if(exists($ActiveTransactions{$Socket})) {
1070: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 1071: } else {
1072: # In the process of conneting, so need to turn that off.
1073:
1074: $LondConnecting = 0;
1.43 albertel 1075: }
1076: $Watcher->cancel();
1077: KillSocket($Socket);
1078: return;
1.42 foxr 1079: }
1080:
1081:
1082:
1.41 albertel 1083: if ($State eq "Connected") {
1.1 foxr 1084:
1.41 albertel 1085: # "init" is being sent...
1.42 foxr 1086:
1.41 albertel 1087: } elsif ($State eq "Initialized") {
1.4 foxr 1088:
1.41 albertel 1089: # Now that init was sent, we switch
1090: # to watching for readability:
1.1 foxr 1091:
1.41 albertel 1092: $Watcher->cb(\&LondReadable);
1093: $Watcher->poll("r");
1094:
1095: } elsif ($State eq "ChallengeReceived") {
1096: # We received the challenge, now we
1097: # are echoing it back. This is a no-op,
1098: # we're waiting for the state to change
1.1 foxr 1099:
1.41 albertel 1100: } elsif ($State eq "ChallengeReplied") {
1101: # The echo was sent back, so we switch
1102: # to watching readability.
1103:
1104: $Watcher->cb(\&LondReadable);
1105: $Watcher->poll("r");
1106: } elsif ($State eq "RequestingVersion") {
1107: # Sending the peer a version request...
1.42 foxr 1108:
1.41 albertel 1109: } elsif ($State eq "ReadingVersionString") {
1110: # Transition to read since we have sent the
1111: # version command and now just need to read the
1112: # version string from the peer:
1.40 foxr 1113:
1.41 albertel 1114: $Watcher->cb(\&LondReadable);
1115: $Watcher->poll("r");
1.40 foxr 1116:
1.41 albertel 1117: } elsif ($State eq "SetHost") {
1118: # Setting the remote domain...
1.42 foxr 1119:
1.41 albertel 1120: } elsif ($State eq "HostSet") {
1121: # Back to readable to get the ok.
1.40 foxr 1122:
1.41 albertel 1123: $Watcher->cb(\&LondReadable);
1124: $Watcher->poll("r");
1.40 foxr 1125:
1126:
1.41 albertel 1127: } elsif ($State eq "RequestingKey") {
1128: # At this time we're requesting the key.
1129: # again, this is essentially a no-op.
1130:
1131: } elsif ($State eq "ReceivingKey") {
1132: # Now we need to wait for the key
1133: # to come back from the peer:
1134:
1135: $Watcher->cb(\&LondReadable);
1136: $Watcher->poll("r");
1137:
1138: } elsif ($State eq "SendingRequest") {
1.40 foxr 1139:
1.41 albertel 1140: # At this time we are sending a request to the
1.1 foxr 1141: # peer... write the next chunk:
1142:
1.41 albertel 1143:
1144: } elsif ($State eq "ReceivingReply") {
1145: # The send has completed. Wait for the
1146: # data to come in for a reply.
1147: Debug(8,"Writable sent request/receiving reply");
1148: $Watcher->cb(\&LondReadable);
1149: $Watcher->poll("r");
1.1 foxr 1150:
1.41 albertel 1151: } else {
1152: # Control only passes here on an error:
1153: # the socket state does not match any
1154: # of the known states... so an error
1155: # must be logged.
1.1 foxr 1156:
1.41 albertel 1157: &Debug(4, "Invalid socket state ".$State."\n");
1158: }
1.1 foxr 1159:
1160: }
1.81 albertel 1161:
1.6 foxr 1162: =pod
1163:
1164: =cut
1.69 matthew 1165:
1.81 albertel 1166:
1.6 foxr 1167: sub QueueDelayed {
1.8 foxr 1168: Debug(3,"QueueDelayed called");
1169:
1.6 foxr 1170: my $path = "$perlvar{'lonSockDir'}/delayed";
1.8 foxr 1171:
1172: Debug(4, "Delayed path: ".$path);
1.6 foxr 1173: opendir(DIRHANDLE, $path);
1.75 albertel 1174:
1.82 albertel 1175: my $host_id_re = '(?:'.join('|',map {quotemeta($_)} (@all_host_ids)).')';
1.75 albertel 1176: my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
1.6 foxr 1177: closedir(DIRHANDLE);
1.75 albertel 1178: foreach my $dfname (sort(@alldelayed)) {
1179: my $reqfile = "$path/$dfname";
1180: my ($host_id) = ($dfname =~ /\.([^.]*)$/);
1181: Debug(4, "queueing ".$reqfile." for $host_id");
1.6 foxr 1182: my $Handle = IO::File->new($reqfile);
1183: my $cmd = <$Handle>;
1.8 foxr 1184: chomp $cmd; # There may or may not be a newline...
1.12 foxr 1185: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1.75 albertel 1186: my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
1.7 foxr 1187: $Transaction->SetDeferred($reqfile);
1188: QueueTransaction($Transaction);
1.6 foxr 1189: }
1190:
1191: }
1.1 foxr 1192:
1193: =pod
1.3 albertel 1194:
1.1 foxr 1195: =head2 MakeLondConnection
1.3 albertel 1196:
1197: Create a new lond connection object, and start it towards its initial
1198: idleness. Once idle, it becomes elligible to receive transactions
1199: from the work queue. If the work queue is not empty when the
1200: connection is completed and becomes idle, it will dequeue an entry and
1201: start off on it.
1202:
1.1 foxr 1203: =cut
1.3 albertel 1204:
1.1 foxr 1205: sub MakeLondConnection {
1206: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1207: .GetServerPort());
1208:
1209: my $Connection = LondConnection->new(&GetServerHost(),
1.81 albertel 1210: &GetServerPort(),
1211: &GetHostId());
1.1 foxr 1212:
1.90 foxr 1213: if($Connection eq undef) {
1.9 foxr 1214: Log("CRITICAL","Failed to make a connection with lond.");
1.10 foxr 1215: $ConnectionRetriesLeft--;
1216: return 0; # Failure.
1.5 foxr 1217: } else {
1.22 foxr 1218:
1.82 albertel 1219: $LondConnecting = 1; # Connection in progress.
1.5 foxr 1220: # The connection needs to have writability
1221: # monitored in order to send the init sequence
1222: # that starts the whole authentication/key
1223: # exchange underway.
