Diff for /loncom/loncnew between versions 1.65 and 1.77

version 1.65, 2004/10/05 11:16:57 version 1.77, 2007/03/27 19:38:36
Line 61  use LONCAPA::LondConnection; Line 61  use LONCAPA::LondConnection;
 use LONCAPA::LondTransaction;  use LONCAPA::LondTransaction;
 use LONCAPA::Configuration;  use LONCAPA::Configuration;
 use LONCAPA::HashIterator;  use LONCAPA::HashIterator;
   use Fcntl qw(:flock);
   
   
 # Read the httpd configuration file to get perl variables  # Read the httpd configuration file to get perl variables
Line 105  my $ConnectionCount = 0; Line 106  my $ConnectionCount = 0;
 my $IdleSeconds     = 0; # Number of seconds idle.  my $IdleSeconds     = 0; # Number of seconds idle.
 my $Status          = ""; # Current status string.  my $Status          = ""; # Current status string.
 my $RecentLogEntry  = "";  my $RecentLogEntry  = "";
 my $ConnectionRetries=2; # Number of connection retries allowed.  my $ConnectionRetries=5; # Number of connection retries allowed.
 my $ConnectionRetriesLeft=2; # Number of connection retries remaining.  my $ConnectionRetriesLeft=5; # Number of connection retries remaining.
 my $LondVersion     = "unknown"; # Version of lond we talk with.  my $LondVersion     = "unknown"; # Version of lond we talk with.
 my $KeyMode         = "";       # e.g. ssl, local, insecure from last connect.  my $KeyMode         = "";       # e.g. ssl, local, insecure from last connect.
 my $LondConnecting  = 0;       # True when a connection is being built.  my $LondConnecting  = 0;       # True when a connection is being built.
Line 114  my $LondConnecting  = 0;       # True wh Line 115  my $LondConnecting  = 0;       # True wh
   
   
 my $DieWhenIdle     = 1; # When true children die when trimmed -> 0.  my $DieWhenIdle     = 1; # When true children die when trimmed -> 0.
   my $hosts_tab       = 1;        # True if we are using a static hosts.tab
 my $I_am_child      = 0; # True if this is the child process.  my $I_am_child      = 0; # True if this is the child process.
   
 #  #
Line 151  sub UpdateStatus { Line 153  sub UpdateStatus {
 Makes an entry into the permanent log file.  Makes an entry into the permanent log file.
   
 =cut  =cut
   
 sub LogPerm {  sub LogPerm {
     my $message=shift;      my $message=shift;
     my $execdir=$perlvar{'lonDaemons'};      my $execdir=$perlvar{'lonDaemons'};
Line 196  sub Log { Line 199  sub Log {
     my $now   = time;      my $now   = time;
     my $local = localtime($now);      my $local = localtime($now);
     my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";      my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
     my $finalformat = $finalformat.$format."\n";      $finalformat = $finalformat.$format."\n";
   
     # open the file and put the result.      # open the file and put the result.
   
Line 270  sub SocketDump { Line 273  sub SocketDump {
  and as what we return in a SIGUSR1   and as what we return in a SIGUSR1
   
 =cut  =cut
   
 sub ShowStatus {  sub ShowStatus {
     my $state = shift;      my $state = shift;
     my $now = time;      my $now = time;
Line 280  sub ShowStatus { Line 284  sub ShowStatus {
   
 =pod  =pod
   
 =head 2 SocketTimeout  =head2 SocketTimeout
   
     Called when an action on the socket times out.  The socket is       Called when an action on the socket times out.  The socket is 
    destroyed and any active transaction is failed.     destroyed and any active transaction is failed.
   
