Diff for /loncom/loncnew between versions 1.2 and 1.31

version 1.2, 2003/04/18 02:59:31 version 1.31, 2003/10/28 10:48:10
Line 7 Line 7
 # Copyright Michigan State University Board of Trustees  # Copyright Michigan State University Board of Trustees
 #  #
 # This file is part of the LearningOnline Network with CAPA (LON-CAPA).  # This file is part of the LearningOnline Network with CAPA (LON-CAPA).
 #  ## LON-CAPA is free software; you can redistribute it and/or modify
 # LON-CAPA is free software; you can redistribute it and/or modify  
 # it under the terms of the GNU General Public License as published by  # it under the terms of the GNU General Public License as published by
 # the Free Software Foundation; either version 2 of the License, or  # the Free Software Foundation; either version 2 of the License, or
 # (at your option) any later version.  # (at your option) any later version.
Line 27 Line 26
 # http://www.lon-capa.org/  # http://www.lon-capa.org/
 #  #
 #  #
 # new lonc handles n requestors spread out bver m connections to londs.  # new lonc handles n request out bver m connections to londs.
 # This module is based on the Event class.  # This module is based on the Event class.
 #   Development iterations:  #   Development iterations:
 #    - Setup basic event loop.   (done)  #    - Setup basic event loop.   (done)
 #    - Add timer dispatch.       (done)  #    - Add timer dispatch.       (done)
 #    - Add ability to accept lonc UNIX domain sockets.  (done)  #    - Add ability to accept lonc UNIX domain sockets.  (done)
 #    - Add ability to create/negotiate lond connections (done).  #    - Add ability to create/negotiate lond connections (done).
 #    - Add general logic for dispatching requests and timeouts.  #    - Add general logic for dispatching requests and timeouts. (done).
 #    - Add support for the lonc/lond requests.  #    - Add support for the lonc/lond requests.          (done).
 #    - Add logging/status monitoring.  #    - Add logging/status monitoring.
 #    - Add Signal handling - HUP restarts. USR1 status report.  #    - Add Signal handling - HUP restarts. USR1 status report.
 #    - Add Configuration file I/O  #    - Add Configuration file I/O                       (done).
 #    - Add Pending request processing on startup.  
 #    - Add management/status request interface.  #    - Add management/status request interface.
   #    - Add deferred request capability.                  (done)
   #    - Detect transmission timeouts.
   #
   
   # Change log:
   #    $Log$
   #    Revision 1.31  2003/10/28 10:48:10  foxr
   #    Call LondConnection::ReadConfig explicitly so that the configuration is loaded
   #    prior to getting the hosts table hashiterator.
   #
   #    Revision 1.30  2003/10/27 10:09:21  foxr
   #    Tighten up a few compares to eq and flip a few debug levels around... nothing
   #    critical
   #
   #    Revision 1.29  2003/10/21 14:24:42  foxr
   #    Fix little typo that may explain growth of connections
   #
   #    Revision 1.28  2003/10/14 15:36:21  albertel
   #    - making it easier to run loncnew,
   #       /etc/init.d/loncontrol startnew
   #       /etc/init.d/loncontrol restartnew
   #      will now start loncnew in place of lonc
   #
   #    Revision 1.27  2003/10/07 11:23:03  foxr
   #    Installed and tested code to process reinit in parent server.
   #
   #    Revision 1.26  2003/09/30 11:11:17  foxr
   #    Add book-keeping hashes to support the re-init procedure.
   #
   #    Revision 1.25  2003/09/23 11:22:14  foxr
   #    Tested ability to receive sigusr2  This is now logged and must be
   #    properly implemented as a re-read of hosts and re-init of appropriate
   #    children.
   #
   #    Revision 1.24  2003/09/16 09:46:42  foxr
   #    Added skeletal infrastructure to support SIGUSR2 update hosts request.
   #
   #    Revision 1.23  2003/09/15 09:24:49  foxr
   #    Add use strict and fix all the fallout from that.
   #
   #    Revision 1.22  2003/09/02 10:34:47  foxr
   #    - Fix errors in host dead detection logic (too many cases where the
   #      retries left were not getting incremented or just not checked).
   #    - Added some additional status to the ps axuww display:
   #      o Remaining retries on a host.
   #      o >>> DEAD <<< indicator if I've given up on a host.
   #    - Tested the SIGHUP will reset the retries remaining count (thanks to
   #      the above status stuff, and get allow the loncnew to re-try again
   #      on the host (thanks to the log).
   #
   #    Revision 1.21  2003/08/26 09:19:51  foxr
   #    How embarrassing... put in the SocketTimeout function in loncnew and forgot
   #    to actually hook it into the LondTransaction.  Added this to MakeLondConnection
   #    where it belongs... hopefully transactions (not just connection attempts) will
   #    timeout more speedily than the socket errors will catch it.
   #
   #    Revision 1.20  2003/08/25 18:48:11  albertel
   #    - fixing a forgotten ;
   #
   #    Revision 1.19  2003/08/19 09:31:46  foxr
   #    Get socket directory from configuration rather than the old hard coded test
   #    way that I forgot to un-hard code.
   #
   #    Revision 1.18  2003/08/06 09:52:29  foxr
   #    Also needed to remember to fail in-flight transactions if their sends fail.
   #
   #    Revision 1.17  2003/08/03 00:44:31  foxr
   #    1. Correct handling of connection failure: Assume it means the host is
   #       unreachable and fail all of the queued transactions.  Note that the
   #       inflight transactions should fail on their own time due either to timeout
   #       or send/receive failures.
   #    2. Correct handling of logs for forced death signals.  Pull the signal
   #       from the event watcher.
   #
   #    Revision 1.16  2003/07/29 02:33:05  foxr
   #    Add SIGINT processing to child processes to toggle annoying trace mode
   #    on/off.. will try to use this to isolate the compute boud process issue.
   #
   #    Revision 1.15  2003/07/15 02:07:05  foxr
   #    Added code for lonc/lond transaction timeouts.  Who knows if it works right.
   #    The intent is for a timeout to fail any transaction in progress and kill
   #    off the sockt that timed out.
   #
   #    Revision 1.14  2003/07/03 02:10:18  foxr
   #    Get all of the signals to work correctly.
   #
   #    Revision 1.13  2003/07/02 01:31:55  foxr
   #    Added kill -HUP logic (restart).
   #
   #    Revision 1.11  2003/06/25 01:54:44  foxr
   #    Fix more problems with transaction failure.
   #
   #    Revision 1.10  2003/06/24 02:46:04  foxr
   #    Put a limit on  the number of times we'll retry a connection.
   #    Start getting the signal stuff put in as well...note that need to get signals
   #    going or else the client will permanently give up on dead servers.
   #
   #    Revision 1.9  2003/06/13 02:38:43  foxr
   #    Add logging in 'expected format'
   #
   #    Revision 1.8  2003/06/11 02:04:35  foxr
   #    Support delayed transactions... this is done uniformly by encapsulating
   #    transactions in an object ... a LondTransaction that is implemented by
   #    LondTransaction.pm
   #
   #    Revision 1.7  2003/06/03 01:59:39  foxr
   #    complete coding to support deferred transactions.
   #
   #
   use strict;
 use lib "/home/httpd/lib/perl/";  use lib "/home/httpd/lib/perl/";
 use lib "/home/foxr/newloncapa/types";  use lib "/home/foxr/newloncapa/types";
 use Event qw(:DEFAULT );  use Event qw(:DEFAULT );
 use POSIX qw(:signal_h);  use POSIX qw(:signal_h);
   use POSIX;
 use IO::Socket;  use IO::Socket;
 use IO::Socket::INET;  use IO::Socket::INET;
 use IO::Socket::UNIX;  use IO::Socket::UNIX;
   use IO::File;
   use IO::Handle;
 use Socket;  use Socket;
 use Crypt::IDEA;  use Crypt::IDEA;
 use LONCAPA::Queue;  use LONCAPA::Queue;
 use LONCAPA::Stack;  use LONCAPA::Stack;
 use LONCAPA::LondConnection;  use LONCAPA::LondConnection;
   use LONCAPA::LondTransaction;
 use LONCAPA::Configuration;  use LONCAPA::Configuration;
 use LONCAPA::HashIterator;  use LONCAPA::HashIterator;
   
 print "Loncnew starting\n";  
   
 #  #
 #   Disable all signals we might receive from outside for now.  #   Disable all signals we might receive from outside for now.
 #  #
 $SIG{QUIT}  = IGNORE;  
 $SIG{HUP}   = IGNORE;  
 $SIG{USR1}  = IGNORE;  
 $SIG{INT}   = IGNORE;  
 $SIG{CHLD}  = IGNORE;  
 $SIG{__DIE__}  = IGNORE;  
   
   
 # Read the httpd configuration file to get perl variables  # Read the httpd configuration file to get perl variables
Line 80  my %perlvar    = %{$perlvarref}; Line 184  my %perlvar    = %{$perlvarref};
 #  parent and shared variables.  #  parent and shared variables.
   
 my %ChildHash; # by pid -> host.  my %ChildHash; # by pid -> host.
   my %HostToPid; # By host -> pid.
   my %HostHash; # by loncapaname -> IP.
   
   
 my $MaxConnectionCount = 5; # Will get from config later.  my $MaxConnectionCount = 10; # Will get from config later.
 my $ClientConnection = 0; # Uniquifier for client events.  my $ClientConnection = 0; # Uniquifier for client events.
   
 my $DebugLevel = 5;  my $DebugLevel = 0;
   my $NextDebugLevel= 2; # So Sigint can toggle this.
 my $IdleTimeout= 3600; # Wait an hour before pruning connections.  my $IdleTimeout= 3600; # Wait an hour before pruning connections.
   