1224: #
1225: my $Socket = $Connection->GetSocket();
1.30 foxr 1226: if($Socket eq undef) {
1.64 foxr 1227: &child_exit(-1, "did not get a socket from the connection");
1.5 foxr 1228: } else {
1229: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1230: }
1.1 foxr 1231:
1.21 foxr 1232: $Connection->SetTimeoutCallback(\&SocketTimeout);
1233:
1.23 foxr 1234: my $event = Event->io(fd => $Socket,
1.5 foxr 1235: poll => 'w',
1236: cb => \&LondWritable,
1.8 foxr 1237: data => $Connection,
1.5 foxr 1238: desc => 'Connection to lond server');
1239: $ActiveConnections{$Connection} = $event;
1.52 foxr 1240: if ($ConnectionCount == 0) {
1241: &SetupTimer; # Need to handle timeouts with connections...
1242: }
1.5 foxr 1243: $ConnectionCount++;
1.95 ! foxr 1244: $Connection->SetClientData($ConnectionCount);
1.8 foxr 1245: Debug(4, "Connection count = ".$ConnectionCount);
1.6 foxr 1246: if($ConnectionCount == 1) { # First Connection:
1247: QueueDelayed;
1248: }
1.9 foxr 1249: Log("SUCESS", "Created connection ".$ConnectionCount
1250: ." to host ".GetServerHost());
1.10 foxr 1251: return 1; # Return success.
1.1 foxr 1252: }
1253:
1254: }
1.3 albertel 1255:
1.1 foxr 1256: =pod
1.3 albertel 1257:
1.1 foxr 1258: =head2 StartRequest
1.3 albertel 1259:
1260: Starts a lond request going on a specified lond connection.
1261: parameters are:
1262:
1263: =item $Lond
1264:
1265: Connection to the lond that will send the transaction and receive the
1266: reply.
1267:
1268: =item $Client
1269:
1270: Connection to the client that is making this request We got the
1271: request from this socket, and when the request has been relayed to
1272: lond and we get a reply back from lond it will get sent to this
1273: socket.
1274:
1275: =item $Request
1276:
1277: The text of the request to send.
1278:
1.1 foxr 1279: =cut
1280:
1281: sub StartRequest {
1.47 foxr 1282:
1283: my ($Lond, $Request) = @_;
1.1 foxr 1284:
1.7 foxr 1285: Debug(6, "StartRequest: ".$Request->getRequest());
1.1 foxr 1286:
1287: my $Socket = $Lond->GetSocket();
1288:
1.7 foxr 1289: $Request->Activate($Lond);
1290: $ActiveTransactions{$Lond} = $Request;
1.1 foxr 1291:
1.7 foxr 1292: $Lond->InitiateTransaction($Request->getRequest());
1.23 foxr 1293: my $event = Event->io(fd => $Socket,
1.1 foxr 1294: poll => "w",
1295: cb => \&LondWritable,
1296: data => $Lond,
1297: desc => "lond transaction connection");
1298: $ActiveConnections{$Lond} = $event;
1299: Debug(8," Start Request made watcher data with ".$event->data."\n");
1300: }
1301:
1302: =pod
1.3 albertel 1303:
1.1 foxr 1304: =head2 QueueTransaction
1.3 albertel 1305:
1306: If there is an idle lond connection, it is put to work doing this
1307: transaction. Otherwise, the transaction is placed in the work queue.
1308: If placed in the work queue and the maximum number of connections has
1309: not yet been created, a new connection will be started. Our goal is
1310: to eventually have a sufficient number of connections that the work
1311: queue will typically be empty. parameters are:
1312:
1313: =item Socket
1314:
1315: open on the lonc client.
1316:
1317: =item Request
1318:
1319: data to send to the lond.
1.1 foxr 1320:
1321: =cut
1.3 albertel 1322:
1.1 foxr 1323: sub QueueTransaction {
1324:
1.7 foxr 1325: my $requestData = shift; # This is a LondTransaction.
1326: my $cmd = $requestData->getRequest();
1327:
1328: Debug(6,"QueueTransaction: ".$cmd);
1.1 foxr 1329:
1330: my $LondSocket = $IdleConnections->pop();
1331: if(!defined $LondSocket) { # Need to queue request.
1.29 foxr 1332: Debug(5,"Must queue...");
1.1 foxr 1333: $WorkQueue->enqueue($requestData);
1.56 foxr 1334: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1335: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1336:
1.22 foxr 1337: if($ConnectionRetriesLeft > 0) {
1.29 foxr 1338: Debug(5,"Starting additional lond connection");
1.56 foxr 1339: if(&MakeLondConnection() == 0) {
1.22 foxr 1340: EmptyQueue(); # Fail transactions, can't make connection.
1.42 foxr 1341: CloseAllLondConnections; # Should all be closed but...
1.22 foxr 1342: }
1343: } else {
1344: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1.56 foxr 1345: $LondConnecting = 0;
1.22 foxr 1346: EmptyQueue(); # It's worse than that ... he's dead Jim.
1.42 foxr 1347: CloseAllLondConnections; # Should all be closed but..
1.17 foxr 1348: }
1.1 foxr 1349: }
1350: } else { # Can start the request:
1351: Debug(8,"Can start...");
1.7 foxr 1352: StartRequest($LondSocket, $requestData);
1.1 foxr 1353: }
1354: }
1355:
1.95 ! foxr 1356: #-------------------------- Lonc UNIX socket handling -------------------
1.1 foxr 1357: =pod
1.3 albertel 1358:
1.1 foxr 1359: =head2 ClientRequest
1.3 albertel 1360: Callback that is called when data can be read from the UNIX domain
1361: socket connecting us with an apache server process.
1.1 foxr 1362:
1363: =cut
1364:
1365: sub ClientRequest {
1366: Debug(6, "ClientRequest");
1367: my $event = shift;
1368: my $watcher = $event->w;
1369: my $socket = $watcher->fd;
1370: my $data = $watcher->data;
1371: my $thisread;
1372:
1373: Debug(9, " Watcher named: ".$watcher->desc);
1374:
1375: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1376: Debug(8, "rcv: data length = ".length($thisread)
1377: ." read =".$thisread);
1.29 foxr 1378: unless (defined $rv && length($thisread)) {
1.1 foxr 1379: # Likely eof on socket.
1380: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1381: close($socket);
1382: $watcher->cancel();
1383: delete($ActiveClients{$socket});
1.10 foxr 1384: return;
1.1 foxr 1385: }
1386: Debug(8,"Data: ".$data." this read: ".$thisread);
1387: $data = $data.$thisread; # Append new data.
1388: $watcher->data($data);
1.44 albertel 1389: if($data =~ /\n$/) { # Request entirely read.
