loncom/loncnew - diff

Return to loncnew CVS log

Up to [LON-CAPA] / loncom

Diff for /loncom/loncnew between versions 1.64 and 1.90

version 1.64, 2004/10/05 10:10:31	version 1.90, 2010/12/20 11:15:51
Line 60 use LONCAPA::Stack;	Line 60 use LONCAPA::Stack;
use LONCAPA::LondConnection;	use LONCAPA::LondConnection;
use LONCAPA::LondTransaction;	use LONCAPA::LondTransaction;
use LONCAPA::Configuration;	use LONCAPA::Configuration;
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 72 my %perlvar = %{$perlvarref};	Line 72 my %perlvar = %{$perlvarref};
#	#
# parent and shared variables.	# parent and shared variables.

my %ChildHash; # by pid -> host.	my %ChildPid; # by pid -> host.
my %HostToPid; # By host -> pid.	my %ChildHost; # by host.
my %HostHash; # by loncapaname -> IP.
my %listening_to; # Socket->host table for who the parent	my %listening_to; # Socket->host table for who the parent
# is listening to.	# is listening to.
my %parent_dispatchers; # host-> listener watcher events.	my %parent_dispatchers; # host-> listener watcher events.

	my %parent_handlers; # Parent signal handlers...

my $MaxConnectionCount = 10; # 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 = 0;	my $DebugLevel = 0;
my $NextDebugLevel= 2; # So Sigint can toggle this.	my $NextDebugLevel= 2; # So Sigint can toggle this.
my $IdleTimeout= 600; # Wait 10 minutes before pruning connections.	my $IdleTimeout= 7200; # Wait two hours before pruning connections.

my $LogTransactions = 0; # When True, all transactions/replies get logged.	my $LogTransactions = 0; # When True, all transactions/replies get logged.
	my $executable = $0; # Get the full path to me.

#	#
# 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 $RemoteHostId; # default lonid of host child is talking to.
	my @all_host_ids;
my $UnixSocketDir= $perlvar{'lonSockDir'};	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.
Line 102 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.


# DO NOT SET THE NEXT VARIABLE TO NON ZERO!!!!!!!!!!!!!!!

my $DieWhenIdle = 0; # When true children die when trimmed -> 0.
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 149 sub UpdateStatus {	Line 151 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'};
my $now=time;	my $now=time;
my $local=localtime($now);	my $local=localtime($now);
my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");	my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
	chomp($message);
print $fh "$now:$message:$local\n";	print $fh "$now:$message:$local\n";
}	}

Line 194 sub Log {	Line 198 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 268 sub SocketDump {	Line 272 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 278 sub ShowStatus {	Line 283 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 302 sub SocketTimeout {	Line 308 sub SocketTimeout {
}	}

}	}

#	#
# This function should be called by the child in all cases where it must	# This function should be called by the child in all cases where it must
# exit. If the child process is running with the DieWhenIdle turned on	# exit. The child process must create a lock file for the AF_UNIX socket
# it must create a lock file for the AF_UNIX socket in order to prevent	# in order to prevent connection requests from lonnet in the time between
# connection requests from lonnet in the time between process exit	# process exit and the parent picking up the listen again.
# and the parent picking up the listen again.	#
# Parameters:	# Parameters:
# exit_code - Exit status value, however see the next parameter.	# exit_code - Exit status value, however see the next parameter.
# message - If this optional parameter is supplied, the exit	# message - If this optional parameter is supplied, the exit
Line 318 sub child_exit {	Line 325 sub child_exit {

# Regardless of how we exit, we may need to do the lock thing:	# Regardless of how we exit, we may need to do the lock thing:

if($DieWhenIdle) {	#
#	# Create a lock file since there will be a time window
# Create a lock file since there will be a time window	# between our exit and the parent's picking up the listen
# between our exit and the parent's picking up the listen	# 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);	unlink(&GetLoncSocketPath());

exit(0);
}
# Now figure out how we exit:

if($message) {	if ($message) {
die $message;	die($message);
} else {	} else {
exit($exit_code);	exit($exit_code);
}	}
Line 369 sub Tick {	Line 372 sub Tick {
KillSocket($Socket);	KillSocket($Socket);
$IdleSeconds = 0; # Otherwise all connections get trimmed to fast.	$IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
UpdateStatus();	UpdateStatus();
if(($ConnectionCount == 0) && $DieWhenIdle) {	if(($ConnectionCount == 0)) {
&child_exit(0);	&child_exit(0);

