Annotation of loncom/loncnew, revision 1.97
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: #
1.97 ! raeburn 5: # $Id: loncnew,v 1.96 2011/01/25 11:02:35 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: #
1.97 ! raeburn 29: # new lonc handles n request out over 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:
1.97 ! raeburn 352: Invoked each timer tick.
1.3 albertel 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:
1.97 ! raeburn 491: This callback is established when a transaction response is
! 492: available from lond. The response is forwarded to the unix socket
1.3 albertel 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) {
1.96 foxr 562: # No action taken...the socket will be writable firing the event again
563: # which will result in a retry of the write.
1.6 foxr 564: } else { # Unanticipated errno.
565: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
566: $Watcher->cancel; # Stop the watcher.
567: $Socket->shutdown(2); # Kill connection
568: $Socket->close(); # Close the socket.
569: }
1.1 foxr 570:
571: }
1.6 foxr 572: } else {
573: $Watcher->cancel(); # A delayed request...just cancel.
1.84 albertel 574: return;
1.1 foxr 575: }
576: }
577:
578: =pod
1.3 albertel 579:
1.1 foxr 580: =head2 CompleteTransaction
1.3 albertel 581:
582: Called when the reply data has been received for a lond
1.1 foxr 583: transaction. The reply data must now be sent to the
584: ultimate client on the other end of the Unix socket. This is
585: done by setting up a writable event for the socket with the
586: data the reply data.
1.3 albertel 587:
1.1 foxr 588: Parameters:
1.3 albertel 589:
590: =item Socket
591:
1.97 ! raeburn 592: Socket on which the lond transaction occurred. This is a
! 593: LondConnection. The data received are in the TransactionReply member.
1.3 albertel 594:
1.7 foxr 595: =item Transaction
1.3 albertel 596:
1.7 foxr 597: The transaction that is being completed.
1.1 foxr 598:
599: =cut
1.3 albertel 600:
1.1 foxr 601: sub CompleteTransaction {
1.29 foxr 602: &Debug(5,"Complete transaction");
1.47 foxr 603:
604: my ($Socket, $Transaction) = @_;
1.1 foxr 605:
1.7 foxr 606: if (!$Transaction->isDeferred()) { # Normal transaction
607: my $data = $Socket->GetReply(); # Data to send.
1.39 foxr 608: if($LogTransactions) {
609: Log("SUCCESS", "Reply from lond: '$data'");
610: }
1.7 foxr 611: StartClientReply($Transaction, $data);
612: } else { # Delete deferred transaction file.
1.9 foxr 613: Log("SUCCESS", "A delayed transaction was completed");
1.86 albertel 614: LogPerm("S:".$Transaction->getClient().":".$Transaction->getRequest());
615: unlink($Transaction->getFile());
1.7 foxr 616: }
1.6 foxr 617: }
1.42 foxr 618:
1.6 foxr 619: =pod
1.42 foxr 620:
1.6 foxr 621: =head1 StartClientReply
622:
623: Initiates a reply to a client where the reply data is a parameter.
624:
1.7 foxr 625: =head2 parameters:
626:
627: =item Transaction
628:
629: The transaction for which we are responding to the client.
630:
631: =item data
632:
1.97 ! raeburn 633: The data to send to apache client.
1.7 foxr 634:
1.6 foxr 635: =cut
1.42 foxr 636:
1.6 foxr 637: sub StartClientReply {
1.1 foxr 638:
1.47 foxr 639: my ($Transaction, $data) = @_;
1.12 foxr 640:
1.7 foxr 641: my $Client = $Transaction->getClient();
642:
1.1 foxr 643: &Debug(8," Reply was: ".$data);
644: my $Serial = $ActiveClients{$Client};
645: my $desc = sprintf("Connection to lonc client %d",
646: $Serial);
647: Event->io(fd => $Client,
648: poll => "w",
649: desc => $desc,
650: cb => \&ClientWritable,
651: data => $data);
652: }
1.42 foxr 653:
1.4 foxr 654: =pod
1.42 foxr 655:
1.4 foxr 656: =head2 FailTransaction
657:
658: Finishes a transaction with failure because the associated lond socket
1.7 foxr 659: disconnected. There are two possibilities:
660: - The transaction is deferred: in which case we just quietly
661: delete the transaction since there is no client connection.
662: - The transaction is 'live' in which case we initiate the sending
663: of "con_lost" to the client.
664:
1.42 foxr 665: Deleting the transaction means killing it from the %ActiveTransactions hash.
1.4 foxr 666:
667: Parameters:
668:
669: =item client
670:
1.7 foxr 671: The LondTransaction we are failing.
672:
1.42 foxr 673:
1.4 foxr 674: =cut
675:
676: sub FailTransaction {
1.7 foxr 677: my $transaction = shift;
1.52 foxr 678:
679: # If the socket is dead, that's already logged.
680:
681: if ($ConnectionRetriesLeft > 0) {
682: Log("WARNING", "Failing transaction "
1.71 albertel 683: .$transaction->getLoggableRequest());
1.52 foxr 684: }
1.71 albertel 685: Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
1.10 foxr 686: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
1.11 foxr 687: my $client = $transaction->getClient();
1.30 foxr 688: Debug(1," Replying con_lost to ".$transaction->getRequest());
1.11 foxr 689: StartClientReply($transaction, "con_lost\n");
1.7 foxr 690: }
1.4 foxr 691:
692: }
693:
694: =pod
1.69 matthew 695:
1.6 foxr 696: =head1 EmptyQueue
1.7 foxr 697:
1.6 foxr 698: Fails all items in the work queue with con_lost.
1.7 foxr 699: Note that each item in the work queue is a transaction.
700:
1.6 foxr 701: =cut
1.69 matthew 702:
1.6 foxr 703: sub EmptyQueue {
1.22 foxr 704: $ConnectionRetriesLeft--; # Counts as connection failure too.
1.6 foxr 705: while($WorkQueue->Count()) {
1.10 foxr 706: my $request = $WorkQueue->dequeue(); # This is a transaction
1.7 foxr 707: FailTransaction($request);
1.6 foxr 708: }
709: }
710:
711: =pod
1.4 foxr 712:
1.9 foxr 713: =head2 CloseAllLondConnections
714:
715: Close all connections open on lond prior to exit e.g.
716:
717: =cut
1.69 matthew 718:
1.9 foxr 719: sub CloseAllLondConnections {
1.23 foxr 720: foreach my $Socket (keys %ActiveConnections) {
1.42 foxr 721: if(exists($ActiveTransactions{$Socket})) {
722: FailTransaction($ActiveTransactions{$Socket});
723: }
724: KillSocket($Socket);
1.9 foxr 725: }
726: }
727:
728: =pod
729:
1.4 foxr 730: =head2 KillSocket
731:
732: Destroys a socket. This function can be called either when a socket
733: has died of 'natural' causes or because a socket needs to be pruned due to
734: idleness. If the socket has died naturally, if there are no longer any
735: live connections a new connection is created (in case there are transactions
736: in the queue). If the socket has been pruned, it is never re-created.
737:
738: Parameters:
1.1 foxr 739:
1.4 foxr 740: =item Socket
741:
742: The socket to kill off.
743:
744: =item Restart
745:
1.97 ! raeburn 746: non-zero if we are allowed to create a new connection.
1.4 foxr 747:
1.69 matthew 748: =cut
1.4 foxr 749:
750: sub KillSocket {
751: my $Socket = shift;
752:
1.17 foxr 753: Log("WARNING", "Shutting down a socket");
1.9 foxr 754: $Socket->Shutdown();
755:
1.7 foxr 756: # If the socket came from the active connection set,
757: # delete its transaction... note that FailTransaction should
758: # already have been called!!!
759: # otherwise it came from the idle set.
760: #
1.4 foxr 761:
762: if(exists($ActiveTransactions{$Socket})) {
763: delete ($ActiveTransactions{$Socket});
764: }
765: if(exists($ActiveConnections{$Socket})) {
1.90 foxr 766: $ActiveConnections{$Socket}->cancel;
1.4 foxr 767: delete($ActiveConnections{$Socket});
1.37 albertel 768: $ConnectionCount--;
769: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
1.4 foxr 770: }
1.6 foxr 771: # If the connection count has gone to zero and there is work in the
772: # work queue, the work all gets failed with con_lost.
773: #
774: if($ConnectionCount == 0) {
1.22 foxr 775: EmptyQueue();
1.42 foxr 776: CloseAllLondConnections; # Should all already be closed but...
1.4 foxr 777: }
1.90 foxr 778: UpdateStatus();
1.4 foxr 779: }
1.1 foxr 780:
781: =pod
1.3 albertel 782:
1.1 foxr 783: =head2 LondReadable
1.3 albertel 784:
1.1 foxr 785: This function is called whenever a lond connection
786: is readable. The action is state dependent:
787:
1.3 albertel 788: =head3 State=Initialized
789:
1.97 ! raeburn 790: We're waiting for the challenge, this is a no-op until the
1.1 foxr 791: state changes.
1.3 albertel 792:
1.1 foxr 793: =head3 State=Challenged
1.3 albertel 794:
795: The challenge has arrived we need to transition to Writable.
1.1 foxr 796: The connection must echo the challenge back.