1.87 albertel 1390: if ($data eq "close_connection_exit\n") {
1.9 foxr 1391: Log("CRITICAL",
1392: "Request Close Connection ... exiting");
1393: CloseAllLondConnections();
1394: exit;
1.87 albertel 1395: } elsif ($data eq "reset_retries\n") {
1396: Log("INFO", "Resetting Connection Retries.");
1397: $ConnectionRetriesLeft = $ConnectionRetries;
1398: &UpdateStatus();
1399: my $Transaction = LondTransaction->new($data);
1400: $Transaction->SetClient($socket);
1401: StartClientReply($Transaction, "ok\n");
1402: $watcher->cancel();
1403: return;
1.9 foxr 1404: }
1.1 foxr 1405: Debug(8, "Complete transaction received: ".$data);
1.87 albertel 1406: if ($LogTransactions) {
1.39 foxr 1407: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1408: }
1.8 foxr 1409: my $Transaction = LondTransaction->new($data);
1.7 foxr 1410: $Transaction->SetClient($socket);
1411: QueueTransaction($Transaction);
1.1 foxr 1412: $watcher->cancel(); # Done looking for input data.
1413: }
1414:
1415: }
1416:
1.62 foxr 1417: #
1418: # Accept a connection request for a client (lonc child) and
1419: # start up an event watcher to keep an eye on input from that
1420: # Event. This can be called both from NewClient and from
1.80 albertel 1421: # ChildProcess.
1.62 foxr 1422: # Parameters:
1423: # $socket - The listener socket.
1424: # Returns:
1425: # NONE
1426: # Side Effects:
1427: # An event is made to watch the accepted connection.
1428: # Active clients hash is updated to reflect the new connection.
1429: # The client connection count is incremented.
1430: #
1431: sub accept_client {
1432: my ($socket) = @_;
1433:
1434: Debug(8, "Entering accept for lonc UNIX socket\n");
1435: my $connection = $socket->accept(); # Accept the client connection.
1436: Debug(8,"Connection request accepted from "
1437: .GetPeername($connection, AF_UNIX));
1438:
1439:
1440: my $description = sprintf("Connection to lonc client %d",
1441: $ClientConnection);
1442: Debug(9, "Creating event named: ".$description);
1443: Event->io(cb => \&ClientRequest,
1444: poll => 'r',
1445: desc => $description,
1446: data => "",
1447: fd => $connection);
1448: $ActiveClients{$connection} = $ClientConnection;
1449: $ClientConnection++;
1450: }
1.1 foxr 1451:
1452: =pod
1.3 albertel 1453:
1.1 foxr 1454: =head2 NewClient
1.3 albertel 1455:
1456: Callback that is called when a connection is received on the unix
1457: socket for a new client of lonc. The callback is parameterized by the
1458: event.. which is a-priori assumed to be an io event, and therefore has
1459: an fd member that is the Listener socket. We Accept the connection
1460: and register a new event on the readability of that socket:
1461:
1.1 foxr 1462: =cut
1.3 albertel 1463:
1.1 foxr 1464: sub NewClient {
1465: Debug(6, "NewClient");
1466: my $event = shift; # Get the event parameters.
1467: my $watcher = $event->w;
1468: my $socket = $watcher->fd; # Get the event' socket.
1469:
1.62 foxr 1470: &accept_client($socket);
1.1 foxr 1471: }
1.3 albertel 1472:
1473: =pod
1474:
1475: =head2 GetLoncSocketPath
1476:
1477: Returns the name of the UNIX socket on which to listen for client
1478: connections.
1.1 foxr 1479:
1.58 foxr 1480: =head2 Parameters:
1481:
1482: host (optional) - Name of the host socket to return.. defaults to
1483: the return from GetServerHost().
1484:
1.1 foxr 1485: =cut
1.3 albertel 1486:
1.1 foxr 1487: sub GetLoncSocketPath {
1.58 foxr 1488:
1489: my $host = GetServerHost(); # Default host.
1490: if (@_) {
1491: ($host) = @_; # Override if supplied.
1492: }
1493: return $UnixSocketDir."/".$host;
1.1 foxr 1494: }
1495:
1.3 albertel 1496: =pod
1497:
1498: =head2 GetServerHost
1499:
1500: Returns the host whose lond we talk with.
1501:
1.1 foxr 1502: =cut
1.3 albertel 1503:
1.7 foxr 1504: sub GetServerHost {
1.1 foxr 1505: return $RemoteHost; # Setup by the fork.
1506: }
1.3 albertel 1507:
1508: =pod
1509:
1.81 albertel 1510: =head2 GetServerId
1511:
1512: Returns the hostid whose lond we talk with.
1513:
1514: =cut
1515:
1516: sub GetHostId {
1517: return $RemoteHostId; # Setup by the fork.
1518: }
1519:
1520: =pod
1521:
1.3 albertel 1522: =head2 GetServerPort
1523:
1524: Returns the lond port number.
1525:
1.1 foxr 1526: =cut
1.3 albertel 1527:
1.7 foxr 1528: sub GetServerPort {
1.1 foxr 1529: return $perlvar{londPort};
1530: }
1.3 albertel 1531:
1532: =pod
1533:
1534: =head2 SetupLoncListener
1535:
1536: Setup a lonc listener event. The event is called when the socket
1537: becomes readable.. that corresponds to the receipt of a new
1538: connection. The event handler established will accept the connection
1539: (creating a communcations channel), that int turn will establish
1540: another event handler to subess requests.
1.1 foxr 1541:
1.58 foxr 1542: =head2 Parameters:
1543:
1544: host (optional) Name of the host to set up a unix socket to.
1545:
1.1 foxr 1546: =cut
1.3 albertel 1547:
1.1 foxr 1548: sub SetupLoncListener {
1.78 albertel 1549: my ($host,$SocketName) = @_;
1550: if (!$host) { $host = &GetServerHost(); }
1551: if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }
1.1 foxr 1552:
1.78 albertel 1553:
1554: unlink($SocketName);
1.58 foxr 1555:
1.1 foxr 1556: my $socket;
1.7 foxr 1557: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1.55 albertel 1558: Listen => 250,
1.1 foxr 1559: Type => SOCK_STREAM)) {
1.64 foxr 1560: if($I_am_child) {
1561: &child_exit(-1, "Failed to create a lonc listener socket");
1562: } else {
1563: die "Failed to create a lonc listner socket";
1564: }
1.1 foxr 1565: }
1.59 foxr 1566: return $socket;
1.1 foxr 1567: }
1568:
1.39 foxr 1569: #
1570: # Toggle transaction logging.
1571: # Implicit inputs:
1572: # LogTransactions
1573: # Implicit Outputs:
1574: # LogTransactions
1575: sub ToggleTransactionLogging {
1576: print STDERR "Toggle transaction logging...\n";
1577: if(!$LogTransactions) {
1578: $LogTransactions = 1;
1579: } else {
1580: $LogTransactions = 0;
1581: }
1582:
1583:
1584: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1585: }
1586:
1.14 foxr 1587: =pod
1588:
1589: =head2 ChildStatus
1590:
1591: Child USR1 signal handler to report the most recent status
1592: into the status file.