 #  #
 #  The variables below are only used by the child processes.  #  The variables below are only used by the child processes.
 #  #
 my $RemoteHost; # Name of host child is talking to.  my $RemoteHost; # Name of host child is talking to.
 my $UnixSocketDir= "/home/httpd/sockets";   my $UnixSocketDir= $perlvar{'lonSockDir'};
 my $IdleConnections = Stack->new(); # Set of idle connections  my $IdleConnections = Stack->new(); # Set of idle connections
 my %ActiveConnections; # Connections to the remote lond.  my %ActiveConnections; # Connections to the remote lond.
 my %ActiveTransactions; # Transactions in flight.  my %ActiveTransactions; # LondTransactions in flight.
 my %ActiveClients; # Serial numbers of active clients by socket.  my %ActiveClients; # Serial numbers of active clients by socket.
 my $WorkQueue       = Queue->new(); # Queue of pending transactions.  my $WorkQueue       = Queue->new(); # Queue of pending transactions.
 my $ClientQueue     = Queue->new(); # Queue of clients causing xactinos.  
 my $ConnectionCount = 0;  my $ConnectionCount = 0;
   my $IdleSeconds     = 0; # Number of seconds idle.
   my $Status          = ""; # Current status string.
   my $RecentLogEntry  = "";
   my $ConnectionRetries=2; # Number of connection retries allowed.
   my $ConnectionRetriesLeft=2; # Number of connection retries remaining.
   
 #  #
   #   The hash below gives the HTML format for log messages
   #   given a severity.
   #    
   my %LogFormats;
   
   $LogFormats{"CRITICAL"} = "<font color=red>CRITICAL: %s</font>";
   $LogFormats{"SUCCESS"}  = "<font color=green>SUCCESS: %s</font>";
   $LogFormats{"INFO"}     = "<font color=yellow>INFO: %s</font>";
   $LogFormats{"WARNING"}  = "<font color=blue>WARNING: %s</font>";
   $LogFormats{"DEFAULT"}  = " %s ";
   
   
   
 =pod  =pod
 =head 2 GetPeerName  
    Returns the name of the host that a socket object is connected  =head2 LogPerm
    to.  
   Makes an entry into the permanent log file.
   
   =cut
   sub LogPerm {
       my $message=shift;
       my $execdir=$perlvar{'lonDaemons'};
       my $now=time;
       my $local=localtime($now);
       my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
       print $fh "$now:$message:$local\n";
   }
   
   =pod
   
   =head2 Log
   
   Logs a message to the log file.
   Parameters:
   
   =item severity
   
   One of CRITICAL, WARNING, INFO, SUCCESS used to select the
   format string used to format the message.  if the severity is
   not a defined severity the Default format string is used.
   
   =item message
   
   The base message.  In addtion to the format string, the message
   will be appended to a string containing the name of our remote
   host and the time will be formatted into the message.
   
   =cut
   
   sub Log {
       my $severity = shift;
       my $message  = shift;
      
       if(!$LogFormats{$severity}) {
    $severity = "DEFAULT";
       }
   
       my $format = $LogFormats{$severity};
       
       #  Put the window dressing in in front of the message format:
   
       my $now   = time;
       my $local = localtime($now);
       my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
       my $finalformat = $finalformat.$format."\n";
   
       # open the file and put the result.
   
       my $execdir = $perlvar{'lonDaemons'};
       my $fh      = IO::File->new(">>$execdir/logs/lonc.log");
       my $msg = sprintf($finalformat, $message);
       $RecentLogEntry = $msg;
       print $fh $msg;
       
       
   }
   
   
   =pod
   
   =head2 GetPeerName
   
   Returns the name of the host that a socket object is connected to.
   
 =cut  =cut
   
 sub GetPeername {  sub GetPeername {
Line 117  sub GetPeername { Line 308  sub GetPeername {
     my $peerip;      my $peerip;
     if($AdrFamily == AF_INET) {      if($AdrFamily == AF_INET) {
  ($peerport, $peerip) = sockaddr_in($peer);   ($peerport, $peerip) = sockaddr_in($peer);
  my $peername    = gethostbyaddr($iaddr, $AdrFamily);   my $peername    = gethostbyaddr($peerip, $AdrFamily);
  return $peername;   return $peername;
     } elsif ($AdrFamily == AF_UNIX) {      } elsif ($AdrFamily == AF_UNIX) {
  my $peerfile;   my $peerfile;
Line 127  sub GetPeername { Line 318  sub GetPeername {
 }  }
 #----------------------------- Timer management ------------------------  #----------------------------- Timer management ------------------------
 =pod  =pod
   
 =head2 Debug  =head2 Debug
    Invoked to issue a debug message.  
   Invoked to issue a debug message.
   
 =cut  =cut
   
 sub Debug {  sub Debug {
     my $level   = shift;      my $level   = shift;
     my $message = shift;      my $message = shift;
     if ($level <= $DebugLevel) {      if ($level <= $DebugLevel) {
  print $message." host = ".$RemoteHost."\n";   Log("INFO", "-Debug- $message host = $RemoteHost");
     }      }
 }  }
   
Line 145  sub SocketDump { Line 340  sub SocketDump {
  $socket->Dump();   $socket->Dump();
     }      }
 }  }
   
   =pod
   
   =head2 ShowStatus
   
    Place some text as our pid status.
    and as what we return in a SIGUSR1
   
   =cut
   sub ShowStatus {
       my $state = shift;
       my $now = time;
       my $local = localtime($now);
       $Status   = $local.": ".$state;
       $0='lonc: '.$state.' '.$local;
   }
   
 =pod  =pod
   
   =head 2 SocketTimeout
   
       Called when an action on the socket times out.  The socket is 
      destroyed and any active transaction is failed.
   
   
   =cut
   sub SocketTimeout {
       my $Socket = shift;
       
       KillSocket($Socket); # A transaction timeout also counts as
                                   # a connection failure:
       $ConnectionRetriesLeft--;
   }
   
   =pod
   
 =head2 Tick  =head2 Tick
     Invoked  each timer tick.  
   Invoked  each timer tick.
   
 =cut  =cut
   
   
 sub Tick {  sub Tick {
     my $client;      my $client;
     Debug(6, "Tick");      if($ConnectionRetriesLeft > 0) {
     Debug(6, "    Current connection count: ".$ConnectionCount);   ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
     foreach $client (keys %ActiveClients) {     ." Retries remaining: ".$ConnectionRetriesLeft);
  Debug(7, "    Have client:  with id: ".$ActiveClients{$client});      } else {
    ShowStatus(GetServerHost()." >> DEAD <<");
       }
       # Is it time to prune connection count:
   
   
       if($IdleConnections->Count()  && 
          ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
    $IdleSeconds++;
    if($IdleSeconds > $IdleTimeout) { # Prune a connection...
       my $Socket = $IdleConnections->pop();
       KillSocket($Socket);
    }
       } else {
    $IdleSeconds = 0; # Reset idle count if not idle.
       }
       #
       #  For each inflight transaction, tick down its timeout counter.
       #
       foreach my $item (keys %ActiveTransactions) {
    my $Socket = $ActiveTransactions{$item}->getServer();
    $Socket->Tick();
       }
       # Do we have work in the queue, but no connections to service them?
       # If so, try to make some new connections to get things going again.
       #
       
       my $Requests = $WorkQueue->Count();
       if (($ConnectionCount == 0)  && ($Requests > 0)) { 
    if ($ConnectionRetriesLeft > 0) {
       my $Connections = ($Requests <= $MaxConnectionCount) ?
    $Requests : $MaxConnectionCount;
       Debug(5,"Work but no connections, start ".$Connections." of them");
       my $successCount = 0;
       for (my $i =0; $i < $Connections; $i++) {
    $successCount += MakeLondConnection();
       }
       if($successCount == 0) { # All connections failed:
    Debug(5,"Work in queue failed to make any connectiouns\n");
    EmptyQueue(); # Fail pending transactions with con_lost.
       }
    } else {
       ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
       Debug(5,"Work in queue, but gave up on connections..flushing\n");
       EmptyQueue(); # Connections can't be established.
    }
          
     }      }
 }  }
   
 =pod  =pod
   
 =head2 SetupTimer  =head2 SetupTimer
    Sets up a 1 per sec recurring timer event.  The event handler is used to:  
   
 =item Trigger timeouts on communications along active sockets.  Sets up a 1 per sec recurring timer event.  The event handler is used to:
 =item Trigger disconnections of idle sockets.  
   =item
   
   Trigger timeouts on communications along active sockets.
   