   
 =cut  =cut
   
 sub SocketTimeout {  sub SocketTimeout {
     my $Socket = shift;      my $Socket = shift;
     Log("WARNING", "A socket timeout was detected");      Log("WARNING", "A socket timeout was detected");
Line 327  sub child_exit { Line 332  sub child_exit {
  #  during which no listens will be done on the   #  during which no listens will be done on the
  #  lonnet client socket.   #  lonnet client socket.
  #   #
  my $lock_file = GetLoncSocketPath().".lock";   my $lock_file = &GetLoncSocketPath().".lock";
  open(LOCK,">$lock_file");   open(LOCK,">$lock_file");
  print LOCK "Contents not important";   print LOCK "Contents not important";
  close(LOCK);   close(LOCK);
    if ($hosts_tab) {
       unlink(&GetLoncSocketPath());
    }
  exit(0);   exit(0);
     }      }
     #  Now figure out how we exit:      #  Now figure out how we exit:
Line 419  sub Tick { Line 426  sub Tick {
  $KeyMode = "";    $KeyMode = ""; 
  $clock_watcher->cancel();   $clock_watcher->cancel();
     }      }
       &UpdateStatus();
 }  }
   
 =pod  =pod
Line 672  sub FailTransaction { Line 680  sub FailTransaction {
   
     if ($ConnectionRetriesLeft > 0) {      if ($ConnectionRetriesLeft > 0) {
  Log("WARNING", "Failing transaction "   Log("WARNING", "Failing transaction "
     .$transaction->getRequest());      .$transaction->getLoggableRequest());
     }      }
     Debug(1, "Failing transaction: ".$transaction->getRequest());      Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
     if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.      if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
  my $client  = $transaction->getClient();   my $client  = $transaction->getClient();
  Debug(1," Replying con_lost to ".$transaction->getRequest());   Debug(1," Replying con_lost to ".$transaction->getRequest());
Line 684  sub FailTransaction { Line 692  sub FailTransaction {
 }  }
   
 =pod  =pod
   
 =head1  EmptyQueue  =head1  EmptyQueue
   
   Fails all items in the work queue with con_lost.    Fails all items in the work queue with con_lost.
   Note that each item in the work queue is a transaction.    Note that each item in the work queue is a transaction.
   
 =cut  =cut
   
 sub EmptyQueue {  sub EmptyQueue {
     $ConnectionRetriesLeft--; # Counts as connection failure too.      $ConnectionRetriesLeft--; # Counts as connection failure too.
     while($WorkQueue->Count()) {      while($WorkQueue->Count()) {
Line 705  sub EmptyQueue { Line 715  sub EmptyQueue {
 Close all connections open on lond prior to exit e.g.  Close all connections open on lond prior to exit e.g.
   
 =cut  =cut
   
 sub CloseAllLondConnections {  sub CloseAllLondConnections {
     foreach my $Socket (keys %ActiveConnections) {      foreach my $Socket (keys %ActiveConnections) {
       if(exists($ActiveTransactions{$Socket})) {        if(exists($ActiveTransactions{$Socket})) {
Line 713  sub CloseAllLondConnections { Line 724  sub CloseAllLondConnections {
       KillSocket($Socket);        KillSocket($Socket);
     }      }
 }  }
 =cut  
   
 =pod  =pod
   
Line 735  Parameters: Line 745  Parameters:
   
 nonzero if we are allowed to create a new connection.  nonzero if we are allowed to create a new connection.
   
   
 =cut  =cut
   
 sub KillSocket {  sub KillSocket {
     my $Socket = shift;      my $Socket = shift;
   
Line 948  sub LondReadable { Line 958  sub LondReadable {
  #  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.
   