1593:
1.22 foxr 1594: We also use this to reset the retries count in order to allow the
1595: client to retry connections with a previously dead server.
1.69 matthew 1596:
1.14 foxr 1597: =cut
1.46 albertel 1598:
1.14 foxr 1599: sub ChildStatus {
1600: my $event = shift;
1601: my $watcher = $event->w;
1602:
1603: Debug(2, "Reporting child status because : ".$watcher->data);
1604: my $docdir = $perlvar{'lonDocRoot'};
1.67 albertel 1605:
1606: open(LOG,">>$docdir/lon-status/loncstatus.txt");
1607: flock(LOG,LOCK_EX);
1608: print LOG $$."\t".$RemoteHost."\t".$Status."\t".
1.14 foxr 1609: $RecentLogEntry."\n";
1.38 foxr 1610: #
1611: # Write out information about each of the connections:
1612: #
1.46 albertel 1613: if ($DebugLevel > 2) {
1.67 albertel 1614: print LOG "Active connection statuses: \n";
1.46 albertel 1615: my $i = 1;
1616: print STDERR "================================= Socket Status Dump:\n";
1617: foreach my $item (keys %ActiveConnections) {
1618: my $Socket = $ActiveConnections{$item}->data;
1619: my $state = $Socket->GetState();
1.67 albertel 1620: print LOG "Connection $i State: $state\n";
1.46 albertel 1621: print STDERR "---------------------- Connection $i \n";
1.48 foxr 1622: $Socket->Dump(-1); # Ensure it gets dumped..
1.46 albertel 1623: $i++;
1624: }
1.38 foxr 1625: }
1.67 albertel 1626: flock(LOG,LOCK_UN);
1627: close(LOG);
1.22 foxr 1628: $ConnectionRetriesLeft = $ConnectionRetries;
1.70 albertel 1629: UpdateStatus();
1.14 foxr 1630: }
1631:
1.1 foxr 1632: =pod
1.3 albertel 1633:
1.10 foxr 1634: =head2 SignalledToDeath
1635:
1636: Called in response to a signal that causes a chid process to die.
1637:
1638: =cut
1639:
1640:
1641: sub SignalledToDeath {
1.14 foxr 1642: my $event = shift;
1643: my $watcher= $event->w;
1644:
1645: Debug(2,"Signalled to death! via ".$watcher->data);
1.17 foxr 1646: my ($signal) = $watcher->data;
1.10 foxr 1647: chomp($signal);
1648: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1649: ."died through "."\"$signal\"");
1.68 albertel 1650: #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1651: # ."\"$signal\"");
1.12 foxr 1652: exit 0;
1.10 foxr 1653:
1654: }
1.16 foxr 1655:
1.69 matthew 1656: =pod
1657:
1.16 foxr 1658: =head2 ToggleDebug
1659:
1660: This sub toggles trace debugging on and off.
1661:
1662: =cut
1663:
1664: sub ToggleDebug {
1665: my $Current = $DebugLevel;
1666: $DebugLevel = $NextDebugLevel;
1667: $NextDebugLevel = $Current;
1668:
1669: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1670:
1671: }
1672:
1.69 matthew 1673: =pod
1674:
1.1 foxr 1675: =head2 ChildProcess
1676:
1677: This sub implements a child process for a single lonc daemon.
1.61 foxr 1678: Optional parameter:
1679: $socket - if provided, this is a socket already open for listen
1680: on the client socket. Otherwise, a new listen is set up.
1.1 foxr 1681:
1682: =cut
1683:
1684: sub ChildProcess {
1.80 albertel 1685: # We've inherited all the
1.62 foxr 1686: # events of our parent and those have to be cancelled or else
1687: # all holy bloody chaos will result.. trust me, I already made
1688: # >that< mistake.
1689:
1690: my $host = GetServerHost();
1691: foreach my $listener (keys %parent_dispatchers) {
1692: my $watcher = $parent_dispatchers{$listener};
1693: my $s = $watcher->fd;
1694: if ($listener ne $host) { # Close everyone but me.
1695: Debug(5, "Closing listen socket for $listener");
1696: $s->close();
1697: }
1698: Debug(5, "Killing watcher for $listener");
1699:
1700: $watcher->cancel();
1.65 foxr 1701: delete($parent_dispatchers{$listener});
1.62 foxr 1702:
1703: }
1.65 foxr 1704:
1705: # kill off the parent's signal handlers too!
1706: #
1707:
1708: for my $handler (keys %parent_handlers) {
1709: my $watcher = $parent_handlers{$handler};
1710: $watcher->cancel();
1711: delete($parent_handlers{$handler});
1712: }
1713:
1.64 foxr 1714: $I_am_child = 1; # Seems like in spite of it all I may still getting
1715: # parent event dispatches.. flag I'm a child.
1.1 foxr 1716:
1717:
1.14 foxr 1718: #
1719: # Signals must be handled by the Event framework...
1.61 foxr 1720: #
1.14 foxr 1721:
1722: Event->signal(signal => "QUIT",
1723: cb => \&SignalledToDeath,
1724: data => "QUIT");
1725: Event->signal(signal => "HUP",
1726: cb => \&ChildStatus,
1727: data => "HUP");
1728: Event->signal(signal => "USR1",
1729: cb => \&ChildStatus,
1730: data => "USR1");
1.39 foxr 1731: Event->signal(signal => "USR2",
1732: cb => \&ToggleTransactionLogging);
1.16 foxr 1733: Event->signal(signal => "INT",
1734: cb => \&ToggleDebug,
1735: data => "INT");
1.1 foxr 1736:
1.93 foxr 1737: # Block the pipe signal we'll get when the socket disconnects. We detect
1738: # socket disconnection via send/receive failures. On disconnect, the
1739: # socket becomes readable .. which will force the disconnect detection.
1740:
1741: my $set = POSIX::SigSet->new(SIGPIPE);
1742: sigprocmask(SIG_BLOCK, $set);
1743:
1.62 foxr 1744: # Figure out if we got passed a socket or need to open one to listen for
1745: # client requests.
1746:
1.61 foxr 1747: my ($socket) = @_;
1748: if (!$socket) {
1749:
1750: $socket = SetupLoncListener();
1751: }
1.62 foxr 1752: # Establish an event to listen for client connection requests.