   =item
   
   Trigger disconnections of idle sockets.
   
 =cut  =cut
   
Line 173  sub SetupTimer { Line 458  sub SetupTimer {
     Debug(6, "SetupTimer");      Debug(6, "SetupTimer");
     Event->timer(interval => 1, debug => 1, cb => \&Tick );      Event->timer(interval => 1, debug => 1, cb => \&Tick );
 }  }
   
 =pod  =pod
   
 =head2 ServerToIdle  =head2 ServerToIdle
    This function is called when a connection to the server is  
 ready for more work.   This function is called when a connection to the server is
    If there is work in the Work queue the top element is dequeued  ready for more work.
   
   If there is work in the Work queue the top element is dequeued
 and the connection will start to work on it.  If the work queue is  and the connection will start to work on it.  If the work queue is
 empty, the connection is pushed on the idle connection stack where  empty, the connection is pushed on the idle connection stack where
 it will either get another work unit, or alternatively, if it sits there  it will either get another work unit, or alternatively, if it sits there
 long enough, it will be shut down and released.  long enough, it will be shut down and released.
   
   
 =cut  =cut
   
 sub ServerToIdle {  sub ServerToIdle {
     my $Socket   = shift; # Get the socket.      my $Socket   = shift; # Get the socket.
       delete($ActiveTransactions{$Socket}); # Server has no transaction
   
     &Debug(6, "Server to idle");      &Debug(5, "Server to idle");
   
     #  If there's work to do, start the transaction:      #  If there's work to do, start the transaction:
   
     $reqdata = $WorkQueue->dequeue();      my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
     Debug(9, "Queue gave request data: ".$reqdata);      if ($reqdata ne undef)  {
     unless($reqdata eq undef)  {   Debug(5, "Queue gave request data: ".$reqdata->getRequest());
  my $unixSocket = $ClientQueue->dequeue();   &StartRequest($Socket,  $reqdata);
  &Debug(6, "Starting new work request");  
  &Debug(7, "Request: ".$reqdata);  
   
  &StartRequest($Socket, $unixSocket, $reqdata);  
     } else {      } else {
   
     #  There's no work waiting, so push the server to idle list.      #  There's no work waiting, so push the server to idle list.
  &Debug(8, "No new work requests, server connection going idle");   &Debug(5, "No new work requests, server connection going idle");
  delete($ActiveTransactions{$Socket});  
  $IdleConnections->push($Socket);   $IdleConnections->push($Socket);
     }      }
 }  }
   
 =pod  =pod
   
 =head2 ClientWritable  =head2 ClientWritable
   Event callback for when a client socket is writable.  
   This callback is established when a transaction reponse is  Event callback for when a client socket is writable.
   avaiable from lond.  The response is forwarded to the unix socket  
   as it becomes writable in this sub.  This callback is established when a transaction reponse is
   avaiable from lond.  The response is forwarded to the unix socket
   as it becomes writable in this sub.
   
 Parameters:  Parameters:
   
 =item Event - The event that has been triggered. Event->w->data is  =item Event
               the data and Event->w->fd is the socket to write.  
   The event that has been triggered. Event->w->data is
   the data and Event->w->fd is the socket to write.
   
 =cut  =cut
   
 sub ClientWritable {  sub ClientWritable {
     my $Event    = shift;      my $Event    = shift;
     my $Watcher  = $Event->w;      my $Watcher  = $Event->w;
Line 231  sub ClientWritable { Line 525  sub ClientWritable {
     &Debug(6, "ClientWritable writing".$Data);      &Debug(6, "ClientWritable writing".$Data);
     &Debug(9, "Socket is: ".$Socket);      &Debug(9, "Socket is: ".$Socket);
   
     my $result = $Socket->send($Data, 0);      if($Socket->connected) {
    my $result = $Socket->send($Data, 0);
     # $result undefined: the write failed.  
     # otherwise $result is the number of bytes written.  
     # Remove that preceding string from the data.  
     # If the resulting data is empty, destroy the watcher  
     # and set up a read event handler to accept the next  
     # request.  
   
     &Debug(9,"Send result is ".$result." Defined: ".defined($result));  
     if(defined($result)) {  
  &Debug(9, "send result was defined");  
  if($result == length($Data)) { # Entire string sent.  
     &Debug(9, "ClientWritable data all written");  
     $Watcher->cancel();  
     #  
     #  Set up to read next request from socket:  
       
     my $descr     = sprintf("Connection to lonc client %d",  
     $ActiveClients{$Socket});  
     Event->io(cb    => \&ClientRequest,  
       poll  => 'r',  
       desc  => $descr,  
       data  => "",  
       fd    => $Socket);  
   
  } else { # Partial string sent.  
     $Watcher->data(substr($Data, $result));  
  }  
   
     } else { # Error of some sort...   # $result undefined: the write failed.
    # otherwise $result is the number of bytes written.
  # Some errnos are possible:   # Remove that preceding string from the data.
  my $errno = $!;   # If the resulting data is empty, destroy the watcher
  if($errno == POSIX::EWOULDBLOCK   ||   # and set up a read event handler to accept the next
    $errno == POSIX::EAGAIN        ||   # request.
    $errno == POSIX::EINTR) {  
     # No action taken?  
  } else { # Unanticipated errno.  
     &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);  
     $Watcher->cancel; # Stop the watcher.  
     $Socket->shutdown(2); # Kill connection  
     $Socket->close(); # Close the socket.  
  }  
   
    &Debug(9,"Send result is ".$result." Defined: ".defined($result));
    if($result ne undef) {
       &Debug(9, "send result was defined");
       if($result == length($Data)) { # Entire string sent.
    &Debug(9, "ClientWritable data all written");
    $Watcher->cancel();
    #
    #  Set up to read next request from socket:
   
    my $descr     = sprintf("Connection to lonc client %d",
    $ActiveClients{$Socket});
    Event->io(cb    => \&ClientRequest,
     poll  => 'r',
     desc  => $descr,
     data  => "",
     fd    => $Socket);
   
       } else { # Partial string sent.
    $Watcher->data(substr($Data, $result));
    if($result == 0) {    # client hung up on us!!
       Log("INFO", "lonc pipe client hung up on us!");
       $Watcher->cancel;
       $Socket->shutdown(2);
       $Socket->close();
    }
       }
       
    } else { # Error of some sort...
       
       # Some errnos are possible:
       my $errno = $!;
       if($errno == POSIX::EWOULDBLOCK   ||
          $errno == POSIX::EAGAIN        ||
          $errno == POSIX::EINTR) {
    # No action taken?
       } else { # Unanticipated errno.
    &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
    $Watcher->cancel; # Stop the watcher.
    $Socket->shutdown(2); # Kill connection
    $Socket->close(); # Close the socket.
       }
       
    }
       } else {
    $Watcher->cancel(); # A delayed request...just cancel.
     }      }
 }  }
   
 =pod  =pod
   
 =head2 CompleteTransaction  =head2 CompleteTransaction
   Called when the reply data has been received for a lond   
   Called when the reply data has been received for a lond 
 transaction.   The reply data must now be sent to the  transaction.   The reply data must now be sent to the
 ultimate client on the other end of the Unix socket.  This is  ultimate client on the other end of the Unix socket.  This is
 done by setting up a writable event for the socket with the  done by setting up a writable event for the socket with the
 data the reply data.  data the reply data.
   
 Parameters:  Parameters:
 =item Socket - Socket on which the lond transaction occured.  This  
       is a LondConnection. The data received is in the   =item Socket
  TransactionReply member.  
 =item Client - Unix domain socket open on the ultimate client.  Socket on which the lond transaction occured.  This is a
   LondConnection. The data received is in the TransactionReply member.
   
   =item Transaction
   
   The transaction that is being completed.
   
 =cut  =cut
   
 sub CompleteTransaction {  sub CompleteTransaction {
     &Debug(6,"Complete transaction");      &Debug(5,"Complete transaction");
     my $Socket = shift;      my $Socket = shift;
     my $Client = shift;      my $Transaction = shift;
   
       if (!$Transaction->isDeferred()) { # Normal transaction
    my $data   = $Socket->GetReply(); # Data to send.
    StartClientReply($Transaction, $data);
       } else { # Delete deferred transaction file.
    Log("SUCCESS", "A delayed transaction was completed");
    LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
    unlink $Transaction->getFile();
       }
   }
   =pod
   =head1 StartClientReply
   
      Initiates a reply to a client where the reply data is a parameter.
   
   =head2  parameters:
   
   =item Transaction
   
     my $data   = $Socket->GetReply(); # Data to send.      The transaction for which we are responding to the client.
   
   =item data
   
       The data to send to apached client.
   
   =cut
   sub StartClientReply {
       my $Transaction   = shift;
       my $data     = shift;
   
   
       my $Client   = $Transaction->getClient();
   
     &Debug(8," Reply was: ".$data);      &Debug(8," Reply was: ".$data);
     my $Serial         = $ActiveClients{$Client};      my $Serial         = $ActiveClients{$Client};
Line 310  sub CompleteTransaction { Line 653  sub CompleteTransaction {
       cb       => \&ClientWritable,        cb       => \&ClientWritable,
       data     => $data);        data     => $data);
 }  }
   =pod
   =head2 FailTransaction
   
     Finishes a transaction with failure because the associated lond socket
     disconnected.  There are two possibilities:
     - The transaction is deferred: in which case we just quietly
       delete the transaction since there is no client connection.
     - The transaction is 'live' in which case we initiate the sending
       of "con_lost" to the client.
   