1.3 albertel 797:
1.1 foxr 798: =head3 State=ChallengeReplied
1.3 albertel 799:
1.97 ! raeburn 800: The challenge has been replied to. Then we are receiving the
1.1 foxr 801: 'ok' from the partner.
1.3 albertel 802:
1.40 foxr 803: =head3 State=ReadingVersionString
804:
805: We have requested the lond version and are reading the
806: version back. Upon completion, we'll store the version away
807: for future use(?).
808:
809: =head3 State=HostSet
810:
811: We have selected the domain name of our peer (multhomed hosts)
812: and are getting the reply (presumably ok) back.
813:
1.1 foxr 814: =head3 State=RequestingKey
1.3 albertel 815:
816: The ok has been received and we need to send the request for
1.1 foxr 817: an encryption key. Transition to writable for that.
1.3 albertel 818:
1.1 foxr 819: =head3 State=ReceivingKey
1.3 albertel 820:
821: The the key has been requested, now we are reading the new key.
822:
1.1 foxr 823: =head3 State=Idle
1.3 albertel 824:
825: The encryption key has been negotiated or we have finished
1.97 ! raeburn 826: reading data from the a transaction. If the callback data have
! 827: a client as well as the socket nformation, then we are
! 828: doing a transaction and the data received are relayed to the client
1.1 foxr 829: before the socket is put on the idle list.
1.3 albertel 830:
1.1 foxr 831: =head3 State=SendingRequest
1.3 albertel 832:
833: I do not think this state can be received here, but if it is,
1.1 foxr 834: the appropriate thing to do is to transition to writable, and send
835: the request.
1.3 albertel 836:
1.1 foxr 837: =head3 State=ReceivingReply
1.3 albertel 838:
839: We finished sending the request to the server and now transition
1.1 foxr 840: to readable to receive the reply.
841:
842: The parameter to this function are:
1.3 albertel 843:
1.1 foxr 844: The event. Implicit in this is the watcher and its data. The data
1.97 ! raeburn 845: contain at least the lond connection object and, if a
1.1 foxr 846: transaction is in progress, the socket attached to the local client.
847:
1.3 albertel 848: =cut
1.1 foxr 849:
850: sub LondReadable {
1.8 foxr 851:
1.41 albertel 852: my $Event = shift;
853: my $Watcher = $Event->w;
854: my $Socket = $Watcher->data;
855: my $client = undef;
1.40 foxr 856:
1.41 albertel 857: &Debug(6,"LondReadable called state = ".$Socket->GetState());
1.40 foxr 858:
859:
1.41 albertel 860: my $State = $Socket->GetState(); # All action depends on the state.
1.40 foxr 861:
1.41 albertel 862: SocketDump(6, $Socket);
863: my $status = $Socket->Readable();
1.40 foxr 864:
1.41 albertel 865: &Debug(2, "Socket->Readable returned: $status");
1.40 foxr 866:
1.41 albertel 867: if($status != 0) {
868: # bad return from socket read. Currently this means that
869: # The socket has become disconnected. We fail the transaction.
1.40 foxr 870:
1.41 albertel 871: Log("WARNING",
872: "Lond connection lost.");
873: if(exists($ActiveTransactions{$Socket})) {
874: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 875: } else {
876: # Socket is connecting and failed... need to mark
877: # no longer connecting.
878:
879: $LondConnecting = 0;
1.41 albertel 880: }
881: $Watcher->cancel();
882: KillSocket($Socket);
883: $ConnectionRetriesLeft--; # Counts as connection failure
884: return;
885: }
886: SocketDump(6,$Socket);
1.17 foxr 887:
1.41 albertel 888: $State = $Socket->GetState(); # Update in case of transition.
889: &Debug(6, "After read, state is ".$State);
1.1 foxr 890:
1.41 albertel 891: if($State eq "Initialized") {
1.1 foxr 892:
893:
1.41 albertel 894: } elsif ($State eq "ChallengeReceived") {
1.1 foxr 895: # The challenge must be echoed back; The state machine
896: # in the connection takes care of setting that up. Just
897: # need to transition to writable:
1.41 albertel 898:
899: $Watcher->cb(\&LondWritable);
900: $Watcher->poll("w");
1.1 foxr 901:
1.41 albertel 902: } elsif ($State eq "ChallengeReplied") {
1.1 foxr 903:
1.41 albertel 904: } elsif ($State eq "RequestingVersion") {
905: # Need to ask for the version... that is writiability:
1.1 foxr 906:
1.41 albertel 907: $Watcher->cb(\&LondWritable);
908: $Watcher->poll("w");
909:
910: } elsif ($State eq "ReadingVersionString") {
911: # Read the rest of the version string...
912: } elsif ($State eq "SetHost") {
913: # Need to request the actual domain get set...
914:
915: $Watcher->cb(\&LondWritable);
916: $Watcher->poll("w");
917: } elsif ($State eq "HostSet") {
918: # Reading the 'ok' from the peer.
919:
920: } elsif ($State eq "RequestingKey") {
1.1 foxr 921: # The ok was received. Now we need to request the key
922: # That requires us to be writable:
923:
1.41 albertel 924: $Watcher->cb(\&LondWritable);
925: $Watcher->poll("w");
1.1 foxr 926:
1.41 albertel 927: } elsif ($State eq "ReceivingKey") {
1.1 foxr 928:
1.41 albertel 929: } elsif ($State eq "Idle") {
1.40 foxr 930:
1.41 albertel 931: # This is as good a spot as any to get the peer version
932: # string:
1.40 foxr 933:
1.41 albertel 934: if($LondVersion eq "unknown") {
935: $LondVersion = $Socket->PeerVersion();
936: Log("INFO", "Connected to lond version: $LondVersion");
937: }
1.1 foxr 938: # If necessary, complete a transaction and then go into the
939: # idle queue.
1.22 foxr 940: # Note that a trasition to idle indicates a live lond
941: # on the other end so reset the connection retries.
942: #
1.41 albertel 943: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
944: $Watcher->cancel();
945: if(exists($ActiveTransactions{$Socket})) {
946: Debug(5,"Completing transaction!!");
947: CompleteTransaction($Socket,
948: $ActiveTransactions{$Socket});
949: } else {
1.95 foxr 950: my $count = $Socket->GetClientData();
951: Log("SUCCESS", "Connection ".$count." to "
1.41 albertel 952: .$RemoteHost." now ready for action");
953: }
954: ServerToIdle($Socket); # Next work unit or idle.
1.54 foxr 955:
956: #
957: $LondConnecting = 0; # Best spot I can think of for this.
958: #
1.6 foxr 959:
1.41 albertel 960: } elsif ($State eq "SendingRequest") {
1.1 foxr 961: # We need to be writable for this and probably don't belong
962: # here inthe first place.
963:
1.73 albertel 964: Debug(6, "SendingRequest state encountered in readable");
1.41 albertel 965: $Watcher->poll("w");
966: $Watcher->cb(\&LondWritable);
1.1 foxr 967:
1.41 albertel 968: } elsif ($State eq "ReceivingReply") {
1.1 foxr 969:
970:
1.41 albertel 971: } else {
972: # Invalid state.
973: Debug(4, "Invalid state in LondReadable");
974: }
1.1 foxr 975: }
1.3 albertel 976:
1.1 foxr 977: =pod
1.3 albertel 978:
1.1 foxr 979: =head2 LondWritable
1.3 albertel 980:
1.1 foxr 981: This function is called whenever a lond connection
982: becomes writable while there is a writeable monitoring
983: event. The action taken is very state dependent:
1.3 albertel 984:
1.1 foxr 985: =head3 State = Connected
1.3 albertel 986:
987: The connection is in the process of sending the 'init' hailing to the
1.97 ! raeburn 988: lond on the remote end. The connection object's Writable member is
1.3 albertel 989: called. On error, ConnectionError is called to destroy the connection
990: and remove it from the ActiveConnections hash
991:
1.1 foxr 992: =head3 Initialized
1.3 albertel 993:
994: 'init' has been sent, writability monitoring is removed and
995: readability monitoring is started with LondReadable as the callback.
996:
1.1 foxr 997: =head3 ChallengeReceived
1.3 albertel 998:
999: The connection has received the who are you challenge from the remote
1000: system, and is in the process of sending the challenge
1001: response. Writable is called.
1002:
1.1 foxr 1003: =head3 ChallengeReplied
1.3 albertel 1004:
1005: The connection has replied to the initial challenge The we switch to
1006: monitoring readability looking for the server to reply with 'ok'.
1007:
1.1 foxr 1008: =head3 RequestingKey
1.3 albertel 1009:
1010: The connection is in the process of requesting its encryption key.
1011: Writable is called.
1012:
1.1 foxr 1013: =head3 ReceivingKey
1.3 albertel 1014:
1015: The connection has sent the request for a key. Switch to readability
1016: monitoring to accept the key
1017:
1.1 foxr 1018: =head3 SendingRequest
1.3 albertel 1019:
1020: The connection is in the process of sending a request to the server.
1021: This request is part of a client transaction. All the states until
1022: now represent the client setup protocol. Writable is called.
1023:
1.1 foxr 1024: =head3 ReceivingReply
1025:
1.3 albertel 1026: The connection has sent a request. Now it must receive a reply.