1753:
1754:
1.59 foxr 1755: Event->io(cb => \&NewClient,
1756: poll => 'r',
1757: desc => 'Lonc Listener Unix Socket',
1758: fd => $socket);
1.1 foxr 1759:
1.76 albertel 1760: $Event::DebugLevel = $DebugLevel;
1.1 foxr 1761:
1762: Debug(9, "Making initial lond connection for ".$RemoteHost);
1763:
1764: # Setup the initial server connection:
1765:
1.62 foxr 1766: # &MakeLondConnection(); // let first work request do it.
1.10 foxr 1767:
1.80 albertel 1768: # need to accept the connection since the event may not fire.
1.62 foxr 1769:
1.80 albertel 1770: &accept_client($socket);
1.5 foxr 1771:
1.1 foxr 1772: Debug(9,"Entering event loop");
1773: my $ret = Event::loop(); # Start the main event loop.
1774:
1775:
1.64 foxr 1776: &child_exit (-1,"Main event loop exited!!!");
1.1 foxr 1777: }
1778:
1779: # Create a new child for host passed in:
1780:
1781: sub CreateChild {
1.81 albertel 1782: my ($host, $hostid) = @_;
1.52 foxr 1783:
1.12 foxr 1784: my $sigset = POSIX::SigSet->new(SIGINT);
1785: sigprocmask(SIG_BLOCK, $sigset);
1.1 foxr 1786: $RemoteHost = $host;
1.91 foxr 1787: ShowStatus('Parent keeping the flock'); # Update time in status message.
1.9 foxr 1788: Log("CRITICAL", "Forking server for ".$host);
1.23 foxr 1789: my $pid = fork;
1.1 foxr 1790: if($pid) { # Parent
1.17 foxr 1791: $RemoteHost = "Parent";
1.83 albertel 1792: $ChildPid{$pid} = $host;
1.12 foxr 1793: sigprocmask(SIG_UNBLOCK, $sigset);
1.82 albertel 1794: undef(@all_host_ids);
1.1 foxr 1795: } else { # child.
1.81 albertel 1796: $RemoteHostId = $hostid;
1.5 foxr 1797: ShowStatus("Connected to ".$RemoteHost);
1.23 foxr 1798: $SIG{INT} = 'DEFAULT';
1.12 foxr 1799: sigprocmask(SIG_UNBLOCK, $sigset);
1.81 albertel 1800: &ChildProcess(); # Does not return.
1.1 foxr 1801: }
1.61 foxr 1802: }
1.1 foxr 1803:
1.61 foxr 1804: # parent_client_connection:
1805: # Event handler that processes client connections for the parent process.
1806: # This sub is called when the parent is listening on a socket and
1807: # a connection request arrives. We must:
1808: # Start a child process to accept the connection request.
1809: # Kill our listen on the socket.
1810: # Parameter:
1811: # event - The event object that was created to monitor this socket.
1812: # event->w->fd is the socket.
1813: # Returns:
1814: # NONE
1815: #
1816: sub parent_client_connection {
1.62 foxr 1817: if ($I_am_child) {
1818: # Should not get here, but seem to anyway:
1819: &Debug(5," Child caught parent client connection event!!");
1820: my ($event) = @_;
1821: my $watcher = $event->w;
1822: $watcher->cancel(); # Try to kill it off again!!
1823: } else {
1824: &Debug(9, "parent_client_connection");
1825: my ($event) = @_;
1826: my $watcher = $event->w;
1827: my $socket = $watcher->fd;
1.81 albertel 1828: my $connection = $socket->accept(); # Accept the client connection.
1829: Event->io(cb => \&get_remote_hostname,
1830: poll => 'r',
1831: data => "",
1832: fd => $connection);
1.77 albertel 1833: }
1834: }
1835:
1836: sub get_remote_hostname {
1.82 albertel 1837: my ($event) = @_;
1838: my $watcher = $event->w;
1839: my $socket = $watcher->fd;
1840:
1841: my $thisread;
1842: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1843: Debug(8, "rcv: data length = ".length($thisread)." read =".$thisread);
1844: if (!defined($rv) || length($thisread) == 0) {
1845: # Likely eof on socket.
1846: Debug(5,"Client Socket closed on lonc for p_c_c");
1847: close($socket);
1848: $watcher->cancel();
1849: return;
1850: }
1851:
1852: my $data = $watcher->data().$thisread;
1853: $watcher->data($data);
1854: if($data =~ /\n$/) { # Request entirely read.
1855: chomp($data);
1856: } else {
1857: return;
1858: }
1.77 albertel 1859:
1.82 albertel 1860: &Debug(5,"Creating child for $data (parent_client_connection)");
1861: (my $hostname,my $lonid,@all_host_ids) = split(':',$data);
1.83 albertel 1862: $ChildHost{$hostname}++;
1863: if ($ChildHost{$hostname} == 1) {
1864: &CreateChild($hostname,$lonid);
1865: } else {
1866: &Log('WARNING',"Request for a second child on $hostname");
1867: }
1.82 albertel 1868: # Clean up the listen since now the child takes over until it exits.
1869: $watcher->cancel(); # Nolonger listening to this event
1870: $socket->send("done\n");
1871: $socket->close();
1.61 foxr 1872: }
1873:
1874: # parent_listen:
1875: # Opens a socket and starts a listen for the parent process on a client UNIX
1876: # domain socket.
1877: #
1878: # This involves:
1879: # Creating a socket for listen.
1880: # Removing any socket lock file
1881: # Adding an event handler for this socket becoming readable
1882: # To the parent's event dispatcher.
1883: # Parameters:
1884: # loncapa_host - LonCAPA cluster name of the host represented by the client
1885: # socket.
1886: # Returns:
1887: # NONE
1888: #
1889: sub parent_listen {
1890: my ($loncapa_host) = @_;
1891: Debug(5, "parent_listen: $loncapa_host");
1892:
1.78 albertel 1893: my ($socket,$file);
1894: if (!$loncapa_host) {
1895: $loncapa_host = 'common_parent';
1896: $file = $perlvar{'lonSockCreate'};
1897: } else {
1898: $file = &GetLoncSocketPath($loncapa_host);
1899: }
1900: $socket = &SetupLoncListener($loncapa_host,$file);
1901:
1.62 foxr 1902: $listening_to{$socket} = $loncapa_host;
1.61 foxr 1903: if (!$socket) {
1904: die "Unable to create a listen socket for $loncapa_host";
1905: }
1906:
1.78 albertel 1907: my $lock_file = $file.".lock";
1.61 foxr 1908: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1909:
1.77 albertel 1910: my $watcher =
1911: Event->io(cb => \&parent_client_connection,
1912: poll => 'r',
1913: desc => "Parent listener unix socket ($loncapa_host)",
1914: data => "",
1915: fd => $socket);
1.62 foxr 1916: $parent_dispatchers{$loncapa_host} = $watcher;
1.61 foxr 1917:
1918: }
1919:
1.77 albertel 1920: sub parent_clean_up {
1921: my ($loncapa_host) = @_;
1.87 albertel 1922: Debug(1, "parent_clean_up: $loncapa_host");
1.77 albertel 1923:
1924: my $socket_file = &GetLoncSocketPath($loncapa_host);
1925: unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
1926: my $lock_file = $socket_file.".lock";
1927: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1928: }
1929:
1.61 foxr 1930:
1.83 albertel 1931:
1932: # This sub initiates a listen on the common unix domain lonc client socket.