   Deleting the transaction means killing it from the 
   %ActiveTransactions hash.
   
   Parameters:
   
   =item client  
    
      The LondTransaction we are failing.
    
   =cut
   
   sub FailTransaction {
       my $transaction = shift;
       Log("WARNING", "Failing transaction ".$transaction->getRequest());
       Debug(1, "Failing transaction: ".$transaction->getRequest());
       if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
    my $client  = $transaction->getClient();
    Debug(1," Replying con_lost to ".$transaction->getRequest());
    StartClientReply($transaction, "con_lost\n");
       }
       if($ConnectionRetriesLeft <= 0) {
    Log("CRITICAL", "Host marked dead: ".GetServerHost());
       }
   
   }
   
   =pod
   =head1  EmptyQueue
   
     Fails all items in the work queue with con_lost.
     Note that each item in the work queue is a transaction.
   
   =cut
   sub EmptyQueue {
       $ConnectionRetriesLeft--; # Counts as connection failure too.
       while($WorkQueue->Count()) {
    my $request = $WorkQueue->dequeue(); # This is a transaction
    FailTransaction($request);
       }
   }
   
   =pod
   
   =head2 CloseAllLondConnections
   
   Close all connections open on lond prior to exit e.g.
   
   =cut
   sub CloseAllLondConnections {
       foreach my $Socket (keys %ActiveConnections) {
    KillSocket($Socket);
       }
   }
   =cut
   
   =pod
   
   =head2 KillSocket
    
   Destroys a socket.  This function can be called either when a socket
   has died of 'natural' causes or because a socket needs to be pruned due to
   idleness.  If the socket has died naturally, if there are no longer any 
   live connections a new connection is created (in case there are transactions
   in the queue).  If the socket has been pruned, it is never re-created.
   
   Parameters:
   
   =item Socket
    
     The socket to kill off.
   
   =item Restart
   
   nonzero if we are allowed to create a new connection.
   
   
   =cut
   sub KillSocket {
       my $Socket = shift;
   
       Log("WARNING", "Shutting down a socket");
       $Socket->Shutdown();
   
       #  If the socket came from the active connection set,
       #  delete its transaction... note that FailTransaction should
       #  already have been called!!!
       #  otherwise it came from the idle set.
       #  
       
       if(exists($ActiveTransactions{$Socket})) {
    delete ($ActiveTransactions{$Socket});
       }
       if(exists($ActiveConnections{$Socket})) {
    delete($ActiveConnections{$Socket});
       }
       $ConnectionCount--;
   
       #  If the connection count has gone to zero and there is work in the
       #  work queue, the work all gets failed with con_lost.
       #
       if($ConnectionCount == 0) {
    EmptyQueue();
       }
   }
   
 =pod  =pod
   
 =head2 LondReadable  =head2 LondReadable
   
 This function is called whenever a lond connection  This function is called whenever a lond connection
 is readable.  The action is state dependent:  is readable.  The action is state dependent:
   
 =head3 State = Initialized  =head3 State=Initialized
     We''re waiting for the challenge, this is a no-op until the  
   We''re waiting for the challenge, this is a no-op until the
 state changes.  state changes.
   
 =head3 State=Challenged   =head3 State=Challenged 
     The challenge has arrived we need to transition to Writable.  
   The challenge has arrived we need to transition to Writable.
 The connection must echo the challenge back.  The connection must echo the challenge back.
   
 =head3 State=ChallengeReplied  =head3 State=ChallengeReplied
     The challenge has been replied to.  The we are receiveing the   
   The challenge has been replied to.  The we are receiveing the 
 'ok' from the partner.  'ok' from the partner.
   
 =head3 State=RequestingKey  =head3 State=RequestingKey
     The ok has been received and we need to send the request for  
   The ok has been received and we need to send the request for
 an encryption key.  Transition to writable for that.  an encryption key.  Transition to writable for that.
   
 =head3 State=ReceivingKey  =head3 State=ReceivingKey
     The the key has been requested, now we are reading the new key.  
   The the key has been requested, now we are reading the new key.
   
 =head3 State=Idle   =head3 State=Idle 
     The encryption key has been negotiated or we have finished   
   The encryption key has been negotiated or we have finished 
 reading data from the a transaction.   If the callback data has  reading data from the a transaction.   If the callback data has
 a client as well as the socket iformation, then we are   a client as well as the socket iformation, then we are 
 doing a transaction and the data received is relayed to the client  doing a transaction and the data received is relayed to the client
 before the socket is put on the idle list.  before the socket is put on the idle list.
   
 =head3 State=SendingRequest  =head3 State=SendingRequest
     I do not think this state can be received here, but if it is,  
   I do not think this state can be received here, but if it is,
 the appropriate thing to do is to transition to writable, and send  the appropriate thing to do is to transition to writable, and send
 the request.  the request.
   
 =head3 State=ReceivingReply  =head3 State=ReceivingReply
     We finished sending the request to the server and now transition  
   We finished sending the request to the server and now transition
 to readable to receive the reply.   to readable to receive the reply. 
   
 The parameter to this function are:  The parameter to this function are:
   
 The event. Implicit in this is the watcher and its data.  The data   The event. Implicit in this is the watcher and its data.  The data 
 contains at least the lond connection object and, if a   contains at least the lond connection object and, if a 
 transaction is in progress, the socket attached to the local client.  transaction is in progress, the socket attached to the local client.
   
   
 =cut  =cut
   
 sub LondReadable {  sub LondReadable {
   
     my $Event      = shift;      my $Event      = shift;
     my $Watcher    = $Event->w;      my $Watcher    = $Event->w;
     my $Socket     = $Watcher->data;      my $Socket     = $Watcher->data;
     my $client     = undef;      my $client     = undef;
   
       &Debug(6,"LondReadable called state = ".$Socket->GetState());
   
   
     my $State = $Socket->GetState(); # All action depends on the state.      my $State = $Socket->GetState(); # All action depends on the state.
   
     &Debug(6,"LondReadable called state = ".$State);  
     SocketDump(6, $Socket);      SocketDump(6, $Socket);
       my $status = $Socket->Readable();
   
     if($Socket->Readable() != 0) {      &Debug(2, "Socket->Readable returned: $status");
  # bad return from socket read.  
       if($status != 0) {
    # bad return from socket read. Currently this means that
    # The socket has become disconnected. We fail the transaction.
   
    Log("WARNING",
       "Lond connection lost.");
    if(exists($ActiveTransactions{$Socket})) {
       FailTransaction($ActiveTransactions{$Socket});
    }
    $Watcher->cancel();
    KillSocket($Socket);
    $ConnectionRetriesLeft--;       # Counts as connection failure
    return;
     }      }
     SocketDump(6,$Socket);      SocketDump(6,$Socket);
   
Line 380  sub LondReadable { Line 870  sub LondReadable {
  # in the connection takes care of setting that up.  Just   # in the connection takes care of setting that up.  Just
  # need to transition to writable:   # need to transition to writable:
   
  $Watcher->poll("w");  
  $Watcher->cb(\&LondWritable);   $Watcher->cb(\&LondWritable);
    $Watcher->poll("w");
   
     } elsif ($State eq "ChallengeReplied") {      } elsif ($State eq "ChallengeReplied") {
   
Line 390  sub LondReadable { Line 880  sub LondReadable {
  #  The ok was received.  Now we need to request the key   #  The ok was received.  Now we need to request the key
  #  That requires us to be writable:   #  That requires us to be writable:
   
  $Watcher->poll("w");  
  $Watcher->cb(\&LondWritable);   $Watcher->cb(\&LondWritable);
    $Watcher->poll("w");
   
     } elsif ($State eq "ReceivingKey") {      } elsif ($State eq "ReceivingKey") {
   
     } elsif ($State eq "Idle") {      } elsif ($State eq "Idle") {
  # If necessary, complete a transaction and then go into the   # If necessary, complete a transaction and then go into the
  # idle queue.   # idle queue.
    #  Note that a trasition to idle indicates a live lond
    # on the other end so reset the connection retries.
    #
    $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
    $Watcher->cancel();
  if(exists($ActiveTransactions{$Socket})) {   if(exists($ActiveTransactions{$Socket})) {
     Debug(8,"Completing transaction!!");      Debug(5,"Completing transaction!!");
     CompleteTransaction($Socket,       CompleteTransaction($Socket, 
  $ActiveTransactions{$Socket});   $ActiveTransactions{$Socket});
    } else {
       Log("SUCCESS", "Connection ".$ConnectionCount." to "
    .$RemoteHost." now ready for action");
  }   }
  $Watcher->cancel();  
  ServerToIdle($Socket); # Next work unit or idle.   ServerToIdle($Socket); # Next work unit or idle.
   