1027: Readability monitoring is requested.
1028:
1029: This function is an event handler and therefore receives as
1.1 foxr 1030: a parameter the event that has fired. The data for the watcher
1031: of this event is a reference to a list of one or two elements,
1032: depending on state. The first (and possibly only) element is the
1033: socket. The second (present only if a request is in progress)
1034: is the socket on which to return a reply to the caller.
1035:
1036: =cut
1.3 albertel 1037:
1.1 foxr 1038: sub LondWritable {
1039: my $Event = shift;
1040: my $Watcher = $Event->w;
1.8 foxr 1041: my $Socket = $Watcher->data;
1042: my $State = $Socket->GetState();
1.1 foxr 1043:
1.8 foxr 1044: Debug(6,"LondWritable State = ".$State."\n");
1.1 foxr 1045:
1.8 foxr 1046:
1.1 foxr 1047: # Figure out what to do depending on the state of the socket:
1048:
1049:
1050:
1051:
1052: SocketDump(6,$Socket);
1053:
1.42 foxr 1054: # If the socket is writable, we must always write.
1055: # Only by writing will we undergo state transitions.
1056: # Old logic wrote in state specific code below, however
1057: # That forces us at least through another invocation of
1058: # this function after writability is possible again.
1059: # This logic also factors out common code for handling
1060: # write failures... in all cases, write failures
1061: # Kill the socket.
1062: # This logic makes the branches of the >big< if below
1063: # so that the writing states are actually NO-OPs.
1064:
1065: if ($Socket->Writable() != 0) {
1.43 albertel 1066: # The write resulted in an error.
1067: # We'll treat this as if the socket got disconnected:
1068: Log("WARNING", "Connection to ".$RemoteHost.
1069: " has been disconnected");
1070: if(exists($ActiveTransactions{$Socket})) {
1071: FailTransaction($ActiveTransactions{$Socket});
1.56 foxr 1072: } else {
1073: # In the process of conneting, so need to turn that off.
1074:
1075: $LondConnecting = 0;
1.43 albertel 1076: }
1077: $Watcher->cancel();
1078: KillSocket($Socket);
1079: return;
1.42 foxr 1080: }
1081:
1082:
1083:
1.41 albertel 1084: if ($State eq "Connected") {
1.1 foxr 1085:
1.41 albertel 1086: # "init" is being sent...
1.42 foxr 1087:
1.41 albertel 1088: } elsif ($State eq "Initialized") {
1.4 foxr 1089:
1.41 albertel 1090: # Now that init was sent, we switch
1091: # to watching for readability:
1.1 foxr 1092:
1.41 albertel 1093: $Watcher->cb(\&LondReadable);
1094: $Watcher->poll("r");
1095:
1096: } elsif ($State eq "ChallengeReceived") {
1097: # We received the challenge, now we
1098: # are echoing it back. This is a no-op,
1099: # we're waiting for the state to change
1.1 foxr 1100:
1.41 albertel 1101: } elsif ($State eq "ChallengeReplied") {
1102: # The echo was sent back, so we switch
1103: # to watching readability.
1104:
1105: $Watcher->cb(\&LondReadable);
1106: $Watcher->poll("r");
1107: } elsif ($State eq "RequestingVersion") {
1108: # Sending the peer a version request...
1.42 foxr 1109:
1.41 albertel 1110: } elsif ($State eq "ReadingVersionString") {
1111: # Transition to read since we have sent the
1112: # version command and now just need to read the
1113: # version string from the peer:
1.40 foxr 1114:
1.41 albertel 1115: $Watcher->cb(\&LondReadable);
1116: $Watcher->poll("r");
1.40 foxr 1117:
1.41 albertel 1118: } elsif ($State eq "SetHost") {
1119: # Setting the remote domain...
1.42 foxr 1120:
1.41 albertel 1121: } elsif ($State eq "HostSet") {
1122: # Back to readable to get the ok.
1.40 foxr 1123:
1.41 albertel 1124: $Watcher->cb(\&LondReadable);
1125: $Watcher->poll("r");
1.40 foxr 1126:
1127:
1.41 albertel 1128: } elsif ($State eq "RequestingKey") {
1129: # At this time we're requesting the key.
1130: # again, this is essentially a no-op.
1131:
1132: } elsif ($State eq "ReceivingKey") {
1133: # Now we need to wait for the key
1134: # to come back from the peer:
1135:
1136: $Watcher->cb(\&LondReadable);
1137: $Watcher->poll("r");
1138:
1139: } elsif ($State eq "SendingRequest") {
1.40 foxr 1140:
1.41 albertel 1141: # At this time we are sending a request to the
1.1 foxr 1142: # peer... write the next chunk:
1143:
1.41 albertel 1144:
1145: } elsif ($State eq "ReceivingReply") {
1146: # The send has completed. Wait for the
1147: # data to come in for a reply.
1148: Debug(8,"Writable sent request/receiving reply");
1149: $Watcher->cb(\&LondReadable);
1150: $Watcher->poll("r");
1.1 foxr 1151:
1.41 albertel 1152: } else {
1153: # Control only passes here on an error:
1154: # the socket state does not match any
1155: # of the known states... so an error
1156: # must be logged.
1.1 foxr 1157:
1.41 albertel 1158: &Debug(4, "Invalid socket state ".$State."\n");
1159: }
1.1 foxr 1160:
1161: }
1.81 albertel 1162:
1.6 foxr 1163: =pod
1164:
1165: =cut
1.69 matthew 1166:
1.81 albertel 1167:
1.6 foxr 1168: sub QueueDelayed {
1.8 foxr 1169: Debug(3,"QueueDelayed called");
1170:
1.6 foxr 1171: my $path = "$perlvar{'lonSockDir'}/delayed";
1.8 foxr 1172:
1173: Debug(4, "Delayed path: ".$path);
1.6 foxr 1174: opendir(DIRHANDLE, $path);
1.75 albertel 1175:
1.82 albertel 1176: my $host_id_re = '(?:'.join('|',map {quotemeta($_)} (@all_host_ids)).')';
1.75 albertel 1177: my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
1.6 foxr 1178: closedir(DIRHANDLE);
1.75 albertel 1179: foreach my $dfname (sort(@alldelayed)) {
1180: my $reqfile = "$path/$dfname";
1181: my ($host_id) = ($dfname =~ /\.([^.]*)$/);
1182: Debug(4, "queueing ".$reqfile." for $host_id");
1.6 foxr 1183: my $Handle = IO::File->new($reqfile);
1184: my $cmd = <$Handle>;
1.8 foxr 1185: chomp $cmd; # There may or may not be a newline...
1.12 foxr 1186: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1.75 albertel 1187: my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
1.7 foxr 1188: $Transaction->SetDeferred($reqfile);
1189: QueueTransaction($Transaction);
1.6 foxr 1190: }
1191:
1192: }
1.1 foxr 1193:
1194: =pod
1.3 albertel 1195:
1.1 foxr 1196: =head2 MakeLondConnection
1.3 albertel 1197:
1198: Create a new lond connection object, and start it towards its initial
1.97 ! raeburn 1199: idleness. Once idle, it becomes eligible to receive transactions
1.3 albertel 1200: from the work queue. If the work queue is not empty when the
1201: connection is completed and becomes idle, it will dequeue an entry and
1202: start off on it.
1203:
1.1 foxr 1204: =cut
1.3 albertel 1205:
1.1 foxr 1206: sub MakeLondConnection {
1207: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1208: .GetServerPort());
1209:
1210: my $Connection = LondConnection->new(&GetServerHost(),
1.81 albertel 1211: &GetServerPort(),
1212: &GetHostId());
1.1 foxr 1213:
1.90 foxr 1214: if($Connection eq undef) {
1.9 foxr 1215: Log("CRITICAL","Failed to make a connection with lond.");
1.10 foxr 1216: $ConnectionRetriesLeft--;
1217: return 0; # Failure.
1.5 foxr 1218: } else {
1.22 foxr 1219:
1.82 albertel 1220: $LondConnecting = 1; # Connection in progress.
1.5 foxr 1221: # The connection needs to have writability
1222: # monitored in order to send the init sequence
1223: # that starts the whole authentication/key
1224: # exchange underway.
1225: #
1226: my $Socket = $Connection->GetSocket();
1.30 foxr 1227: if($Socket eq undef) {
1.64 foxr 1228: &child_exit(-1, "did not get a socket from the connection");
1.5 foxr 1229: } else {
1230: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1231: }
1.1 foxr 1232:
1.21 foxr 1233: $Connection->SetTimeoutCallback(\&SocketTimeout);
1234:
1.23 foxr 1235: my $event = Event->io(fd => $Socket,
1.5 foxr 1236: poll => 'w',
1237: cb => \&LondWritable,
1.8 foxr 1238: data => $Connection,
1.5 foxr 1239: desc => 'Connection to lond server');
1240: $ActiveConnections{$Connection} = $event;
1.52 foxr 1241: if ($ConnectionCount == 0) {
1242: &SetupTimer; # Need to handle timeouts with connections...