1933: # loncnew starts up with no children, and only spawns off children when a
1934: # connection request occurs on the common client unix socket. The spawned
1935: # child continues to run until it has been idle a while at which point it
1936: # eventually exits and once more the parent picks up the listen.
1.61 foxr 1937: #
1938: # Parameters:
1939: # NONE
1940: # Implicit Inputs:
1941: # The configuration file that has been read in by LondConnection.
1942: # Returns:
1943: # NONE
1944: #
1.77 albertel 1945: sub listen_on_common_socket {
1946: Debug(5, "listen_on_common_socket");
1.78 albertel 1947: &parent_listen();
1.77 albertel 1948: }
1949:
1.63 foxr 1950: # server_died is called whenever a child process exits.
1951: # Since this is dispatched via a signal, we must process all
1952: # dead children until there are no more left. The action
1953: # is to:
1954: # - Remove the child from the bookeeping hashes
1955: # - Re-establish a listen on the unix domain socket associated
1956: # with that host.
1957: # Parameters:
1958: # The event, but we don't actually care about it.
1959: sub server_died {
1960: &Debug(9, "server_died called...");
1961:
1962: while(1) { # Loop until waitpid nowait fails.
1963: my $pid = waitpid(-1, WNOHANG);
1964: if($pid <= 0) {
1965: return; # Nothing left to wait for.
1966: }
1967: # need the host to restart:
1968:
1.83 albertel 1969: my $host = $ChildPid{$pid};
1.63 foxr 1970: if($host) { # It's for real...
1971: &Debug(9, "Caught sigchild for $host");
1.83 albertel 1972: delete($ChildPid{$pid});
1973: delete($ChildHost{$host});
1.81 albertel 1974: &parent_clean_up($host);
1975:
1.63 foxr 1976: } else {
1977: &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
1978: }
1979: }
1980:
1981: }
1982:
1.1 foxr 1983: #
1984: # Parent process logic pass 1:
1985: # For each entry in the hosts table, we will
1986: # fork off an instance of ChildProcess to service the transactions
1987: # to that host. Each pid will be entered in a global hash
1988: # with the value of the key, the host.
1989: # The parent will then enter a loop to wait for process exits.
1990: # Each exit gets logged and the child gets restarted.
1991: #
1992:
1.5 foxr 1993: #
1994: # Fork and start in new session so hang-up isn't going to
1995: # happen without intent.
1996: #
1997:
1998:
1.6 foxr 1999:
2000:
1.8 foxr 2001:
1.6 foxr 2002:
2003: ShowStatus("Forming new session");
2004: my $childpid = fork;
2005: if ($childpid != 0) {
2006: sleep 4; # Give child a chacne to break to
2007: exit 0; # a new sesion.
2008: }
1.8 foxr 2009: #
2010: # Write my pid into the pid file so I can be located
2011: #
2012:
2013: ShowStatus("Parent writing pid file:");
1.23 foxr 2014: my $execdir = $perlvar{'lonDaemons'};
1.8 foxr 2015: open (PIDSAVE, ">$execdir/logs/lonc.pid");
2016: print PIDSAVE "$$\n";
2017: close(PIDSAVE);
1.6 foxr 2018:
1.17 foxr 2019:
2020:
1.6 foxr 2021: if (POSIX::setsid() < 0) {
2022: print "Could not create new session\n";
2023: exit -1;
2024: }
1.5 foxr 2025:
2026: ShowStatus("Forking node servers");
2027:
1.9 foxr 2028: Log("CRITICAL", "--------------- Starting children ---------------");
2029:
1.31 foxr 2030: LondConnection::ReadConfig; # Read standard config files.
1.1 foxr 2031:
1.80 albertel 2032: $RemoteHost = "[parent]";
1.81 albertel 2033: &listen_on_common_socket();
1.60 foxr 2034:
1.12 foxr 2035: $RemoteHost = "Parent Server";
1.1 foxr 2036:
2037: # Maintain the population:
1.5 foxr 2038:
2039: ShowStatus("Parent keeping the flock");
1.1 foxr 2040:
1.12 foxr 2041:
1.80 albertel 2042: # We need to setup a SIGChild event to handle the exit (natural or otherwise)
2043: # of the children.
1.61 foxr 2044:
1.80 albertel 2045: Event->signal(cb => \&server_died,
2046: desc => "Child exit handler",
2047: signal => "CHLD");
2048:
2049:
2050: # Set up all the other signals we set up.
2051:
2052: $parent_handlers{INT} = Event->signal(cb => \&Terminate,
2053: desc => "Parent INT handler",
2054: signal => "INT");
2055: $parent_handlers{TERM} = Event->signal(cb => \&Terminate,
2056: desc => "Parent TERM handler",
2057: signal => "TERM");
1.81 albertel 2058: $parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
2059: desc => "Parent HUP handler.",
2060: signal => "HUP");
1.80 albertel 2061: $parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
2062: desc => "Parent USR1 handler",
2063: signal => "USR1");
2064: $parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
2065: desc => "Parent USR2 handler.",
2066: signal => "USR2");
2067:
2068: # Start procdesing events.
2069:
2070: $Event::DebugLevel = $DebugLevel;
2071: Debug(9, "Parent entering event loop");
2072: my $ret = Event::loop();
2073: die "Main Event loop exited: $ret";
1.14 foxr 2074:
2075: =pod
2076:
2077: =head1 CheckKids
2078:
2079: Since kids do not die as easily in this implementation
2080: as the previous one, there is no need to restart the
2081: dead ones (all dead kids get restarted when they die!!)
2082: The only thing this function does is to pass USR1 to the
2083: kids so that they report their status.