  Deubg(6, "SendingRequest state encountered in readable");   Debug(6, "SendingRequest state encountered in readable");
  $Watcher->poll("w");   $Watcher->poll("w");
  $Watcher->cb(\&LondWritable);   $Watcher->cb(\&LondWritable);
   
Line 1149  sub LondWritable { Line 1159  sub LondWritable {
 =pod  =pod
           
 =cut  =cut
   
 sub QueueDelayed {  sub QueueDelayed {
     Debug(3,"QueueDelayed called");      Debug(3,"QueueDelayed called");
   
Line 1156  sub QueueDelayed { Line 1167  sub QueueDelayed {
   
     Debug(4, "Delayed path: ".$path);      Debug(4, "Delayed path: ".$path);
     opendir(DIRHANDLE, $path);      opendir(DIRHANDLE, $path);
       
     my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;      my @all_host_ids;
       my $host_iterator =   &LondConnection::GetHostIterator();
       while (!$host_iterator->end()) {
    my ($host_id,$host_name) = @{$host_iterator->get()}[0,3];
    if ($host_name eq $RemoteHost) {
       push(@all_host_ids, $host_id);
    }
    $host_iterator->next();
       }
       my $host_id_re = '(?:'.join('|',@all_host_ids).')';
       my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
     closedir(DIRHANDLE);      closedir(DIRHANDLE);
     my $dfname;      foreach my $dfname (sort(@alldelayed)) {
     my $reqfile;   my $reqfile = "$path/$dfname";
     foreach $dfname (sort  @alldelayed) {   my ($host_id) = ($dfname =~ /\.([^.]*)$/);
  $reqfile = "$path/$dfname";   Debug(4, "queueing ".$reqfile." for $host_id");
  Debug(4, "queueing ".$reqfile);  
  my $Handle = IO::File->new($reqfile);   my $Handle = IO::File->new($reqfile);
  my $cmd    = <$Handle>;   my $cmd    = <$Handle>;
  chomp $cmd; # There may or may not be a newline...   chomp $cmd; # There may or may not be a newline...
  $cmd = $cmd."\n"; # now for sure there's exactly one newline.   $cmd = $cmd."\n"; # now for sure there's exactly one newline.
  my $Transaction = LondTransaction->new($cmd);   my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
  $Transaction->SetDeferred($reqfile);   $Transaction->SetDeferred($reqfile);
  QueueTransaction($Transaction);   QueueTransaction($Transaction);
     }      }
Line 1557  into the status file. Line 1577  into the status file.
   
 We also use this to reset the retries count in order to allow the  We also use this to reset the retries count in order to allow the
 client to retry connections with a previously dead server.  client to retry connections with a previously dead server.
   
 =cut  =cut
   
 sub ChildStatus {  sub ChildStatus {
Line 1565  sub ChildStatus { Line 1586  sub ChildStatus {
   
     Debug(2, "Reporting child status because : ".$watcher->data);      Debug(2, "Reporting child status because : ".$watcher->data);
     my $docdir = $perlvar{'lonDocRoot'};      my $docdir = $perlvar{'lonDocRoot'};
     my $fh = IO::File->new(">>$docdir/lon-status/loncstatus.txt");      
     print $fh $$."\t".$RemoteHost."\t".$Status."\t".      open(LOG,">>$docdir/lon-status/loncstatus.txt");
       flock(LOG,LOCK_EX);
       print LOG $$."\t".$RemoteHost."\t".$Status."\t".
  $RecentLogEntry."\n";   $RecentLogEntry."\n";
     #      #
     #  Write out information about each of the connections:      #  Write out information about each of the connections:
     #      #
     if ($DebugLevel > 2) {      if ($DebugLevel > 2) {
  print $fh "Active connection statuses: \n";   print LOG "Active connection statuses: \n";
  my $i = 1;   my $i = 1;
  print STDERR  "================================= Socket Status Dump:\n";   print STDERR  "================================= Socket Status Dump:\n";
  foreach my $item (keys %ActiveConnections) {   foreach my $item (keys %ActiveConnections) {
     my $Socket = $ActiveConnections{$item}->data;      my $Socket = $ActiveConnections{$item}->data;
     my $state  = $Socket->GetState();      my $state  = $Socket->GetState();
     print $fh "Connection $i State: $state\n";      print LOG "Connection $i State: $state\n";
     print STDERR "---------------------- Connection $i \n";      print STDERR "---------------------- Connection $i \n";
     $Socket->Dump(-1); # Ensure it gets dumped..      $Socket->Dump(-1); # Ensure it gets dumped..
     $i++;      $i++;
  }   }
     }      }
       flock(LOG,LOCK_UN);
       close(LOG);
     $ConnectionRetriesLeft = $ConnectionRetries;      $ConnectionRetriesLeft = $ConnectionRetries;
       UpdateStatus();
 }  }
   