1243: }
1.5 foxr 1244: $ConnectionCount++;
1.95 foxr 1245: $Connection->SetClientData($ConnectionCount);
1.8 foxr 1246: Debug(4, "Connection count = ".$ConnectionCount);
1.6 foxr 1247: if($ConnectionCount == 1) { # First Connection:
1248: QueueDelayed;
1249: }
1.97 ! raeburn 1250: Log("SUCCESS", "Created connection ".$ConnectionCount
1.9 foxr 1251: ." to host ".GetServerHost());
1.10 foxr 1252: return 1; # Return success.
1.1 foxr 1253: }
1254:
1255: }
1.3 albertel 1256:
1.1 foxr 1257: =pod
1.3 albertel 1258:
1.1 foxr 1259: =head2 StartRequest
1.3 albertel 1260:
1261: Starts a lond request going on a specified lond connection.
1262: parameters are:
1263:
1264: =item $Lond
1265:
1266: Connection to the lond that will send the transaction and receive the
1267: reply.
1268:
1269: =item $Client
1270:
1.97 ! raeburn 1271: Connection to the client that is making this request. We got the
1.3 albertel 1272: request from this socket, and when the request has been relayed to
1273: lond and we get a reply back from lond it will get sent to this
1274: socket.
1275:
1276: =item $Request
1277:
1278: The text of the request to send.
1279:
1.1 foxr 1280: =cut
1281:
1282: sub StartRequest {
1.47 foxr 1283:
1284: my ($Lond, $Request) = @_;
1.1 foxr 1285:
1.7 foxr 1286: Debug(6, "StartRequest: ".$Request->getRequest());
1.1 foxr 1287:
1288: my $Socket = $Lond->GetSocket();
1289:
1.7 foxr 1290: $Request->Activate($Lond);
1291: $ActiveTransactions{$Lond} = $Request;
1.1 foxr 1292:
1.7 foxr 1293: $Lond->InitiateTransaction($Request->getRequest());
1.23 foxr 1294: my $event = Event->io(fd => $Socket,
1.1 foxr 1295: poll => "w",
1296: cb => \&LondWritable,
1297: data => $Lond,
1298: desc => "lond transaction connection");
1299: $ActiveConnections{$Lond} = $event;
1300: Debug(8," Start Request made watcher data with ".$event->data."\n");
1301: }
1302:
1303: =pod
1.3 albertel 1304:
1.1 foxr 1305: =head2 QueueTransaction
1.3 albertel 1306:
1307: If there is an idle lond connection, it is put to work doing this
1308: transaction. Otherwise, the transaction is placed in the work queue.
1309: If placed in the work queue and the maximum number of connections has
1310: not yet been created, a new connection will be started. Our goal is
1311: to eventually have a sufficient number of connections that the work
1312: queue will typically be empty. parameters are:
1313:
1314: =item Socket
1315:
1316: open on the lonc client.
1317:
1318: =item Request
1319:
1320: data to send to the lond.
1.1 foxr 1321:
1322: =cut
1.3 albertel 1323:
1.1 foxr 1324: sub QueueTransaction {
1325:
1.7 foxr 1326: my $requestData = shift; # This is a LondTransaction.
1327: my $cmd = $requestData->getRequest();
1328:
1329: Debug(6,"QueueTransaction: ".$cmd);
1.1 foxr 1330:
1331: my $LondSocket = $IdleConnections->pop();
1332: if(!defined $LondSocket) { # Need to queue request.
1.29 foxr 1333: Debug(5,"Must queue...");
1.1 foxr 1334: $WorkQueue->enqueue($requestData);
1.56 foxr 1335: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1336: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1337:
1.22 foxr 1338: if($ConnectionRetriesLeft > 0) {
1.29 foxr 1339: Debug(5,"Starting additional lond connection");
1.56 foxr 1340: if(&MakeLondConnection() == 0) {
1.22 foxr 1341: EmptyQueue(); # Fail transactions, can't make connection.
1.42 foxr 1342: CloseAllLondConnections; # Should all be closed but...
1.22 foxr 1343: }
1344: } else {
1345: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1.56 foxr 1346: $LondConnecting = 0;
1.22 foxr 1347: EmptyQueue(); # It's worse than that ... he's dead Jim.
1.42 foxr 1348: CloseAllLondConnections; # Should all be closed but..
1.17 foxr 1349: }
1.1 foxr 1350: }
1351: } else { # Can start the request:
1352: Debug(8,"Can start...");
1.7 foxr 1353: StartRequest($LondSocket, $requestData);
1.1 foxr 1354: }
1355: }
1356:
1.95 foxr 1357: #-------------------------- Lonc UNIX socket handling -------------------
1.1 foxr 1358: =pod
1.3 albertel 1359:
1.1 foxr 1360: =head2 ClientRequest
1.97 ! raeburn 1361:
1.3 albertel 1362: Callback that is called when data can be read from the UNIX domain
1363: socket connecting us with an apache server process.
1.1 foxr 1364:
1365: =cut
1366:
1367: sub ClientRequest {
1368: Debug(6, "ClientRequest");
1369: my $event = shift;
1370: my $watcher = $event->w;
1371: my $socket = $watcher->fd;
1372: my $data = $watcher->data;
1373: my $thisread;
1374:
1375: Debug(9, " Watcher named: ".$watcher->desc);
1376:
1377: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1378: Debug(8, "rcv: data length = ".length($thisread)
1379: ." read =".$thisread);
1.29 foxr 1380: unless (defined $rv && length($thisread)) {
1.1 foxr 1381: # Likely eof on socket.
1382: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1383: close($socket);
1384: $watcher->cancel();
1385: delete($ActiveClients{$socket});
1.10 foxr 1386: return;
1.1 foxr 1387: }
1388: Debug(8,"Data: ".$data." this read: ".$thisread);
1389: $data = $data.$thisread; # Append new data.
1390: $watcher->data($data);
1.44 albertel 1391: if($data =~ /\n$/) { # Request entirely read.
1.87 albertel 1392: if ($data eq "close_connection_exit\n") {
1.9 foxr 1393: Log("CRITICAL",
1394: "Request Close Connection ... exiting");
1395: CloseAllLondConnections();
1396: exit;
1.87 albertel 1397: } elsif ($data eq "reset_retries\n") {
1398: Log("INFO", "Resetting Connection Retries.");
1399: $ConnectionRetriesLeft = $ConnectionRetries;
1400: &UpdateStatus();
1401: my $Transaction = LondTransaction->new($data);
1402: $Transaction->SetClient($socket);
1403: StartClientReply($Transaction, "ok\n");
1404: $watcher->cancel();
1405: return;
1.9 foxr 1406: }
1.1 foxr 1407: Debug(8, "Complete transaction received: ".$data);
1.87 albertel 1408: if ($LogTransactions) {
1.39 foxr 1409: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1410: }
1.8 foxr 1411: my $Transaction = LondTransaction->new($data);
1.7 foxr 1412: $Transaction->SetClient($socket);
1413: QueueTransaction($Transaction);
1.1 foxr 1414: $watcher->cancel(); # Done looking for input data.
1415: }
1416:
1417: }
1418:
1.62 foxr 1419: #
1420: # Accept a connection request for a client (lonc child) and
1421: # start up an event watcher to keep an eye on input from that
1422: # Event. This can be called both from NewClient and from
1.80 albertel 1423: # ChildProcess.
1.62 foxr 1424: # Parameters:
1425: # $socket - The listener socket.
1426: # Returns:
1427: # NONE
1428: # Side Effects:
1429: # An event is made to watch the accepted connection.
1430: # Active clients hash is updated to reflect the new connection.
1431: # The client connection count is incremented.
1432: #
1433: sub accept_client {
1434: my ($socket) = @_;
1435:
1436: Debug(8, "Entering accept for lonc UNIX socket\n");
1437: my $connection = $socket->accept(); # Accept the client connection.
1438: Debug(8,"Connection request accepted from "
1439: .GetPeername($connection, AF_UNIX));
1440:
1441:
1442: my $description = sprintf("Connection to lonc client %d",
1443: $ClientConnection);
1444: Debug(9, "Creating event named: ".$description);
1445: Event->io(cb => \&ClientRequest,
1446: poll => 'r',
1447: desc => $description,
1448: data => "",
1449: fd => $connection);
1450: $ActiveClients{$connection} = $ClientConnection;
1451: $ClientConnection++;
1452: }
1.1 foxr 1453:
1454: =pod
1.3 albertel 1455:
1.1 foxr 1456: =head2 NewClient
1.3 albertel 1457:
1458: Callback that is called when a connection is received on the unix
1459: socket for a new client of lonc. The callback is parameterized by the
1460: event.. which is a-priori assumed to be an io event, and therefore has
1.97 ! raeburn 1461: an fd member that is the Listener socket. We accept the connection
1.3 albertel 1462: and register a new event on the readability of that socket:
1463:
1.1 foxr 1464: =cut
1.3 albertel 1465:
1.1 foxr 1466: sub NewClient {
1467: Debug(6, "NewClient");
1468: my $event = shift; # Get the event parameters.
1469: my $watcher = $event->w;
1470: my $socket = $watcher->fd; # Get the event' socket.
1471:
1.62 foxr 1472: &accept_client($socket);
1.1 foxr 1473: }
1.3 albertel 1474:
1475: =pod
1476:
1477: =head2 GetLoncSocketPath
1478:
1479: Returns the name of the UNIX socket on which to listen for client
1480: connections.