     } elsif ($State eq "SendingRequest") {      } elsif ($State eq "SendingRequest") {
  #  We need to be writable for this and probably don't belong   #  We need to be writable for this and probably don't belong
  #  here inthe first place.   #  here inthe first place.
Line 422  sub LondReadable { Line 919  sub LondReadable {
  Debug(4, "Invalid state in LondReadable");   Debug(4, "Invalid state in LondReadable");
     }      }
 }  }
   
 =pod  =pod
   
 =head2 LondWritable  =head2 LondWritable
   
 This function is called whenever a lond connection  This function is called whenever a lond connection
 becomes writable while there is a writeable monitoring  becomes writable while there is a writeable monitoring
 event.  The action taken is very state dependent:  event.  The action taken is very state dependent:
   
 =head3 State = Connected   =head3 State = Connected 
   The connection is in the process of sending the  
   'init' hailing to the lond on the remote end.  The connection is in the process of sending the 'init' hailing to the
    The connection object''s Writable member is called.  lond on the remote end.  The connection object''s Writable member is
    On error, ConnectionError is called to destroy  called.  On error, ConnectionError is called to destroy the connection
    the connection and remove it from the ActiveConnections  and remove it from the ActiveConnections hash
    hash  
 =head3 Initialized  =head3 Initialized
    'init' has been sent, writability monitoring is   
     removed and readability monitoring is started  'init' has been sent, writability monitoring is removed and
     with LondReadable as the callback.  readability monitoring is started with LondReadable as the callback.
   
 =head3 ChallengeReceived  =head3 ChallengeReceived
     The connection has received the who are you   
    challenge from the remote system, and is in the  The connection has received the who are you challenge from the remote
     process of sending the challenge response.   system, and is in the process of sending the challenge
     Writable is called.  response. Writable is called.
   
 =head3 ChallengeReplied  =head3 ChallengeReplied
     The connection has replied to the initial challenge  
     The we switch to monitoring readability looking  The connection has replied to the initial challenge The we switch to
     for the server to reply with 'ok'.  monitoring readability looking for the server to reply with 'ok'.
   
 =head3 RequestingKey  =head3 RequestingKey
     The connection is in the process of requesting its  
     encryption key.  Writable is called.  The connection is in the process of requesting its encryption key.
   Writable is called.
   
 =head3 ReceivingKey  =head3 ReceivingKey
     The connection has sent the request for a key.  
     Switch to readability monitoring to accept the key  The connection has sent the request for a key.  Switch to readability
   monitoring to accept the key
   
 =head3 SendingRequest  =head3 SendingRequest
     The connection is in the process of sending a  
     request to the server.  This request is part of  The connection is in the process of sending a request to the server.
     a client transaction.  All the states until now  This request is part of a client transaction.  All the states until
     represent the client setup protocol. Writable  now represent the client setup protocol. Writable is called.
     is called.  
 =head3 ReceivingReply  =head3 ReceivingReply
     The connection has sent a request.  Now it must  
     receive a reply.  Readability monitoring is  
     requested.  
   
  This function is an event handler and therefore receives as  The connection has sent a request.  Now it must receive a reply.
   Readability monitoring is requested.
   
   This function is an event handler and therefore receives as
 a parameter the event that has fired.  The data for the watcher  a parameter the event that has fired.  The data for the watcher
 of this event is a reference to a list of one or two elements,  of this event is a reference to a list of one or two elements,
 depending on state. The first (and possibly only) element is the  depending on state. The first (and possibly only) element is the
Line 472  socket.  The second (present only if a r Line 980  socket.  The second (present only if a r
 is the socket on which to return a reply to the caller.  is the socket on which to return a reply to the caller.
   
 =cut  =cut
   
 sub LondWritable {  sub LondWritable {
     my $Event   = shift;      my $Event   = shift;
     my $Watcher = $Event->w;      my $Watcher = $Event->w;
     my @data    = $Watcher->data;      my $Socket  = $Watcher->data;
     Debug(6,"LondWritable State = ".$State." data has ".@data." elts.\n");      my $State   = $Socket->GetState();
   
     my $Socket  = $data[0]; # I know there's at least a socket.      Debug(6,"LondWritable State = ".$State."\n");
   
    
     #  Figure out what to do depending on the state of the socket:      #  Figure out what to do depending on the state of the socket:
           
   
     my $State   = $Socket->GetState();  
   
   
     SocketDump(6,$Socket);      SocketDump(6,$Socket);
   
     if      ($State eq "Connected")         {      if      ($State eq "Connected")         {
  #  "init" is being sent...  
   
  if ($Socket->Writable() != 0) {   if ($Socket->Writable() != 0) {
     #  The write resulted in an error.      #  The write resulted in an error.
       # We'll treat this as if the socket got disconnected:
       Log("WARNING", "Connection to ".$RemoteHost.
    " has been disconnected");
       FailTransaction($ActiveTransactions{$Socket});
       $Watcher->cancel();
       KillSocket($Socket);
       return;
  }   }
    #  "init" is being sent...
   
   
     } elsif ($State eq "Initialized")       {      } elsif ($State eq "Initialized")       {
   
  # Now that init was sent, we switch    # Now that init was sent, we switch 
  # to watching for readability:   # to watching for readability:
   
  $Watcher->poll("r");  
  $Watcher->cb(\&LondReadable);   $Watcher->cb(\&LondReadable);
    $Watcher->poll("r");
   
     } elsif ($State eq "ChallengeReceived") {      } elsif ($State eq "ChallengeReceived") {
  # We received the challenge, now we    # We received the challenge, now we 
Line 509  sub LondWritable { Line 1026  sub LondWritable {
  # we're waiting for the state to change   # we're waiting for the state to change
   
  if($Socket->Writable() != 0) {   if($Socket->Writable() != 0) {
     # Write of the next chunk resulted in an error.  
       $Watcher->cancel();
       KillSocket($Socket);
       return;
  }   }
   
     } elsif ($State eq "ChallengeReplied")  {      } elsif ($State eq "ChallengeReplied")  {
  # The echo was sent back, so we switch   # The echo was sent back, so we switch
  # to watching readability.   # to watching readability.
   
  $Watcher->poll("r");  
  $Watcher->cb(\&LondReadable);   $Watcher->cb(\&LondReadable);
    $Watcher->poll("r");
   
     } elsif ($State eq "RequestingKey")     {      } elsif ($State eq "RequestingKey")     {
  # At this time we're requesting the key.   # At this time we're requesting the key.
Line 527  sub LondWritable { Line 1047  sub LondWritable {
   
  if($Socket->Writable() != 0) {   if($Socket->Writable() != 0) {
     # Write resulted in an error.      # Write resulted in an error.
  }  
   
       $Watcher->cancel();
       KillSocket($Socket);
       return;
   
    }
     } elsif ($State eq "ReceivingKey")      {      } elsif ($State eq "ReceivingKey")      {
  # Now we need to wait for the key   # Now we need to wait for the key
  # to come back from the peer:   # to come back from the peer:
   
  $Watcher->poll("r");  
  $Watcher->cb(\&LondReadable);   $Watcher->cb(\&LondReadable);
    $Watcher->poll("r");
   
     } elsif ($State eq "SendingRequest")    {      } elsif ($State eq "SendingRequest")    {
  # At this time we are sending a request to the   # At this time we are sending a request to the
  # peer... write the next chunk:   # peer... write the next chunk:
   
  if($Socket->Writable() != 0) {   if($Socket->Writable() != 0) {
     # Write resulted in an error.  
   
       if(exists($ActiveTransactions{$Socket})) {
    Debug(3, "Lond connection lost, failing transactions");
    FailTransaction($ActiveTransactions{$Socket});
       }
       $Watcher->cancel();
       KillSocket($Socket);
       return;
       
  }   }
   
     } elsif ($State eq "ReceivingReply")    {      } elsif ($State eq "ReceivingReply")    {
  # The send has completed.  Wait for the   # The send has completed.  Wait for the
  # data to come in for a reply.   # data to come in for a reply.
  Debug(8,"Writable sent request/receiving reply");   Debug(8,"Writable sent request/receiving reply");
  $Watcher->poll("r");  
  $Watcher->cb(\&LondReadable);   $Watcher->cb(\&LondReadable);
    $Watcher->poll("r");
   
     } else {      } else {
  #  Control only passes here on an error:    #  Control only passes here on an error: 
Line 562  sub LondWritable { Line 1093  sub LondWritable {
     }      }
           
 }  }
   =pod
       
   =cut
   sub QueueDelayed {
       Debug(3,"QueueDelayed called");
   
       my $path = "$perlvar{'lonSockDir'}/delayed";
   
       Debug(4, "Delayed path: ".$path);
       opendir(DIRHANDLE, $path);
       
       my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
       closedir(DIRHANDLE);
       my $dfname;
       my $reqfile;
       foreach $dfname (sort  @alldelayed) {
    $reqfile = "$path/$dfname";
    Debug(4, "queueing ".$reqfile);
    my $Handle = IO::File->new($reqfile);
    my $cmd    = <$Handle>;
    chomp $cmd; # There may or may not be a newline...
    $cmd = $cmd."\n"; # now for sure there's exactly one newline.
    my $Transaction = LondTransaction->new($cmd);
    $Transaction->SetDeferred($reqfile);
    QueueTransaction($Transaction);
       }
       
   }
   
 =pod  =pod
   
 =head2 MakeLondConnection  =head2 MakeLondConnection
     Create a new lond connection object, and start it towards  
     its initial idleness.  Once idle, it becomes elligible to  Create a new lond connection object, and start it towards its initial
     receive transactions from the work queue.  If the work queue  idleness.  Once idle, it becomes elligible to receive transactions
     is not empty when the connection is completed and becomes idle,  from the work queue.  If the work queue is not empty when the
     it will dequeue an entry and start off on it.  connection is completed and becomes idle, it will dequeue an entry and
   start off on it.
   