 =pod  =pod
Line 1605  sub SignalledToDeath { Line 1631  sub SignalledToDeath {
     chomp($signal);      chomp($signal);
     Log("CRITICAL", "Abnormal exit.  Child $$ for $RemoteHost "      Log("CRITICAL", "Abnormal exit.  Child $$ for $RemoteHost "
  ."died through "."\"$signal\"");   ."died through "."\"$signal\"");
     LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "      #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
     ."\"$signal\"");  #    ."\"$signal\"");
     exit 0;      exit 0;
   
 }  }
   
   =pod
   
 =head2 ToggleDebug  =head2 ToggleDebug
   
 This sub toggles trace debugging on and off.  This sub toggles trace debugging on and off.
Line 1626  sub ToggleDebug { Line 1654  sub ToggleDebug {
   
 }  }
   
   =pod
   
 =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 1704  sub ChildProcess { Line 1734  sub ChildProcess {
       desc => 'Lonc Listener Unix Socket',        desc => 'Lonc Listener Unix Socket',
       fd   => $socket);        fd   => $socket);
           
     $Event::Debuglevel = $DebugLevel;      $Event::DebugLevel = $DebugLevel;
           
     Debug(9, "Making initial lond connection for ".$RemoteHost);      Debug(9, "Making initial lond connection for ".$RemoteHost);
   
Line 1778  sub parent_client_connection { Line 1808  sub parent_client_connection {
  my ($event)   = @_;   my ($event)   = @_;
  my $watcher   = $event->w;   my $watcher   = $event->w;
  my $socket    = $watcher->fd;   my $socket    = $watcher->fd;
    if ($hosts_tab) {
   
  # Lookup the host associated with this socket:      # Lookup the host associated with this socket:
   
  my $host = $listening_to{$socket};      my $host = $listening_to{$socket};
   
  # Start the child:      # Start the child:
   
   
   
       &Debug(9,"Creating child for $host (parent_client_connection)");
       &CreateChild($host, $socket);
       
       # Clean up the listen since now the child takes over until it exits.
   
       $watcher->cancel(); # Nolonger listening to this event
       delete($listening_to{$socket});
       delete($parent_dispatchers{$host});
       $socket->close();
   
    } else {
       my $connection = $socket->accept(); # Accept the client connection.
       Event->io(cb      => \&get_remote_hostname,
         poll    => 'r',
         data    => "",
         fd      => $connection);
    }
       }
   }
   
   sub get_remote_hostname {
    my ($event)   = @_;
    my $watcher   = $event->w;
    my $socket    = $watcher->fd;
   
    my $thisread;
    my $rv = $socket->recv($thisread, 1, 0);
    Debug(8, "rcv:  data length = ".length($thisread)." read =".$thisread);
    if (!defined($rv) || length($thisread) == 0) {
       # Likely eof on socket.
       Debug(5,"Client Socket closed on lonc for p_c_c");
       close($socket);
       $watcher->cancel();
       return;
    }
   
    my $data    = $watcher->data().$thisread;
    $watcher->data($data);
    if($data =~ /\n$/) { # Request entirely read.
       chomp($data);
    } else {
       return;
    }
   
  &Debug(9,"Creating child for $host (parent_client_connection)");   &Debug(5,"Creating child for $data (parent_client_connection)");
  &CreateChild($host, $socket);   &CreateChild($data);
   
  # Clean up the listen since now the child takes over until it exits.   # Clean up the listen since now the child takes over until it exits.
   