1.1 foxr 1481:
1.58 foxr 1482: =head2 Parameters:
1483:
1484: host (optional) - Name of the host socket to return.. defaults to
1485: the return from GetServerHost().
1486:
1.1 foxr 1487: =cut
1.3 albertel 1488:
1.1 foxr 1489: sub GetLoncSocketPath {
1.58 foxr 1490:
1491: my $host = GetServerHost(); # Default host.
1492: if (@_) {
1493: ($host) = @_; # Override if supplied.
1494: }
1495: return $UnixSocketDir."/".$host;
1.1 foxr 1496: }
1497:
1.3 albertel 1498: =pod
1499:
1500: =head2 GetServerHost
1501:
1502: Returns the host whose lond we talk with.
1503:
1.1 foxr 1504: =cut
1.3 albertel 1505:
1.7 foxr 1506: sub GetServerHost {
1.1 foxr 1507: return $RemoteHost; # Setup by the fork.
1508: }
1.3 albertel 1509:
1510: =pod
1511:
1.81 albertel 1512: =head2 GetServerId
1513:
1514: Returns the hostid whose lond we talk with.
1515:
1516: =cut
1517:
1518: sub GetHostId {
1519: return $RemoteHostId; # Setup by the fork.
1520: }
1521:
1522: =pod
1523:
1.3 albertel 1524: =head2 GetServerPort
1525:
1526: Returns the lond port number.
1527:
1.1 foxr 1528: =cut
1.3 albertel 1529:
1.7 foxr 1530: sub GetServerPort {
1.1 foxr 1531: return $perlvar{londPort};
1532: }
1.3 albertel 1533:
1534: =pod
1535:
1536: =head2 SetupLoncListener
1537:
1538: Setup a lonc listener event. The event is called when the socket
1539: becomes readable.. that corresponds to the receipt of a new
1540: connection. The event handler established will accept the connection
1.97 ! raeburn 1541: (creating a communcations channel), that in turn will establish
1.3 albertel 1542: another event handler to subess requests.
1.1 foxr 1543:
1.58 foxr 1544: =head2 Parameters:
1545:
1546: host (optional) Name of the host to set up a unix socket to.
1547:
1.1 foxr 1548: =cut
1.3 albertel 1549:
1.1 foxr 1550: sub SetupLoncListener {
1.78 albertel 1551: my ($host,$SocketName) = @_;
1552: if (!$host) { $host = &GetServerHost(); }
1553: if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }
1.1 foxr 1554:
1.78 albertel 1555:
1556: unlink($SocketName);
1.58 foxr 1557:
1.1 foxr 1558: my $socket;
1.7 foxr 1559: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1.55 albertel 1560: Listen => 250,
1.1 foxr 1561: Type => SOCK_STREAM)) {
1.64 foxr 1562: if($I_am_child) {
1563: &child_exit(-1, "Failed to create a lonc listener socket");
1564: } else {
1565: die "Failed to create a lonc listner socket";
1566: }
1.1 foxr 1567: }
1.59 foxr 1568: return $socket;
1.1 foxr 1569: }
1570:
1.39 foxr 1571: #
1572: # Toggle transaction logging.
1573: # Implicit inputs:
1574: # LogTransactions
1575: # Implicit Outputs:
1576: # LogTransactions
1577: sub ToggleTransactionLogging {
1578: print STDERR "Toggle transaction logging...\n";
1579: if(!$LogTransactions) {
1580: $LogTransactions = 1;
1581: } else {
1582: $LogTransactions = 0;
1583: }
1584:
1585:
1586: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1587: }
1588:
1.14 foxr 1589: =pod
1590:
1591: =head2 ChildStatus
1592:
1593: Child USR1 signal handler to report the most recent status
1594: into the status file.
1595:
1.22 foxr 1596: We also use this to reset the retries count in order to allow the
1597: client to retry connections with a previously dead server.
1.69 matthew 1598:
1.14 foxr 1599: =cut
1.46 albertel 1600:
1.14 foxr 1601: sub ChildStatus {
1602: my $event = shift;
1603: my $watcher = $event->w;
1604:
1605: Debug(2, "Reporting child status because : ".$watcher->data);
1606: my $docdir = $perlvar{'lonDocRoot'};
1.67 albertel 1607:
1608: open(LOG,">>$docdir/lon-status/loncstatus.txt");
1609: flock(LOG,LOCK_EX);
1610: print LOG $$."\t".$RemoteHost."\t".$Status."\t".
1.14 foxr 1611: $RecentLogEntry."\n";
1.38 foxr 1612: #
1613: # Write out information about each of the connections:
1614: #
1.46 albertel 1615: if ($DebugLevel > 2) {
1.67 albertel 1616: print LOG "Active connection statuses: \n";
1.46 albertel 1617: my $i = 1;
1618: print STDERR "================================= Socket Status Dump:\n";
1619: foreach my $item (keys %ActiveConnections) {
1620: my $Socket = $ActiveConnections{$item}->data;
1621: my $state = $Socket->GetState();
1.67 albertel 1622: print LOG "Connection $i State: $state\n";
1.46 albertel 1623: print STDERR "---------------------- Connection $i \n";
1.48 foxr 1624: $Socket->Dump(-1); # Ensure it gets dumped..
1.46 albertel 1625: $i++;
1626: }
1.38 foxr 1627: }
1.67 albertel 1628: flock(LOG,LOCK_UN);
1629: close(LOG);
1.22 foxr 1630: $ConnectionRetriesLeft = $ConnectionRetries;
1.70 albertel 1631: UpdateStatus();
1.14 foxr 1632: }
1633:
1.1 foxr 1634: =pod
1.3 albertel 1635:
1.10 foxr 1636: =head2 SignalledToDeath
1637:
1638: Called in response to a signal that causes a chid process to die.
1639:
1640: =cut
1641:
1642:
1643: sub SignalledToDeath {
1.14 foxr 1644: my $event = shift;
1645: my $watcher= $event->w;
1646:
1647: Debug(2,"Signalled to death! via ".$watcher->data);
1.17 foxr 1648: my ($signal) = $watcher->data;
1.10 foxr 1649: chomp($signal);
1650: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1651: ."died through "."\"$signal\"");
1.68 albertel 1652: #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1653: # ."\"$signal\"");
1.12 foxr 1654: exit 0;
1.10 foxr 1655:
1656: }
1.16 foxr 1657:
1.69 matthew 1658: =pod
1659:
1.16 foxr 1660: =head2 ToggleDebug
1661:
1662: This sub toggles trace debugging on and off.
1663:
1664: =cut
1665:
1666: sub ToggleDebug {
1667: my $Current = $DebugLevel;
1668: $DebugLevel = $NextDebugLevel;
1669: $NextDebugLevel = $Current;
1670:
1671: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1672:
1673: }
1674:
1.69 matthew 1675: =pod
1676:
1.1 foxr 1677: =head2 ChildProcess
1678:
1679: This sub implements a child process for a single lonc daemon.
1.61 foxr 1680: Optional parameter:
1.97 ! raeburn 1681: $socket - if provided, this is a socket already open for listening
! 1682: on the client socket. Otherwise, a new listener is set up.
1.1 foxr 1683:
1684: =cut
1685:
1686: sub ChildProcess {
1.80 albertel 1687: # We've inherited all the
1.62 foxr 1688: # events of our parent and those have to be cancelled or else
1689: # all holy bloody chaos will result.. trust me, I already made
1690: # >that< mistake.
1691:
1692: my $host = GetServerHost();
1693: foreach my $listener (keys %parent_dispatchers) {
1694: my $watcher = $parent_dispatchers{$listener};
1695: my $s = $watcher->fd;
1696: if ($listener ne $host) { # Close everyone but me.
1697: Debug(5, "Closing listen socket for $listener");
1698: $s->close();
1699: }
1700: Debug(5, "Killing watcher for $listener");
1701:
1702: $watcher->cancel();
1.65 foxr 1703: delete($parent_dispatchers{$listener});
1.62 foxr 1704:
1705: }
1.65 foxr 1706:
1707: # kill off the parent's signal handlers too!
1708: #
1709:
1710: for my $handler (keys %parent_handlers) {
1711: my $watcher = $parent_handlers{$handler};
1712: $watcher->cancel();
1713: delete($parent_handlers{$handler});
1714: }
1715:
1.64 foxr 1716: $I_am_child = 1; # Seems like in spite of it all I may still getting
1717: # parent event dispatches.. flag I'm a child.
1.1 foxr 1718:
1719:
1.14 foxr 1720: #
1721: # Signals must be handled by the Event framework...
1.61 foxr 1722: #
1.14 foxr 1723:
1724: Event->signal(signal => "QUIT",
1725: cb => \&SignalledToDeath,
1726: data => "QUIT");
1727: Event->signal(signal => "HUP",
1728: cb => \&ChildStatus,
1729: data => "HUP");
1730: Event->signal(signal => "USR1",
1731: cb => \&ChildStatus,
1732: data => "USR1");
1.39 foxr 1733: Event->signal(signal => "USR2",
1734: cb => \&ToggleTransactionLogging);
1.16 foxr 1735: Event->signal(signal => "INT",
1736: cb => \&ToggleDebug,
1737: data => "INT");
1.1 foxr 1738:
1.93 foxr 1739: # Block the pipe signal we'll get when the socket disconnects. We detect
1740: # socket disconnection via send/receive failures. On disconnect, the
1741: # socket becomes readable .. which will force the disconnect detection.