}	}
Line 417 sub Tick {	Line 420 sub Tick {
$KeyMode = "";	$KeyMode = "";
$clock_watcher->cancel();	$clock_watcher->cancel();
}	}
	&UpdateStatus();
}	}

=pod	=pod
Line 499 the data and Event->w->fd is the socket	Line 503 the data and Event->w->fd is the socket
sub ClientWritable {	sub ClientWritable {
my $Event = shift;	my $Event = shift;
my $Watcher = $Event->w;	my $Watcher = $Event->w;
	if (!defined($Watcher)) {
	&child_exit(-1,'No watcher for event in ClientWritable');
	}
my $Data = $Watcher->data;	my $Data = $Watcher->data;
my $Socket = $Watcher->fd;	my $Socket = $Watcher->fd;

Line 562 sub ClientWritable {	Line 569 sub ClientWritable {
}	}
} else {	} else {
$Watcher->cancel(); # A delayed request...just cancel.	$Watcher->cancel(); # A delayed request...just cancel.
	return;
}	}
}	}

Line 601 sub CompleteTransaction {	Line 609 sub CompleteTransaction {
StartClientReply($Transaction, $data);	StartClientReply($Transaction, $data);
} else { # Delete deferred transaction file.	} else { # Delete deferred transaction file.
Log("SUCCESS", "A delayed transaction was completed");	Log("SUCCESS", "A delayed transaction was completed");
LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());	LogPerm("S:".$Transaction->getClient().":".$Transaction->getRequest());
unlink $Transaction->getFile();	unlink($Transaction->getFile());
}	}
}	}

Line 670 sub FailTransaction {	Line 678 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 682 sub FailTransaction {	Line 690 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 703 sub EmptyQueue {	Line 713 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{$Socket})) {
Line 711 sub CloseAllLondConnections {	Line 722 sub CloseAllLondConnections {
KillSocket($Socket);	KillSocket($Socket);
}	}
}	}
=cut

=pod	=pod

Line 733 Parameters:	Line 743 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 751 sub KillSocket {	Line 761 sub KillSocket {
delete ($ActiveTransactions{$Socket});	delete ($ActiveTransactions{$Socket});
}	}
if(exists($ActiveConnections{$Socket})) {	if(exists($ActiveConnections{$Socket})) {
	$ActiveConnections{$Socket}->cancel;
delete($ActiveConnections{$Socket});	delete($ActiveConnections{$Socket});
$ConnectionCount--;	$ConnectionCount--;
if ($ConnectionCount < 0) { $ConnectionCount = 0; }	if ($ConnectionCount < 0) { $ConnectionCount = 0; }
Line 762 sub KillSocket {	Line 773 sub KillSocket {
EmptyQueue();	EmptyQueue();
CloseAllLondConnections; # Should all already be closed but...	CloseAllLondConnections; # Should all already be closed but...
}	}
	UpdateStatus();
}	}

=pod	=pod
Line 946 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 1144 sub LondWritable {	Line 1156 sub LondWritable {
}	}

}	}

=pod	=pod

=cut	=cut


sub QueueDelayed {	sub QueueDelayed {
Debug(3,"QueueDelayed called");	Debug(3,"QueueDelayed called");