 =cut  =cut
   
 sub MakeLondConnection {       sub MakeLondConnection {     
     Debug(4,"MakeLondConnection to ".GetServerHost()." on port "      Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
   .GetServerPort());    .GetServerPort());
Line 578  sub MakeLondConnection { Line 1141  sub MakeLondConnection {
     my $Connection = LondConnection->new(&GetServerHost(),      my $Connection = LondConnection->new(&GetServerHost(),
  &GetServerPort());   &GetServerPort());
   
     if($Connection == undef) { # Needs to be more robust later.      if($Connection eq undef) { # Needs to be more robust later.
  die "Failed to make a connection!!".$!."\n";   Log("CRITICAL","Failed to make a connection with lond.");
    $ConnectionRetriesLeft--;
    return 0; # Failure.
       }  else {
   
    # The connection needs to have writability 
    # monitored in order to send the init sequence
    # that starts the whole authentication/key
    # exchange underway.
    #
    my $Socket = $Connection->GetSocket();
    if($Socket eq undef) {
       die "did not get a socket from the connection";
    } else {
       &Debug(9,"MakeLondConnection got socket: ".$Socket);
    }
   
     }    $Connection->SetTimeoutCallback(\&SocketTimeout);
     # The connection needs to have writability   
     # monitored in order to send the init sequence  
     # that starts the whole authentication/key  
     # exchange underway.  
     #  
     my $Socket = $Connection->GetSocket();  
     if($Socket == undef) {  
  die "did not get a socket from the connection";  
     } else {  
  &Debug(9,"MakeLondConnection got socket: ".$Socket);  
     }  
   
        my $event = Event->io(fd       => $Socket,
     $event = Event->io(fd       => $Socket,     poll     => 'w',
        poll     => 'w',     cb       => \&LondWritable,
        cb       => \&LondWritable,     data     => $Connection,
        data     => ($Connection, undef),     desc => 'Connection to lond server');
        desc => 'Connection to lond server');   $ActiveConnections{$Connection} = $event;
     $ActiveConnections{$Lond} = $event;  
    $ConnectionCount++;
     $ConnectionCount++;   Debug(4, "Connection count = ".$ConnectionCount);
       if($ConnectionCount == 1) { # First Connection:
       QueueDelayed;
    }
    Log("SUCESS", "Created connection ".$ConnectionCount
       ." to host ".GetServerHost());
    return 1; # Return success.
       }
           
 }  }
   
 =pod  =pod
   
 =head2 StartRequest  =head2 StartRequest
    Starts a lond request going on a specified lond connection.  
     parameters are:  Starts a lond request going on a specified lond connection.
 =item $Lond - Connection to the lond that will send the transaction  parameters are:
     and receive the reply.  
 =item $Client - Connection to the client that is making this request  =item $Lond
     We got the request from this socket, and when the request has  
     been relayed to lond and we get a reply back from lond it will  Connection to the lond that will send the transaction and receive the
     get sent to this socket.  reply.
 =item $Request - The text of the request to send.  
   =item $Client
   
   Connection to the client that is making this request We got the
   request from this socket, and when the request has been relayed to
   lond and we get a reply back from lond it will get sent to this
   socket.
   
   =item $Request
   
   The text of the request to send.
   
 =cut  =cut
   
 sub StartRequest {  sub StartRequest {
     my $Lond     = shift;      my $Lond     = shift;
     my $Client   = shift;      my $Request  = shift; # This is a LondTransaction.
     my $Request  = shift;  
           
     Debug(6, "StartRequest: ".$Request);      Debug(6, "StartRequest: ".$Request->getRequest());
   
     my $Socket = $Lond->GetSocket();      my $Socket = $Lond->GetSocket();
           
     $ActiveTransactions{$Lond} = $Client; # Socket to relay to client.      $Request->Activate($Lond);
       $ActiveTransactions{$Lond} = $Request;
   
     $Lond->InitiateTransaction($Request);      $Lond->InitiateTransaction($Request->getRequest());
     $event = Event->io(fd      => $Lond->GetSocket(),      my $event = Event->io(fd      => $Socket,
        poll    => "w",         poll    => "w",
        cb      => \&LondWritable,         cb      => \&LondWritable,
        data    => $Lond,         data    => $Lond,
Line 641  sub StartRequest { Line 1227  sub StartRequest {
 }  }
   
 =pod  =pod
   
 =head2 QueueTransaction  =head2 QueueTransaction
     - If there is an idle lond connection, it is put to   
     work doing this transaction.  Otherwise, the transaction is  If there is an idle lond connection, it is put to work doing this
     placed in the work queue.  If placed in the work queue and the  transaction.  Otherwise, the transaction is placed in the work queue.
     maximum number of connections has not yet been created, a new  If placed in the work queue and the maximum number of connections has
     connection will be started.  Our goal is to eventually have   not yet been created, a new connection will be started.  Our goal is
     a sufficient number of connections that the work queue will  to eventually have a sufficient number of connections that the work
     typically be empty.  queue will typically be empty.  parameters are:
     parameters are:  
 =item Socket open on the lonc client.  =item Socket
 =item Request data to send to the lond.  
   open on the lonc client.
   
   =item Request
   
   data to send to the lond.
   
 =cut  =cut
   
 sub QueueTransaction {  sub QueueTransaction {
     my $requestSocket = shift;  
     my $requestData   = shift;  
   
     Debug(6,"QueueTransaction: ".$requestData);      my $requestData   = shift; # This is a LondTransaction.
       my $cmd           = $requestData->getRequest();
   
       Debug(6,"QueueTransaction: ".$cmd);
   
     my $LondSocket    = $IdleConnections->pop();      my $LondSocket    = $IdleConnections->pop();
     if(!defined $LondSocket) { # Need to queue request.      if(!defined $LondSocket) { # Need to queue request.
  Debug(8,"Must queue...");   Debug(5,"Must queue...");
  $ClientQueue->enqueue($requestSocket);  
  $WorkQueue->enqueue($requestData);   $WorkQueue->enqueue($requestData);
  if($ConnectionCount < $MaxConnectionCount) {   if($ConnectionCount < $MaxConnectionCount) {
     Debug(4,"Starting additional lond connection");      if($ConnectionRetriesLeft > 0) {
     MakeLondConnection();   Debug(5,"Starting additional lond connection");
    if(MakeLondConnection() == 0) {
       EmptyQueue(); # Fail transactions, can't make connection.
    }
       } else {
    ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
    EmptyQueue(); # It's worse than that ... he's dead Jim.
       }
  }   }
     } else { # Can start the request:      } else { # Can start the request:
  Debug(8,"Can start...");   Debug(8,"Can start...");
  StartRequest($LondSocket, $requestSocket, $requestData);   StartRequest($LondSocket,  $requestData);
     }      }
 }  }
   
 #-------------------------- Lonc UNIX socket handling ---------------------  #-------------------------- Lonc UNIX socket handling ---------------------
   
 =pod  =pod
   
 =head2 ClientRequest  =head2 ClientRequest
    Callback that is called when data can be read from the   Callback that is called when data can be read from the UNIX domain
    UNIX domain socket connecting us with an apache server process.  socket connecting us with an apache server process.
   