  $watcher->cancel(); # Nolonger listening to this event   $watcher->cancel(); # Nolonger listening to this event
  delete($listening_to{$socket});   $socket->send("done\n");
  delete($parent_dispatchers{$host});  
  $socket->close();   $socket->close();
     }  
 }  }
   
 # parent_listen:  # parent_listen:
Line 1818  sub parent_listen { Line 1890  sub parent_listen {
     my ($loncapa_host) = @_;      my ($loncapa_host) = @_;
     Debug(5, "parent_listen: $loncapa_host");      Debug(5, "parent_listen: $loncapa_host");
   
     my $socket    = &SetupLoncListener($loncapa_host);      my $socket   = &SetupLoncListener($loncapa_host);
     $listening_to{$socket} = $loncapa_host;      $listening_to{$socket} = $loncapa_host;
     if (!$socket) {      if (!$socket) {
  die "Unable to create a listen socket for $loncapa_host";   die "Unable to create a listen socket for $loncapa_host";
Line 1827  sub parent_listen { Line 1899  sub parent_listen {
     my $lock_file = &GetLoncSocketPath($loncapa_host).".lock";      my $lock_file = &GetLoncSocketPath($loncapa_host).".lock";
     unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]      unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
   
     my $watcher = Event->io(cb    => \&parent_client_connection,      my $watcher = 
       poll  => 'r',   Event->io(cb    => \&parent_client_connection,
       desc  => "Parent listener unix socket ($loncapa_host)",    poll  => 'r',
       fd    => $socket);    desc  => "Parent listener unix socket ($loncapa_host)",
     data => "",
     fd    => $socket);
     $parent_dispatchers{$loncapa_host} = $watcher;      $parent_dispatchers{$loncapa_host} = $watcher;
   
 }  }
   
   sub parent_clean_up {
       my ($loncapa_host) = @_;
       Debug(5, "parent_clean_up: $loncapa_host");
   
       my $socket_file = &GetLoncSocketPath($loncapa_host);
       unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
       my $lock_file   = $socket_file.".lock";
       unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
   }
   
   
 # listen_on_all_unix_sockets:  # listen_on_all_unix_sockets:
 #    This sub initiates a listen on all unix domain lonc client sockets.  #    This sub initiates a listen on all unix domain lonc client sockets.
Line 1857  sub listen_on_all_unix_sockets { Line 1941  sub listen_on_all_unix_sockets {
     my $host_iterator      =   &LondConnection::GetHostIterator();      my $host_iterator      =   &LondConnection::GetHostIterator();
     while (!$host_iterator->end()) {      while (!$host_iterator->end()) {
  my $host_entry_ref =   $host_iterator->get();   my $host_entry_ref =   $host_iterator->get();
  my $host_name      = $host_entry_ref->[0];   my $host_name      = $host_entry_ref->[3];
  Debug(9, "Listen for $host_name");   Debug(9, "Listen for $host_name");
  &parent_listen($host_name);   &parent_listen($host_name);
  $host_iterator->next();   $host_iterator->next();
     }      }
 }  }
   
   sub listen_on_common_socket {
       Debug(5, "listen_on_common_socket");
       &parent_listen('common');
   }
   
 #   server_died is called whenever a child process exits.  #   server_died is called whenever a child process exits.
 #   Since this is dispatched via a signal, we must process all  #   Since this is dispatched via a signal, we must process all
 #   dead children until there are no more left.  The action  #   dead children until there are no more left.  The action
Line 1888  sub server_died { Line 1977  sub server_died {
     &Debug(9, "Caught sigchild for $host");      &Debug(9, "Caught sigchild for $host");
     delete($ChildHash{$pid});      delete($ChildHash{$pid});
     delete($HostToPid{$host});      delete($HostToPid{$host});
     &parent_listen($host);      if ($hosts_tab) {
    &parent_listen($host);
       } else {
    &parent_clean_up($host);
       }
       