1742:
1743: my $set = POSIX::SigSet->new(SIGPIPE);
1744: sigprocmask(SIG_BLOCK, $set);
1745:
1.62 foxr 1746: # Figure out if we got passed a socket or need to open one to listen for
1747: # client requests.
1748:
1.61 foxr 1749: my ($socket) = @_;
1750: if (!$socket) {
1751:
1752: $socket = SetupLoncListener();
1753: }
1.62 foxr 1754: # Establish an event to listen for client connection requests.
1755:
1756:
1.59 foxr 1757: Event->io(cb => \&NewClient,
1758: poll => 'r',
1759: desc => 'Lonc Listener Unix Socket',
1760: fd => $socket);
1.1 foxr 1761:
1.76 albertel 1762: $Event::DebugLevel = $DebugLevel;
1.1 foxr 1763:
1764: Debug(9, "Making initial lond connection for ".$RemoteHost);
1765:
1766: # Setup the initial server connection:
1767:
1.62 foxr 1768: # &MakeLondConnection(); // let first work request do it.
1.10 foxr 1769:
1.80 albertel 1770: # need to accept the connection since the event may not fire.
1.62 foxr 1771:
1.80 albertel 1772: &accept_client($socket);
1.5 foxr 1773:
1.1 foxr 1774: Debug(9,"Entering event loop");
1775: my $ret = Event::loop(); # Start the main event loop.
1776:
1777:
1.64 foxr 1778: &child_exit (-1,"Main event loop exited!!!");
1.1 foxr 1779: }
1780:
1781: # Create a new child for host passed in:
1782:
1783: sub CreateChild {
1.81 albertel 1784: my ($host, $hostid) = @_;
1.52 foxr 1785:
1.12 foxr 1786: my $sigset = POSIX::SigSet->new(SIGINT);
1787: sigprocmask(SIG_BLOCK, $sigset);
1.1 foxr 1788: $RemoteHost = $host;
1.91 foxr 1789: ShowStatus('Parent keeping the flock'); # Update time in status message.
1.9 foxr 1790: Log("CRITICAL", "Forking server for ".$host);
1.23 foxr 1791: my $pid = fork;
1.1 foxr 1792: if($pid) { # Parent
1.17 foxr 1793: $RemoteHost = "Parent";
1.83 albertel 1794: $ChildPid{$pid} = $host;
1.12 foxr 1795: sigprocmask(SIG_UNBLOCK, $sigset);
1.82 albertel 1796: undef(@all_host_ids);
1.1 foxr 1797: } else { # child.
1.81 albertel 1798: $RemoteHostId = $hostid;
1.5 foxr 1799: ShowStatus("Connected to ".$RemoteHost);
1.23 foxr 1800: $SIG{INT} = 'DEFAULT';
1.12 foxr 1801: sigprocmask(SIG_UNBLOCK, $sigset);
1.81 albertel 1802: &ChildProcess(); # Does not return.
1.1 foxr 1803: }
1.61 foxr 1804: }
1.1 foxr 1805:
1.61 foxr 1806: # parent_client_connection:
1807: # Event handler that processes client connections for the parent process.
1808: # This sub is called when the parent is listening on a socket and
1809: # a connection request arrives. We must:
1810: # Start a child process to accept the connection request.
1811: # Kill our listen on the socket.
1812: # Parameter:
1813: # event - The event object that was created to monitor this socket.
1814: # event->w->fd is the socket.
1815: # Returns:
1816: # NONE
1817: #
1818: sub parent_client_connection {
1.62 foxr 1819: if ($I_am_child) {
1820: # Should not get here, but seem to anyway:
1821: &Debug(5," Child caught parent client connection event!!");
1822: my ($event) = @_;
1823: my $watcher = $event->w;
1824: $watcher->cancel(); # Try to kill it off again!!
1825: } else {
1826: &Debug(9, "parent_client_connection");
1827: my ($event) = @_;
1828: my $watcher = $event->w;
1829: my $socket = $watcher->fd;
1.81 albertel 1830: my $connection = $socket->accept(); # Accept the client connection.
1831: Event->io(cb => \&get_remote_hostname,
1832: poll => 'r',
1833: data => "",
1834: fd => $connection);
1.77 albertel 1835: }
1836: }
1837:
1838: sub get_remote_hostname {
1.82 albertel 1839: my ($event) = @_;
1840: my $watcher = $event->w;
1841: my $socket = $watcher->fd;
1842:
1843: my $thisread;
1844: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1845: Debug(8, "rcv: data length = ".length($thisread)." read =".$thisread);
1846: if (!defined($rv) || length($thisread) == 0) {
1847: # Likely eof on socket.
1848: Debug(5,"Client Socket closed on lonc for p_c_c");
1849: close($socket);
1850: $watcher->cancel();
1851: return;
1852: }
1853:
1854: my $data = $watcher->data().$thisread;
1855: $watcher->data($data);
1856: if($data =~ /\n$/) { # Request entirely read.
1857: chomp($data);
1858: } else {
1859: return;
1860: }
1.77 albertel 1861:
1.82 albertel 1862: &Debug(5,"Creating child for $data (parent_client_connection)");
1863: (my $hostname,my $lonid,@all_host_ids) = split(':',$data);
1.83 albertel 1864: $ChildHost{$hostname}++;
1865: if ($ChildHost{$hostname} == 1) {
1866: &CreateChild($hostname,$lonid);
1867: } else {
1868: &Log('WARNING',"Request for a second child on $hostname");
1869: }
1.82 albertel 1870: # Clean up the listen since now the child takes over until it exits.
1871: $watcher->cancel(); # Nolonger listening to this event
1872: $socket->send("done\n");
1873: $socket->close();
1.61 foxr 1874: }
1875:
1876: # parent_listen:
1877: # Opens a socket and starts a listen for the parent process on a client UNIX
1878: # domain socket.
1879: #
1880: # This involves:
1881: # Creating a socket for listen.
1882: # Removing any socket lock file
1883: # Adding an event handler for this socket becoming readable
1884: # To the parent's event dispatcher.
1885: # Parameters:
1886: # loncapa_host - LonCAPA cluster name of the host represented by the client
1887: # socket.
1888: # Returns:
1889: # NONE
1890: #
1891: sub parent_listen {
1892: my ($loncapa_host) = @_;
1893: Debug(5, "parent_listen: $loncapa_host");
1894:
1.78 albertel 1895: my ($socket,$file);
1896: if (!$loncapa_host) {
1897: $loncapa_host = 'common_parent';
1898: $file = $perlvar{'lonSockCreate'};
1899: } else {
1900: $file = &GetLoncSocketPath($loncapa_host);
1901: }
1902: $socket = &SetupLoncListener($loncapa_host,$file);
1903:
1.62 foxr 1904: $listening_to{$socket} = $loncapa_host;
1.61 foxr 1905: if (!$socket) {
1906: die "Unable to create a listen socket for $loncapa_host";
1907: }
1908:
1.78 albertel 1909: my $lock_file = $file.".lock";
1.61 foxr 1910: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1911:
1.77 albertel 1912: my $watcher =
1913: Event->io(cb => \&parent_client_connection,
1914: poll => 'r',
1915: desc => "Parent listener unix socket ($loncapa_host)",
1916: data => "",
1917: fd => $socket);
1.62 foxr 1918: $parent_dispatchers{$loncapa_host} = $watcher;
1.61 foxr 1919:
1920: }
1921:
1.77 albertel 1922: sub parent_clean_up {
1923: my ($loncapa_host) = @_;
1.87 albertel 1924: Debug(1, "parent_clean_up: $loncapa_host");
1.77 albertel 1925:
1926: my $socket_file = &GetLoncSocketPath($loncapa_host);
1927: unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
1928: my $lock_file = $socket_file.".lock";
1929: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1930: }
1931:
1.61 foxr 1932:
1.83 albertel 1933:
1934: # This sub initiates a listen on the common unix domain lonc client socket.
1935: # loncnew starts up with no children, and only spawns off children when a
1936: # connection request occurs on the common client unix socket. The spawned
1937: # child continues to run until it has been idle a while at which point it
1938: # eventually exits and once more the parent picks up the listen.
1.61 foxr 1939: #
1940: # Parameters:
1941: # NONE
1942: # Implicit Inputs:
1943: # The configuration file that has been read in by LondConnection.
1944: # Returns:
1945: # NONE
1946: #
1.77 albertel 1947: sub listen_on_common_socket {
1948: Debug(5, "listen_on_common_socket");
1.78 albertel 1949: &parent_listen();
1.77 albertel 1950: }
1951:
1.63 foxr 1952: # server_died is called whenever a child process exits.
1953: # Since this is dispatched via a signal, we must process all
1954: # dead children until there are no more left. The action
1955: # is to:
1956: # - Remove the child from the bookeeping hashes
1957: # - Re-establish a listen on the unix domain socket associated
1958: # with that host.