 =cut  =cut
   
Line 696  sub ClientRequest { Line 1298  sub ClientRequest {
     my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);      my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
     Debug(8, "rcv:  data length = ".length($thisread)      Debug(8, "rcv:  data length = ".length($thisread)
   ." read =".$thisread);    ." read =".$thisread);
     unless (defined $rv && length($thisread)) {      unless (defined $rv  && length($thisread)) {
  # Likely eof on socket.   # Likely eof on socket.
  Debug(5,"Client Socket closed on lonc for ".$RemoteHost);   Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
  close($socket);   close($socket);
  $watcher->cancel();   $watcher->cancel();
  delete($ActiveClients{$socket});   delete($ActiveClients{$socket});
    return;
     }      }
     Debug(8,"Data: ".$data." this read: ".$thisread);      Debug(8,"Data: ".$data." this read: ".$thisread);
     $data = $data.$thisread; # Append new data.      $data = $data.$thisread; # Append new data.
     $watcher->data($data);      $watcher->data($data);
     if($data =~ /(.*\n)/) { # Request entirely read.      if($data =~ /(.*\n)/) { # Request entirely read.
    if($data eq "close_connection_exit\n") {
       Log("CRITICAL",
    "Request Close Connection ... exiting");
       CloseAllLondConnections();
       exit;
    }
  Debug(8, "Complete transaction received: ".$data);   Debug(8, "Complete transaction received: ".$data);
  QueueTransaction($socket, $data);   my $Transaction = LondTransaction->new($data);
    $Transaction->SetClient($socket);
    QueueTransaction($Transaction);
  $watcher->cancel(); # Done looking for input data.   $watcher->cancel(); # Done looking for input data.
     }      }
   
Line 716  sub ClientRequest { Line 1327  sub ClientRequest {
   
   
 =pod  =pod
   
 =head2  NewClient  =head2  NewClient
     Callback that is called when a connection is received on the  
     unix socket for a new client of lonc.  The callback is parameterized  Callback that is called when a connection is received on the unix
     by the event.. which is a-priori assumed to be an  io event, and therefore  socket for a new client of lonc.  The callback is parameterized by the
     has an fd member that is the Listener socket.   We Accept the connection  event.. which is a-priori assumed to be an io event, and therefore has
     and register a new event on the readability of that socket:  an fd member that is the Listener socket.  We Accept the connection
   and register a new event on the readability of that socket:
   
 =cut  =cut
   
 sub NewClient {  sub NewClient {
     Debug(6, "NewClient");      Debug(6, "NewClient");
     my $event      = shift; # Get the event parameters.      my $event      = shift; # Get the event parameters.
Line 744  sub NewClient { Line 1359  sub NewClient {
     $ActiveClients{$connection} = $ClientConnection;      $ActiveClients{$connection} = $ClientConnection;
     $ClientConnection++;      $ClientConnection++;
 }  }
 =pod GetLoncSocketPath  
     Returns the name of the UNIX socket on which to listen for client  =pod
     connections.  
   =head2 GetLoncSocketPath
   
   Returns the name of the UNIX socket on which to listen for client
   connections.
   
 =cut  =cut
   
 sub GetLoncSocketPath {  sub GetLoncSocketPath {
     return $UnixSocketDir."/".GetServerHost();      return $UnixSocketDir."/".GetServerHost();
 }  }
   
 =pod GetServerHost   =pod
    Returns the host whose lond we talk with.  
   =head2 GetServerHost
   
   Returns the host whose lond we talk with.
   
 =cut  =cut
 sub GetServerHost { # Stub - get this from config.  
   sub GetServerHost {
     return $RemoteHost; # Setup by the fork.      return $RemoteHost; # Setup by the fork.
 }  }
 =pod GetServerPort  
    Returns the lond port number.  =pod
   
   =head2 GetServerPort
   
   Returns the lond port number.
   
 =cut  =cut
 sub GetServerPort { # Stub - get this from config.  
   sub GetServerPort {
     return $perlvar{londPort};      return $perlvar{londPort};
 }  }
 =pod SetupLoncListener  
    Setup a lonc listener event.  The event is called when  =pod
    the socket becomes readable.. that corresponds to the   
    receipt of a new connection.  The event handler established  =head2 SetupLoncListener
    will accept the connection (creating a communcations channel), that  
    int turn will establish another event handler to subess requests.  Setup a lonc listener event.  The event is called when the socket
   becomes readable.. that corresponds to the receipt of a new
   connection.  The event handler established will accept the connection
   (creating a communcations channel), that int turn will establish
   another event handler to subess requests.
   
 =cut  =cut
   
 sub SetupLoncListener {  sub SetupLoncListener {
   
     my $socket;      my $socket;
     my $SocketName = GetLoncSocketPath();      my $SocketName = GetLoncSocketPath();
     unlink($SocketName);      unlink($SocketName);
     unless ($socket = IO::Socket::UNIX->new(Local  => $SocketName,      unless ($socket =IO::Socket::UNIX->new(Local  => $SocketName,
     Listen => 10,       Listen => 10, 
     Type   => SOCK_STREAM)) {      Type   => SOCK_STREAM)) {
  die "Failed to create a lonc listner socket";   die "Failed to create a lonc listner socket";
Line 789  sub SetupLoncListener { Line 1425  sub SetupLoncListener {
       fd     => $socket);        fd     => $socket);
 }  }
   
   =pod 
   
   =head2 ChildStatus
    
   Child USR1 signal handler to report the most recent status
   into the status file.
   
   We also use this to reset the retries count in order to allow the
   client to retry connections with a previously dead server.
   =cut
   sub ChildStatus {
       my $event = shift;
       my $watcher = $event->w;
   
       Debug(2, "Reporting child status because : ".$watcher->data);
       my $docdir = $perlvar{'lonDocRoot'};
       my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");
       print $fh $$."\t".$RemoteHost."\t".$Status."\t".
    $RecentLogEntry."\n";
       $ConnectionRetriesLeft = $ConnectionRetries;
   }
   
 =pod  =pod
   
   =head2 SignalledToDeath
   
   Called in response to a signal that causes a chid process to die.
   
   =cut
   
   
   sub SignalledToDeath {
       my $event  = shift;
       my $watcher= $event->w;
   
       Debug(2,"Signalled to death! via ".$watcher->data);
       my ($signal) = $watcher->data;
       chomp($signal);
       Log("CRITICAL", "Abnormal exit.  Child $$ for $RemoteHost "
    ."died through "."\"$signal\"");
       LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
       ."\"$signal\"");
       exit 0;
   
   }
   
   =head2 ToggleDebug
   
   This sub toggles trace debugging on and off.
   
   =cut
   
   sub ToggleDebug {
       my $Current    = $DebugLevel;
          $DebugLevel = $NextDebugLevel;
          $NextDebugLevel = $Current;
   
       Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
   
   }
   
 =head2 ChildProcess  =head2 ChildProcess
   
 This sub implements a child process for a single lonc daemon.  This sub implements a child process for a single lonc daemon.
Line 798  This sub implements a child process for Line 1494  This sub implements a child process for
   
 sub ChildProcess {  sub ChildProcess {
   
     print "Loncnew\n";  
   
     # For now turn off signals.      #
           #  Signals must be handled by the Event framework...
     $SIG{QUIT}  = IGNORE;  #
     $SIG{HUP}   = IGNORE;  
     $SIG{USR1}  = IGNORE;      Event->signal(signal   => "QUIT",
     $SIG{INT}   = IGNORE;    cb       => \&SignalledToDeath,
     $SIG{CHLD}  = IGNORE;    data     => "QUIT");
     $SIG{__DIE__}  = IGNORE;      Event->signal(signal   => "HUP",
     cb       => \&ChildStatus,
     data     => "HUP");
       Event->signal(signal   => "USR1",
     cb       => \&ChildStatus,
     data     => "USR1");
       Event->signal(signal   => "INT",
     cb       => \&ToggleDebug,
     data     => "INT");
   
     SetupTimer();      SetupTimer();
           
Line 819  sub ChildProcess { Line 1522  sub ChildProcess {
   
 # Setup the initial server connection:  # Setup the initial server connection:
           
     &MakeLondConnection();       # &MakeLondConnection(); // let first work requirest do it.
       
   
     Debug(9,"Entering event loop");      Debug(9,"Entering event loop");
     my $ret = Event::loop(); #  Start the main event loop.      my $ret = Event::loop(); #  Start the main event loop.
           
Line 831  sub ChildProcess { Line 1535  sub ChildProcess {
 #  Create a new child for host passed in:  #  Create a new child for host passed in:
   
 sub CreateChild {  sub CreateChild {
       my $sigset = POSIX::SigSet->new(SIGINT);
       sigprocmask(SIG_BLOCK, $sigset);
     my $host = shift;      my $host = shift;
     $RemoteHost = $host;      $RemoteHost = $host;
     Debug(3, "Forking off child for ".$RemoteHost);      Log("CRITICAL", "Forking server for ".$host);
     sleep(5);      my $pid          = fork;
     $pid          = fork;  
     if($pid) { # Parent      if($pid) { # Parent
  $ChildHash{$pid} = $RemoteHost;   $RemoteHost = "Parent";
    $ChildHash{$pid} = $host;
    $HostToPid{$host}= $pid;
    sigprocmask(SIG_UNBLOCK, $sigset);
   
     } else { # child.      } else { # child.
  ChildProcess;   ShowStatus("Connected to ".$RemoteHost);
    $SIG{INT} = 'DEFAULT';
    sigprocmask(SIG_UNBLOCK, $sigset);
    ChildProcess; # Does not return.
     }      }
   
 }  }
Line 853  sub CreateChild { Line 1565  sub CreateChild {
 #  Each exit gets logged and the child gets restarted.  #  Each exit gets logged and the child gets restarted.
 #  #
   
   #
   #   Fork and start in new session so hang-up isn't going to 
   #   happen without intent.
   #
   
   
   
   
   
   
   ShowStatus("Forming new session");
   my $childpid = fork;
   if ($childpid != 0) {
       sleep 4; # Give child a chacne to break to
       exit 0; # a new sesion.
   }
   #
   #   Write my pid into the pid file so I can be located
   #
   
   ShowStatus("Parent writing pid file:");
   my $execdir = $perlvar{'lonDaemons'};
   open (PIDSAVE, ">$execdir/logs/lonc.pid");
   print PIDSAVE "$$\n";
   close(PIDSAVE);
   