2084:
2085: =cut
2086:
2087: sub CheckKids {
2088: Debug(2, "Checking status of children");
2089: my $docdir = $perlvar{'lonDocRoot'};
2090: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
2091: my $now=time;
2092: my $local=localtime($now);
2093: print $fh "LONC status $local - parent $$ \n\n";
1.65 foxr 2094: foreach my $host (keys %parent_dispatchers) {
2095: print $fh "LONC Parent process listening for $host\n";
2096: }
1.83 albertel 2097: foreach my $pid (keys %ChildPid) {
1.14 foxr 2098: Debug(2, "Sending USR1 -> $pid");
2099: kill 'USR1' => $pid; # Tell Child to report status.
2100: }
1.65 foxr 2101:
1.14 foxr 2102: }
1.24 foxr 2103:
2104: =pod
2105:
2106: =head1 UpdateKids
2107:
1.25 foxr 2108: parent's SIGUSR2 handler. This handler:
1.24 foxr 2109:
2110: =item
2111:
2112: Rereads the hosts file.
2113:
2114: =item
2115:
2116: Kills off (via sigint) children for hosts that have disappeared.
2117:
2118: =item
2119:
1.27 foxr 2120: QUITs children for hosts that already exist (this just forces a status display
1.24 foxr 2121: and resets the connection retry count for that host.
2122:
2123: =item
2124:
2125: Starts new children for hosts that have been added to the hosts.tab file since
2126: the start of the master program and maintains them.
2127:
2128: =cut
2129:
2130: sub UpdateKids {
1.27 foxr 2131:
1.25 foxr 2132: Log("INFO", "Updating connections via SIGUSR2");
1.27 foxr 2133:
1.65 foxr 2134: # I'm not sure what I was thinking in the first implementation.
2135: # someone will have to work hard to convince me the effect is any
2136: # different than Restart, especially now that we don't start up
2137: # per host servers automatically, may as well just restart.
2138: # The down side is transactions that are in flight will get timed out
2139: # (lost unless they are critical).
1.27 foxr 2140:
1.81 albertel 2141: &KillThemAll();
1.24 foxr 2142: }
2143:
1.14 foxr 2144:
1.13 foxr 2145: =pod
2146:
2147: =head1 Restart
2148:
2149: Signal handler for HUP... all children are killed and
2150: we self restart. This is an el-cheapo way to re read
2151: the config file.
2152:
2153: =cut
2154:
2155: sub Restart {
1.23 foxr 2156: &KillThemAll; # First kill all the children.
1.13 foxr 2157: Log("CRITICAL", "Restarting");
2158: my $execdir = $perlvar{'lonDaemons'};
2159: unlink("$execdir/logs/lonc.pid");
1.65 foxr 2160: exec("$executable");
1.10 foxr 2161: }
1.12 foxr 2162:
2163: =pod
2164:
2165: =head1 KillThemAll
2166:
2167: Signal handler that kills all children by sending them a
1.17 foxr 2168: SIGHUP. Responds to sigint and sigterm.
1.12 foxr 2169:
2170: =cut
2171:
1.10 foxr 2172: sub KillThemAll {
1.12 foxr 2173: Debug(2, "Kill them all!!");
1.85 albertel 2174:
2175: #local($SIG{CHLD}) = 'IGNORE';
2176: # Our children >will< die.
2177: # but we need to catch their death and cleanup after them in case this is
2178: # a restart set of kills
2179: my @allpids = keys(%ChildPid);
2180: foreach my $pid (@allpids) {
1.83 albertel 2181: my $serving = $ChildPid{$pid};
1.52 foxr 2182: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2183: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
1.17 foxr 2184: kill 'QUIT' => $pid;
1.12 foxr 2185: }
1.85 albertel 2186: ShowStatus("Finished killing child processes off.");
1.1 foxr 2187: }
1.12 foxr 2188:
1.52 foxr 2189:
2190: #
2191: # Kill all children via KILL. Just in case the
2192: # first shot didn't get them.
2193:
2194: sub really_kill_them_all_dammit
2195: {
2196: Debug(2, "Kill them all Dammit");
2197: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
1.83 albertel 2198: foreach my $pid (keys %ChildPid) {
2199: my $serving = $ChildPid{$pid};
1.52 foxr 2200: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2201: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2202: kill 'KILL' => $pid;
1.83 albertel 2203: delete($ChildPid{$pid});
1.52 foxr 2204: my $execdir = $perlvar{'lonDaemons'};
2205: unlink("$execdir/logs/lonc.pid");
2206: }
2207: }
1.69 matthew 2208:
1.14 foxr 2209: =pod
2210:
2211: =head1 Terminate
2212:
2213: Terminate the system.
2214:
2215: =cut
2216:
2217: sub Terminate {
1.52 foxr 2218: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2219: &KillThemAll;
2220: #
2221: # By now they really should all be dead.. but just in case
2222: # send them all SIGKILL's after a bit of waiting:
2223:
2224: sleep(4);
2225: &Log("CRITICAL", "Now kill children nasty");
2226: &really_kill_them_all_dammit;
1.17 foxr 2227: Log("CRITICAL","Master process exiting");
2228: exit 0;
1.14 foxr 2229:
2230: }
1.81 albertel 2231:
2232: sub my_hostname {
2233: use Sys::Hostname;
2234: my $name = &hostname();
2235: &Debug(9,"Name is $name");
2236: return $name;
2237: }
2238:
1.12 foxr 2239: =pod
1.1 foxr 2240:
2241: =head1 Theory
1.3 albertel 2242:
2243: The event class is used to build this as a single process with an
2244: event driven model. The following events are handled:
1.1 foxr 2245:
2246: =item UNIX Socket connection Received
2247:
2248: =item Request data arrives on UNIX data transfer socket.
2249:
2250: =item lond connection becomes writable.
2251:
2252: =item timer fires at 1 second intervals.
2253:
2254: All sockets are run in non-blocking mode. Timeouts managed by the timer
2255: handler prevents hung connections.
2256:
2257: Key data structures:
2258:
1.3 albertel 2259: =item RequestQueue
2260:
2261: A queue of requests received from UNIX sockets that are
2262: waiting for a chance to be forwarded on a lond connection socket.
2263:
2264: =item ActiveConnections
2265:
2266: A hash of lond connections that have transactions in process that are
2267: available to be timed out.
2268:
2269: =item ActiveTransactions
2270:
2271: A hash indexed by lond connections that contain the client reply
2272: socket for each connection that has an active transaction on it.
2273:
2274: =item IdleConnections
2275:
2276: A hash of lond connections that have no work to do. These connections
2277: can be closed if they are idle for a long enough time.
1.1 foxr 2278:
2279: =cut
1.88 foxr 2280:
2281: =pod
2282:
2283: =head1 Log messages
2284:
2285: The following is a list of log messages that can appear in the
2286: lonc.log file. Each log file has a severity and a message.