1959: # Parameters:
1960: # The event, but we don't actually care about it.
1961: sub server_died {
1962: &Debug(9, "server_died called...");
1963:
1964: while(1) { # Loop until waitpid nowait fails.
1965: my $pid = waitpid(-1, WNOHANG);
1966: if($pid <= 0) {
1967: return; # Nothing left to wait for.
1968: }
1969: # need the host to restart:
1970:
1.83 albertel 1971: my $host = $ChildPid{$pid};
1.63 foxr 1972: if($host) { # It's for real...
1973: &Debug(9, "Caught sigchild for $host");
1.83 albertel 1974: delete($ChildPid{$pid});
1975: delete($ChildHost{$host});
1.81 albertel 1976: &parent_clean_up($host);
1977:
1.63 foxr 1978: } else {
1979: &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
1980: }
1981: }
1982:
1983: }
1984:
1.1 foxr 1985: #
1986: # Parent process logic pass 1:
1987: # For each entry in the hosts table, we will
1988: # fork off an instance of ChildProcess to service the transactions
1989: # to that host. Each pid will be entered in a global hash
1990: # with the value of the key, the host.
1991: # The parent will then enter a loop to wait for process exits.
1992: # Each exit gets logged and the child gets restarted.
1993: #
1994:
1.5 foxr 1995: #
1996: # Fork and start in new session so hang-up isn't going to
1997: # happen without intent.
1998: #
1999:
2000:
1.6 foxr 2001:
2002:
1.8 foxr 2003:
1.6 foxr 2004:
2005: ShowStatus("Forming new session");
2006: my $childpid = fork;
2007: if ($childpid != 0) {
2008: sleep 4; # Give child a chacne to break to
2009: exit 0; # a new sesion.
2010: }
1.8 foxr 2011: #
2012: # Write my pid into the pid file so I can be located
2013: #
2014:
2015: ShowStatus("Parent writing pid file:");
1.23 foxr 2016: my $execdir = $perlvar{'lonDaemons'};
1.8 foxr 2017: open (PIDSAVE, ">$execdir/logs/lonc.pid");
2018: print PIDSAVE "$$\n";
2019: close(PIDSAVE);
1.6 foxr 2020:
1.17 foxr 2021:
2022:
1.6 foxr 2023: if (POSIX::setsid() < 0) {
2024: print "Could not create new session\n";
2025: exit -1;
2026: }
1.5 foxr 2027:
2028: ShowStatus("Forking node servers");
2029:
1.9 foxr 2030: Log("CRITICAL", "--------------- Starting children ---------------");
2031:
1.31 foxr 2032: LondConnection::ReadConfig; # Read standard config files.
1.1 foxr 2033:
1.80 albertel 2034: $RemoteHost = "[parent]";
1.81 albertel 2035: &listen_on_common_socket();
1.60 foxr 2036:
1.12 foxr 2037: $RemoteHost = "Parent Server";
1.1 foxr 2038:
2039: # Maintain the population:
1.5 foxr 2040:
2041: ShowStatus("Parent keeping the flock");
1.1 foxr 2042:
1.12 foxr 2043:
1.80 albertel 2044: # We need to setup a SIGChild event to handle the exit (natural or otherwise)
2045: # of the children.
1.61 foxr 2046:
1.80 albertel 2047: Event->signal(cb => \&server_died,
2048: desc => "Child exit handler",
2049: signal => "CHLD");
2050:
2051:
2052: # Set up all the other signals we set up.
2053:
2054: $parent_handlers{INT} = Event->signal(cb => \&Terminate,
2055: desc => "Parent INT handler",
2056: signal => "INT");
2057: $parent_handlers{TERM} = Event->signal(cb => \&Terminate,
2058: desc => "Parent TERM handler",
2059: signal => "TERM");
1.81 albertel 2060: $parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
2061: desc => "Parent HUP handler.",
2062: signal => "HUP");
1.80 albertel 2063: $parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
2064: desc => "Parent USR1 handler",
2065: signal => "USR1");
2066: $parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
2067: desc => "Parent USR2 handler.",
2068: signal => "USR2");
2069:
2070: # Start procdesing events.
2071:
2072: $Event::DebugLevel = $DebugLevel;
2073: Debug(9, "Parent entering event loop");
2074: my $ret = Event::loop();
2075: die "Main Event loop exited: $ret";
1.14 foxr 2076:
2077: =pod
2078:
2079: =head1 CheckKids
2080:
2081: Since kids do not die as easily in this implementation
1.97 ! raeburn 2082: as the previous one, there is no need to restart the
1.14 foxr 2083: dead ones (all dead kids get restarted when they die!!)
2084: The only thing this function does is to pass USR1 to the
2085: kids so that they report their status.
2086:
2087: =cut
2088:
2089: sub CheckKids {
2090: Debug(2, "Checking status of children");
2091: my $docdir = $perlvar{'lonDocRoot'};
2092: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
2093: my $now=time;
2094: my $local=localtime($now);
2095: print $fh "LONC status $local - parent $$ \n\n";
1.65 foxr 2096: foreach my $host (keys %parent_dispatchers) {
2097: print $fh "LONC Parent process listening for $host\n";
2098: }
1.83 albertel 2099: foreach my $pid (keys %ChildPid) {
1.14 foxr 2100: Debug(2, "Sending USR1 -> $pid");
2101: kill 'USR1' => $pid; # Tell Child to report status.
2102: }
1.65 foxr 2103:
1.14 foxr 2104: }
1.24 foxr 2105:
2106: =pod
2107:
2108: =head1 UpdateKids
2109:
1.25 foxr 2110: parent's SIGUSR2 handler. This handler:
1.24 foxr 2111:
2112: =item
2113:
2114: Rereads the hosts file.
2115:
2116: =item
2117:
2118: Kills off (via sigint) children for hosts that have disappeared.
2119:
2120: =item
2121:
1.27 foxr 2122: QUITs children for hosts that already exist (this just forces a status display
1.24 foxr 2123: and resets the connection retry count for that host.
2124:
2125: =item
2126:
2127: Starts new children for hosts that have been added to the hosts.tab file since
2128: the start of the master program and maintains them.
2129:
2130: =cut
2131:
2132: sub UpdateKids {
1.27 foxr 2133:
1.25 foxr 2134: Log("INFO", "Updating connections via SIGUSR2");
1.27 foxr 2135:
1.65 foxr 2136: # I'm not sure what I was thinking in the first implementation.
2137: # someone will have to work hard to convince me the effect is any
2138: # different than Restart, especially now that we don't start up
2139: # per host servers automatically, may as well just restart.
2140: # The down side is transactions that are in flight will get timed out
2141: # (lost unless they are critical).
1.27 foxr 2142:
1.81 albertel 2143: &KillThemAll();
1.24 foxr 2144: }
2145:
1.14 foxr 2146:
1.13 foxr 2147: =pod
2148:
2149: =head1 Restart
2150:
2151: Signal handler for HUP... all children are killed and
1.97 ! raeburn 2152: we self restart. This is an el-cheapo way to re-read
1.13 foxr 2153: the config file.
2154:
2155: =cut
2156:
2157: sub Restart {
1.23 foxr 2158: &KillThemAll; # First kill all the children.
1.13 foxr 2159: Log("CRITICAL", "Restarting");
2160: my $execdir = $perlvar{'lonDaemons'};
2161: unlink("$execdir/logs/lonc.pid");
1.65 foxr 2162: exec("$executable");
1.10 foxr 2163: }
1.12 foxr 2164:
2165: =pod
2166:
2167: =head1 KillThemAll
2168:
2169: Signal handler that kills all children by sending them a
1.17 foxr 2170: SIGHUP. Responds to sigint and sigterm.
1.12 foxr 2171:
2172: =cut
2173:
1.10 foxr 2174: sub KillThemAll {
1.12 foxr 2175: Debug(2, "Kill them all!!");
1.85 albertel 2176:
2177: #local($SIG{CHLD}) = 'IGNORE';
2178: # Our children >will< die.
2179: # but we need to catch their death and cleanup after them in case this is
2180: # a restart set of kills
2181: my @allpids = keys(%ChildPid);
2182: foreach my $pid (@allpids) {
1.83 albertel 2183: my $serving = $ChildPid{$pid};
1.52 foxr 2184: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2185: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
1.17 foxr 2186: kill 'QUIT' => $pid;
1.12 foxr 2187: }
1.85 albertel 2188: ShowStatus("Finished killing child processes off.");
1.1 foxr 2189: }
1.12 foxr 2190:
1.52 foxr 2191:
2192: #
2193: # Kill all children via KILL. Just in case the
2194: # first shot didn't get them.
2195:
2196: sub really_kill_them_all_dammit
2197: {
2198: Debug(2, "Kill them all Dammit");
2199: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
1.83 albertel 2200: foreach my $pid (keys %ChildPid) {
2201: my $serving = $ChildPid{$pid};
1.52 foxr 2202: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2203: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2204: kill 'KILL' => $pid;
1.83 albertel 2205: delete($ChildPid{$pid});
1.52 foxr 2206: my $execdir = $perlvar{'lonDaemons'};
2207: unlink("$execdir/logs/lonc.pid");
2208: }
2209: }
1.69 matthew 2210:
1.14 foxr 2211: =pod
2212:
2213: =head1 Terminate
2214:
2215: Terminate the system.