   
   
   if (POSIX::setsid() < 0) {
       print "Could not create new session\n";
       exit -1;
   }
   
   ShowStatus("Forking node servers");
   
   Log("CRITICAL", "--------------- Starting children ---------------");
   
   LondConnection::ReadConfig;               # Read standard config files.
 my $HostIterator = LondConnection::GetHostIterator;  my $HostIterator = LondConnection::GetHostIterator;
 while (! $HostIterator->end()) {  while (! $HostIterator->end()) {
   
     $hostentryref = $HostIterator->get();      my $hostentryref = $HostIterator->get();
     CreateChild($hostentryref->[0]);      CreateChild($hostentryref->[0]);
       $HostHash{$hostentryref->[0]} = $hostentryref->[4];
     $HostIterator->next();      $HostIterator->next();
 }  }
   $RemoteHost = "Parent Server";
   
 # Maintain the population:  # Maintain the population:
   
   ShowStatus("Parent keeping the flock");
   
   #
   #   Set up parent signals:
   #
   
   $SIG{INT}  = \&Terminate;
   $SIG{TERM} = \&Terminate; 
   $SIG{HUP}  = \&Restart;
   $SIG{USR1} = \&CheckKids; 
   $SIG{USR2} = \&UpdateKids; # LonManage update request.
   
 while(1) {  while(1) {
     $deadchild = wait();      my $deadchild = wait();
     if(exists $ChildHash{$deadchild}) { # need to restart.      if(exists $ChildHash{$deadchild}) { # need to restart.
  $deadhost = $ChildHash{$deadchild};   my $deadhost = $ChildHash{$deadchild};
    delete($HostToPid{$deadhost});
  delete($ChildHash{$deadchild});   delete($ChildHash{$deadchild});
  Debug(4,"Lost child pid= ".$deadchild.   Log("WARNING","Lost child pid= ".$deadchild.
       "Connected to host ".$deadhost);        "Connected to host ".$deadhost);
    Log("INFO", "Restarting child procesing ".$deadhost);
  CreateChild($deadhost);   CreateChild($deadhost);
     }      }
 }  }
   
   
   
   =pod
   
   =head1 CheckKids
   
     Since kids do not die as easily in this implementation
   as the previous one, there  is no need to restart the
   dead ones (all dead kids get restarted when they die!!)
   The only thing this function does is to pass USR1 to the
   kids so that they report their status.
   
   =cut
   
   sub CheckKids {
       Debug(2, "Checking status of children");
       my $docdir = $perlvar{'lonDocRoot'};
       my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
       my $now=time;
       my $local=localtime($now);
       print $fh "LONC status $local - parent $$ \n\n";
       foreach my $pid (keys %ChildHash) {
    Debug(2, "Sending USR1 -> $pid");
    kill 'USR1' => $pid; # Tell Child to report status.
    sleep 1; # Wait so file doesn't intermix.
       }
   }
   
   =pod
   
   =head1  UpdateKids
   
   parent's SIGUSR2 handler.  This handler:
   
   =item
   
   Rereads the hosts file.
   
   =item
    
   Kills off (via sigint) children for hosts that have disappeared.
   
   =item
   
   QUITs  children for hosts that already exist (this just forces a status display
   and resets the connection retry count for that host.
   
   =item
   
   Starts new children for hosts that have been added to the hosts.tab file since
   the start of the master program and maintains them.
   
   =cut
   
   sub UpdateKids {
   
       Log("INFO", "Updating connections via SIGUSR2");
   
       #  Just in case we need to kill our own lonc, we wait a few seconds to
       #  give it a chance to receive and relay lond's response to the 
       #  re-init command.
       #
   
       sleep(2); # Wait a couple of seconds.
   
       my %hosts;                   # Indexed by loncapa hostname, value=ip.
       
       # Need to re-read  the host table:
       
       
       LondConnection::ReadConfig();
       my $I = LondConnection::GetHostIterator;
       while (! $I->end()) {
    my $item = $I->get();
    $hosts{$item->[0]} = $item->[4];
    $I->next();
       }
   
       #  The logic below is written for clarity not for efficiency.
       #  Since I anticipate that this function is only rarely called, that's
       #  appropriate.  There are certainly ways to combine the loops below,
       #  and anyone wishing to obscure the logic is welcome to go for it.
       #  Note that we don't re-direct sigchild.  Instead we do what's needed
       #  to the data structures that keep track of children to ensure that
       #  when sigchild is honored, no new child is born.
       #
   
       #  For each existing child; if it's host doesn't exist, kill the child.
   
       foreach my $child (keys %ChildHash) {
    my $oldhost = $ChildHash{$child};
    if (!(exists $hosts{$oldhost})) {
       Log("CRITICAL", "Killing child for $oldhost  host no longer exists");
       delete $ChildHash{$child};
       delete $HostToPid{$oldhost};
       kill 'QUIT' => $child;
    }
       }
       # For each remaining existing child; if it's host's ip has changed,
       # Restart the child on the new IP.
   
       foreach my $child (keys %ChildHash) {
    my $oldhost = $ChildHash{$child};
    my $oldip   = $HostHash{$oldhost};
    if ($hosts{$oldhost} ne $oldip) {
   
       # kill the old child.
   
       Log("CRITICAL", "Killing child for $oldhost host ip has changed...");
       delete $ChildHash{$child};
       delete $HostToPid{$oldhost};
       kill 'QUIT' => $child;
   
       # Do the book-keeping needed to start a new child on the
       # new ip.
   
       $HostHash{$oldhost} = $hosts{$oldhost};
       CreateChild($oldhost);
    }
       }
       # Finally, for each new host, not in the host hash, create a
       # enter the host and create a new child.
       # Force a status display of any existing process.
   
       foreach my $host (keys %hosts) {
    if(!(exists $HostHash{$host})) {
       Log("INFO", "New host $host discovered in hosts.tab...");
       $HostHash{$host} = $hosts{$host};
       CreateChild($host);
    } else {
       kill 'HUP' => $HostToPid{$host};    # status display.
    }
       }
   }
   
   
   =pod
   
   =head1 Restart
   
   Signal handler for HUP... all children are killed and
   we self restart.  This is an el-cheapo way to re read
   the config file.
   
   =cut
   
   sub Restart {
       &KillThemAll; # First kill all the children.
       Log("CRITICAL", "Restarting");
       my $execdir = $perlvar{'lonDaemons'};
       unlink("$execdir/logs/lonc.pid");
       exec("$execdir/loncnew");
   }
   
   =pod
   
   =head1 KillThemAll
   
   Signal handler that kills all children by sending them a 
   SIGHUP.  Responds to sigint and sigterm.
   
   =cut
   
   sub KillThemAll {
       Debug(2, "Kill them all!!");
       local($SIG{CHLD}) = 'IGNORE';      # Our children >will< die.
       foreach my $pid (keys %ChildHash) {
    my $serving = $ChildHash{$pid};
    Debug(2, "Killing lonc for $serving pid = $pid");
    ShowStatus("Killing lonc for $serving pid = $pid");
    Log("CRITICAL", "Killing lonc for $serving pid = $pid");
    kill 'QUIT' => $pid;
    delete($ChildHash{$pid});
       }
       my $execdir = $perlvar{'lonDaemons'};
       unlink("$execdir/logs/lonc.pid");
   
   }
   
   =pod
   
   =head1 Terminate
    
   Terminate the system.
   
   =cut
   
   sub Terminate {
       KillThemAll;
       Log("CRITICAL","Master process exiting");
       exit 0;
   
   }
   =pod
   
 =head1 Theory  =head1 Theory
    The event class is used to build this as a single process with  
    an event driven model.  The following events are handled:  The event class is used to build this as a single process with an
   event driven model.  The following events are handled:
   
 =item UNIX Socket connection Received  =item UNIX Socket connection Received
   
Line 890  All sockets are run in non-blocking mode Line 1852  All sockets are run in non-blocking mode
 handler prevents hung connections.  handler prevents hung connections.
   
 Key data structures:  Key data structures:
 =item RequestQueue - A queue of requests received from UNIX sockets that are  
     waiting for a chance to be forwarded on a lond connection socket.  
   
 =item ActiveConnections - A hash of lond connections that have transactions  =item RequestQueue
     in process that are available to be timed out.  
   A queue of requests received from UNIX sockets that are
   waiting for a chance to be forwarded on a lond connection socket.
   
   =item ActiveConnections
   
   A hash of lond connections that have transactions in process that are
   available to be timed out.
   
   =item ActiveTransactions
   
   A hash indexed by lond connections that contain the client reply
   socket for each connection that has an active transaction on it.
   
   =item IdleConnections
   
   A hash of lond connections that have no work to do.  These connections
   can be closed if they are idle for a long enough time.
   
 =item ActiveTransactions - A hash indexed by lond connections that  
     contain the client reply socket for each connection that has an active  
     transaction on it.  
   
 =item IdleConnections - A hash of lond connections that have no work  
     to do.  These connections can be closed if they are idle for a long  
     enough time.  
 =cut  =cut

Removed from v.1.2  
changed lines
  Added in v.1.31


FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>