2287:
2288: =over 2
2289:
2290: =item Warning A socket timeout was detected
2291:
2292: If there are pending transactions in the socket's queue,
2293: they are failed (saved if critical). If the connection
2294: retry count gets exceeded by this, the
2295: remote host is marked as dead.
2296: Called when timeouts occured during the connection and
2297: connection dialog with a remote host.
2298:
2299: =item Critical Host makred DEAD <hostname>
2300:
2301: The numer of retry counts for contacting a host was
2302: exceeded. The host is marked dead an no
2303: further attempts will be made by that child.
2304:
2305: =item Info lonc pipe client hung up on us
2306:
2307: Write to the client pipe indicated no data transferred
2308: Socket to remote host is shut down. Reply to the client
2309: is discarded. Note: This is commented out in &ClientWriteable
2310:
2311: =item Success Reply from lond: <data>
2312:
2313: Can be enabled for debugging by setting LogTransactions to nonzero.
2314: Indicates a successful transaction with lond, <data> is the data received
2315: from the remote lond.
2316:
2317: =item Success A delayed transaction was completed
2318:
2319: A transaction that must be reliable was executed and completed
2320: as lonc restarted. This is followed by a mesage of the form
2321:
2322: S: client-name : request
2323:
2324: =item WARNING Failing transaction <cmd>:<subcmd>
2325:
2326: Transaction failed on a socket, but the failure retry count for the remote
2327: node has not yet been exhausted (the node is not yet marked dead).
2328: cmd is the command, subcmd is the subcommand. This results from a con_lost
2329: when communicating with lond.
2330:
2331: =item WARNING Shutting down a socket
2332:
2333: Called when a socket is being closed to lond. This is emitted both when
2334: idle pruning is being done and when the socket has been disconnected by the remote.
2335:
2336: =item WARNING Lond connection lost.
2337:
2338: Called when a read from lond's socket failed indicating lond has closed the
2339: connection or died. This should be followed by one or more
2340:
2341: "WARNING Failing transaction..." msgs for each in-flight or queued transaction.
2342:
2343: =item INFO Connected to lond version: <version>
2344:
2345: When connection negotiation is complete, the lond version is requested and logged here.
2346:
2347: =item SUCCESS Connection n to host now ready for action
2348:
2349: Emitted when connection has been completed with lond. n is then number of
2350: concurrent connections and host, the host to which the connection has just
2351: been established.
2352:
2353: =item WARNING Connection to host has been disconnected
2354:
2355: Write to a lond resulted in failure status. Connection to lond is dropped.
2356:
2357: =item SUCCESS Created connection n to host host
2358:
2359: Initial connection request to host..(before negotiation).
2360:
2361: =item CRITICAL Request Close Connection ... exiting
2362:
2363: Client has sent "close_connection_exit" The loncnew server is exiting.
2364:
2365: =item INFO Resetting Connection Retries
2366:
2367: Client has sent "reset_retries" The lond connection retries are reset to zero for the
2368: corresponding lond.
2369:
2370: =item SUCCESS Transaction <data>
2371:
2372: Only emitted if the global variable $LogTransactions was set to true.
2373: A client has requested a lond transaction <data> is the contents of the request.
2374:
2375: =item SUCCESS Toggled transaction logging <LogTransactions>
2376:
2377: The state of the $LogTransactions global has been toggled, and its current value
2378: (after being toggled) is displayed. When non zero additional logging of transactions
2379: is enabled for debugging purposes. Transaction logging is toggled on receipt of a USR2
2380: signal.
2381:
2382: =item CRITICAL Abnormal exit. Child <pid> for <host> died thorugh signal.
2383:
2384: QUIT signal received. lonc child process is exiting.
2385:
2386: =item SUCCESS New debugging level for <RemoteHost> now <DebugLevel>
2387:
2388: Debugging toggled for the host loncnew is talking with.
2389: Currently debugging is a level based scheme with higher number
2390: conveying more information. The daemon starts out at
2391: DebugLevel 0 and can toggle back and forth between that and
2392: DebugLevel 2 These are controlled by
2393: the global variables $DebugLevel and $NextDebugLevel
2394: The debug level can go up to 9.
2395: SIGINT toggles the debug level. The higher the debug level the
2396: more debugging information is spewed. See the Debug
2397: sub in loncnew.
2398:
2399: =item CRITICAL Forking server for host
2400:
2401: A child is being created to service requests for the specified host.
2402:
2403:
2404: =item WARNING Request for a second child on hostname
2405:
2406: Somehow loncnew was asked to start a second child on a host that already had a child
2407: servicing it. This request is not honored, but themessage is emitted. This could happen
2408: due to a race condition. When a client attempts to contact loncnew for a new host, a child
2409: is forked off to handle the requests for that server. The parent then backs off the Unix
2410: domain socket leaving it for the child to service all requests. If in the time between
2411: creating the child, and backing off, a new connection request comes in to the unix domain
2412: socket, this could trigger (unlikely but remotely possible),.
2413:
2414: =item CRITICAL ------ Starting Children ----
2415:
2416: This message should probably be changed to "Entering event loop" as the loncnew only starts
2417: children as needed. This message is emitted as new events are established and
2418: the event processing loop is entered.
2419:
2420: =item INFO Updating connections via SIGUSR2
2421:
2422: SIGUSR2 received. The original code would kill all clients, re-read the host file,
2423: then restart children for each host. Now that childrean aree started on demand, this
2424: just kills all child processes and lets requests start them as needed again.
2425:
2426:
2427: =item CRITICAL Restarting
2428:
2429: SigHUP received. all the children are killed and the script exec's itself to start again.
2430:
2431: =item CRITICAL Nicely killing lonc for host pid = <pid>
2432:
2433: Attempting to kill the child that is serving the specified host (pid given) cleanly via
2434: SIGQUIT The child should handle that, clean up nicely and exit.
2435:
2436: =item CRITICAL Nastily killing lonc for host pid = <pid>
2437:
2438: The child specified did not die when requested via SIGQUIT. Therefore it is killed
2439: via SIGKILL.
2440:
2441: =item CRITICAL Asked to kill children.. first be nice..
2442:
2443: In the parent's INT handler. INT kills the child processes. This inidicate loncnew
2444: is about to attempt to kill all known children via SIGQUIT. This message should be followed
2445: by one "Nicely killing" message for each extant child.
2446:
2447: =item CRITICAL Now kill children nasty
2448:
2449: In the parent's INT handler. remaining children are about to be killed via
2450: SIGKILL. Should be followed by a Nastily killing... for each lonc child that
2451: refused to die.
2452:
2453: =item CRITICAL Master process exiting
2454:
2455: In the parent's INT handler. just prior to the exit 0 call.
2456:
2457: =back
2458:
2459: =cut
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>