  } else {   } else {
     &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");      &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
  }   }
Line 1949  my $HostIterator = LondConnection::GetHo Line 2042  my $HostIterator = LondConnection::GetHo
   
 if ($DieWhenIdle) {  if ($DieWhenIdle) {
     $RemoteHost = "[parent]";      $RemoteHost = "[parent]";
     &listen_on_all_unix_sockets();      if ($hosts_tab) {
    &listen_on_all_unix_sockets();
       } else {
    &listen_on_common_socket();
       }
 } else {  } else {
           
     while (! $HostIterator->end()) {      while (! $HostIterator->end()) {
Line 1987  if ($DieWhenIdle) { Line 2084  if ($DieWhenIdle) {
     $parent_handlers{TERM} = Event->signal(cb       => \&Terminate,      $parent_handlers{TERM} = Event->signal(cb       => \&Terminate,
    desc     => "Parent TERM handler",     desc     => "Parent TERM handler",
    signal   => "TERM");     signal   => "TERM");
     $parent_handlers{HUP}  = Event->signal(cb       => \&Restart,      if ($hosts_tab) {
    desc     => "Parent HUP handler.",   $parent_handlers{HUP}  = Event->signal(cb       => \&Restart,
    signal   => "HUP");         desc     => "Parent HUP handler.",
          signal   => "HUP");
       } else {
    $parent_handlers{HUP}  = Event->signal(cb       => \&KillThemAll,
          desc     => "Parent HUP handler.",
          signal   => "HUP");
       }
     $parent_handlers{USR1} = Event->signal(cb       => \&CheckKids,      $parent_handlers{USR1} = Event->signal(cb       => \&CheckKids,
    desc     => "Parent USR1 handler",     desc     => "Parent USR1 handler",
    signal   => "USR1");     signal   => "USR1");
Line 2012  if ($DieWhenIdle) { Line 2115  if ($DieWhenIdle) {
           
     $SIG{INT}  = \&Terminate;      $SIG{INT}  = \&Terminate;
     $SIG{TERM} = \&Terminate;       $SIG{TERM} = \&Terminate; 
     $SIG{HUP}  = \&Restart;      if ($hosts_tab) {
    $SIG{HUP}  = \&Restart;
       } else {
    $SIG{HUP}  = \&KillThemAll;
       }
     $SIG{USR1} = \&CheckKids;       $SIG{USR1} = \&CheckKids; 
     $SIG{USR2} = \&UpdateKids; # LonManage update request.      $SIG{USR2} = \&UpdateKids; # LonManage update request.
           
Line 2056  sub CheckKids { Line 2163  sub CheckKids {
     foreach my $pid (keys %ChildHash) {      foreach my $pid (keys %ChildHash) {
  Debug(2, "Sending USR1 -> $pid");   Debug(2, "Sending USR1 -> $pid");
  kill 'USR1' => $pid; # Tell Child to report status.   kill 'USR1' => $pid; # Tell Child to report status.
  sleep 1; # Wait so file doesn't intermix.  
     }      }
   
 }  }
Line 2098  sub UpdateKids { Line 2204  sub UpdateKids {
     # The down side is transactions that are in flight will get timed out      # The down side is transactions that are in flight will get timed out
     # (lost unless they are critical).      # (lost unless they are critical).
   
     &Restart();      if ($hosts_tab) {
    &Restart();
       } else {
    &KillThemAll();
       }
 }  }
   
   
Line 2139  sub KillThemAll { Line 2248  sub KillThemAll {
  Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");   Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
  kill 'QUIT' => $pid;   kill 'QUIT' => $pid;
     }      }
   
   
 }  }
   
   
Line 2162  sub really_kill_them_all_dammit Line 2269  sub really_kill_them_all_dammit
  unlink("$execdir/logs/lonc.pid");   unlink("$execdir/logs/lonc.pid");
     }      }
 }  }
   
 =pod  =pod
   
 =head1 Terminate  =head1 Terminate

Removed from v.1.65  
changed lines
  Added in v.1.77


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