2216:
2217: =cut
2218:
2219: sub Terminate {
1.52 foxr 2220: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2221: &KillThemAll;
2222: #
2223: # By now they really should all be dead.. but just in case
2224: # send them all SIGKILL's after a bit of waiting:
2225:
2226: sleep(4);
2227: &Log("CRITICAL", "Now kill children nasty");
2228: &really_kill_them_all_dammit;
1.17 foxr 2229: Log("CRITICAL","Master process exiting");
2230: exit 0;
1.14 foxr 2231:
2232: }
1.81 albertel 2233:
2234: sub my_hostname {
2235: use Sys::Hostname;
2236: my $name = &hostname();
2237: &Debug(9,"Name is $name");
2238: return $name;
2239: }
2240:
1.12 foxr 2241: =pod
1.1 foxr 2242:
2243: =head1 Theory
1.3 albertel 2244:
2245: The event class is used to build this as a single process with an
2246: event driven model. The following events are handled:
1.1 foxr 2247:
2248: =item UNIX Socket connection Received
2249:
2250: =item Request data arrives on UNIX data transfer socket.
2251:
2252: =item lond connection becomes writable.
2253:
2254: =item timer fires at 1 second intervals.
2255:
2256: All sockets are run in non-blocking mode. Timeouts managed by the timer
2257: handler prevents hung connections.
2258:
2259: Key data structures:
2260:
1.3 albertel 2261: =item RequestQueue
2262:
2263: A queue of requests received from UNIX sockets that are
2264: waiting for a chance to be forwarded on a lond connection socket.
2265:
2266: =item ActiveConnections
2267:
2268: A hash of lond connections that have transactions in process that are
2269: available to be timed out.
2270:
2271: =item ActiveTransactions
2272:
2273: A hash indexed by lond connections that contain the client reply
2274: socket for each connection that has an active transaction on it.
2275:
2276: =item IdleConnections
2277:
2278: A hash of lond connections that have no work to do. These connections
2279: can be closed if they are idle for a long enough time.
1.1 foxr 2280:
2281: =cut
1.88 foxr 2282:
2283: =pod
2284:
2285: =head1 Log messages
2286:
2287: The following is a list of log messages that can appear in the
2288: lonc.log file. Each log file has a severity and a message.
2289:
2290: =over 2
2291:
2292: =item Warning A socket timeout was detected
2293:
2294: If there are pending transactions in the socket's queue,
2295: they are failed (saved if critical). If the connection
2296: retry count gets exceeded by this, the
2297: remote host is marked as dead.
1.97 ! raeburn 2298: Called when timeouts occurred during the connection and
1.88 foxr 2299: connection dialog with a remote host.
2300:
2301: =item Critical Host makred DEAD <hostname>
2302:
2303: The numer of retry counts for contacting a host was
2304: exceeded. The host is marked dead an no
2305: further attempts will be made by that child.
2306:
2307: =item Info lonc pipe client hung up on us
2308:
2309: Write to the client pipe indicated no data transferred
2310: Socket to remote host is shut down. Reply to the client
2311: is discarded. Note: This is commented out in &ClientWriteable
2312:
2313: =item Success Reply from lond: <data>
2314:
2315: Can be enabled for debugging by setting LogTransactions to nonzero.
2316: Indicates a successful transaction with lond, <data> is the data received
2317: from the remote lond.
2318:
2319: =item Success A delayed transaction was completed
2320:
2321: A transaction that must be reliable was executed and completed
2322: as lonc restarted. This is followed by a mesage of the form
2323:
2324: S: client-name : request
2325:
2326: =item WARNING Failing transaction <cmd>:<subcmd>
2327:
2328: Transaction failed on a socket, but the failure retry count for the remote
2329: node has not yet been exhausted (the node is not yet marked dead).
2330: cmd is the command, subcmd is the subcommand. This results from a con_lost
2331: when communicating with lond.
2332:
2333: =item WARNING Shutting down a socket
2334:
2335: Called when a socket is being closed to lond. This is emitted both when
2336: idle pruning is being done and when the socket has been disconnected by the remote.
2337:
2338: =item WARNING Lond connection lost.
2339:
2340: Called when a read from lond's socket failed indicating lond has closed the
2341: connection or died. This should be followed by one or more
2342:
2343: "WARNING Failing transaction..." msgs for each in-flight or queued transaction.
2344:
2345: =item INFO Connected to lond version: <version>
2346:
2347: When connection negotiation is complete, the lond version is requested and logged here.
2348:
2349: =item SUCCESS Connection n to host now ready for action
2350:
2351: Emitted when connection has been completed with lond. n is then number of
2352: concurrent connections and host, the host to which the connection has just
2353: been established.
2354:
2355: =item WARNING Connection to host has been disconnected
2356:
2357: Write to a lond resulted in failure status. Connection to lond is dropped.
2358:
2359: =item SUCCESS Created connection n to host host
2360:
2361: Initial connection request to host..(before negotiation).
2362:
2363: =item CRITICAL Request Close Connection ... exiting
2364:
2365: Client has sent "close_connection_exit" The loncnew server is exiting.
2366:
2367: =item INFO Resetting Connection Retries
2368:
2369: Client has sent "reset_retries" The lond connection retries are reset to zero for the
2370: corresponding lond.
2371:
2372: =item SUCCESS Transaction <data>
2373:
2374: Only emitted if the global variable $LogTransactions was set to true.
2375: A client has requested a lond transaction <data> is the contents of the request.
2376:
2377: =item SUCCESS Toggled transaction logging <LogTransactions>
2378:
2379: The state of the $LogTransactions global has been toggled, and its current value
2380: (after being toggled) is displayed. When non zero additional logging of transactions
2381: is enabled for debugging purposes. Transaction logging is toggled on receipt of a USR2
2382: signal.
2383:
2384: =item CRITICAL Abnormal exit. Child <pid> for <host> died thorugh signal.
2385:
2386: QUIT signal received. lonc child process is exiting.
2387:
2388: =item SUCCESS New debugging level for <RemoteHost> now <DebugLevel>
2389:
2390: Debugging toggled for the host loncnew is talking with.
2391: Currently debugging is a level based scheme with higher number
2392: conveying more information. The daemon starts out at
2393: DebugLevel 0 and can toggle back and forth between that and
2394: DebugLevel 2 These are controlled by
2395: the global variables $DebugLevel and $NextDebugLevel
2396: The debug level can go up to 9.
2397: SIGINT toggles the debug level. The higher the debug level the
2398: more debugging information is spewed. See the Debug
2399: sub in loncnew.
2400:
2401: =item CRITICAL Forking server for host
2402:
2403: A child is being created to service requests for the specified host.
2404:
2405:
2406: =item WARNING Request for a second child on hostname
2407:
2408: Somehow loncnew was asked to start a second child on a host that already had a child
2409: servicing it. This request is not honored, but themessage is emitted. This could happen
2410: due to a race condition. When a client attempts to contact loncnew for a new host, a child
2411: is forked off to handle the requests for that server. The parent then backs off the Unix
2412: domain socket leaving it for the child to service all requests. If in the time between
2413: creating the child, and backing off, a new connection request comes in to the unix domain
2414: socket, this could trigger (unlikely but remotely possible),.
2415:
2416: =item CRITICAL ------ Starting Children ----
2417:
2418: This message should probably be changed to "Entering event loop" as the loncnew only starts
2419: children as needed. This message is emitted as new events are established and
2420: the event processing loop is entered.
2421:
2422: =item INFO Updating connections via SIGUSR2
2423:
2424: SIGUSR2 received. The original code would kill all clients, re-read the host file,
1.97 ! raeburn 2425: then restart children for each host. Now that children are started on demand, this
1.88 foxr 2426: just kills all child processes and lets requests start them as needed again.
2427:
2428:
2429: =item CRITICAL Restarting
2430:
2431: SigHUP received. all the children are killed and the script exec's itself to start again.
2432:
2433: =item CRITICAL Nicely killing lonc for host pid = <pid>
2434:
2435: Attempting to kill the child that is serving the specified host (pid given) cleanly via
1.97 ! raeburn 2436: SIGQUIT. The child should handle that, clean up nicely and exit.
1.88 foxr 2437:
2438: =item CRITICAL Nastily killing lonc for host pid = <pid>
2439:
2440: The child specified did not die when requested via SIGQUIT. Therefore it is killed
2441: via SIGKILL.
2442:
2443: =item CRITICAL Asked to kill children.. first be nice..
2444:
2445: In the parent's INT handler. INT kills the child processes. This inidicate loncnew
2446: is about to attempt to kill all known children via SIGQUIT. This message should be followed
2447: by one "Nicely killing" message for each extant child.
2448:
2449: =item CRITICAL Now kill children nasty
2450:
2451: In the parent's INT handler. remaining children are about to be killed via
2452: SIGKILL. Should be followed by a Nastily killing... for each lonc child that
2453: refused to die.
2454:
2455: =item CRITICAL Master process exiting
2456:
2457: In the parent's INT handler. just prior to the exit 0 call.
2458:
2459: =back
2460:
2461: =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