Line 1154 sub QueueDelayed {	Line 1169 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 $host_id_re = '(?:'.join('\|',map {quotemeta($_)} (@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 1190 sub MakeLondConnection {	Line 1205 sub MakeLondConnection {
.GetServerPort());	.GetServerPort());

my $Connection = LondConnection->new(&GetServerHost(),	my $Connection = LondConnection->new(&GetServerHost(),
&GetServerPort());	&GetServerPort(),
	&GetHostId());

if($Connection eq undef) { # Needs to be more robust later.	if($Connection eq undef) {
Log("CRITICAL","Failed to make a connection with lond.");	Log("CRITICAL","Failed to make a connection with lond.");
$ConnectionRetriesLeft--;	$ConnectionRetriesLeft--;
return 0; # Failure.	return 0; # Failure.
} else {	} else {

	$LondConnecting = 1; # Connection in progress.
# The connection needs to have writability	# The connection needs to have writability
# monitored in order to send the init sequence	# monitored in order to send the init sequence
# that starts the whole authentication/key	# that starts the whole authentication/key
Line 1228 sub MakeLondConnection {	Line 1245 sub MakeLondConnection {
}	}
Log("SUCESS", "Created connection ".$ConnectionCount	Log("SUCESS", "Created connection ".$ConnectionCount
." to host ".GetServerHost());	." to host ".GetServerHost());
$LondConnecting = 1; # Connection in progress.
return 1; # Return success.	return 1; # Return success.
}	}

Line 1369 sub ClientRequest {	Line 1385 sub ClientRequest {
$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($data eq "close_connection_exit\n") {	if ($data eq "close_connection_exit\n") {
Log("CRITICAL",	Log("CRITICAL",
"Request Close Connection ... exiting");	"Request Close Connection ... exiting");
CloseAllLondConnections();	CloseAllLondConnections();
exit;	exit;
	} elsif ($data eq "reset_retries\n") {
	Log("INFO", "Resetting Connection Retries.");
	$ConnectionRetriesLeft = $ConnectionRetries;
	&UpdateStatus();
	my $Transaction = LondTransaction->new($data);
	$Transaction->SetClient($socket);
	StartClientReply($Transaction, "ok\n");
	$watcher->cancel();
	return;
}	}
Debug(8, "Complete transaction received: ".$data);	Debug(8, "Complete transaction received: ".$data);
if($LogTransactions) {	if ($LogTransactions) {
Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.	Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
}	}
my $Transaction = LondTransaction->new($data);	my $Transaction = LondTransaction->new($data);
Line 1391 sub ClientRequest {	Line 1416 sub ClientRequest {
# Accept a connection request for a client (lonc child) and	# Accept a connection request for a client (lonc child) and
# start up an event watcher to keep an eye on input from that	# start up an event watcher to keep an eye on input from that
# Event. This can be called both from NewClient and from	# Event. This can be called both from NewClient and from
# ChildProcess if we are started in DieWhenIdle mode.	# ChildProcess.
# Parameters:	# Parameters:
# $socket - The listener socket.	# $socket - The listener socket.
# Returns:	# Returns:
Line 1480 sub GetServerHost {	Line 1505 sub GetServerHost {

=pod	=pod

	=head2 GetServerId

	Returns the hostid whose lond we talk with.

	=cut

	sub GetHostId {
	return $RemoteHostId; # Setup by the fork.
	}

	=pod

=head2 GetServerPort	=head2 GetServerPort

Returns the lond port number.	Returns the lond port number.
Line 1507 another event handler to subess requests	Line 1544 another event handler to subess requests
=cut	=cut

sub SetupLoncListener {	sub SetupLoncListener {
	my ($host,$SocketName) = @_;
	if (!$host) { $host = &GetServerHost(); }
	if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }

my $host = GetServerHost(); # Default host.
if (@_) {
($host) = @_ # Override host with parameter.
}

my $socket;
my $SocketName = GetLoncSocketPath($host);
unlink($SocketName);	unlink($SocketName);

	my $socket;
unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,	unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
Listen => 250,	Listen => 250,
Type => SOCK_STREAM)) {	Type => SOCK_STREAM)) {
Line 1555 into the status file.	Line 1591 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 1563 sub ChildStatus {	Line 1600 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 1603 sub SignalledToDeath {	Line 1645 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 1624 sub ToggleDebug {	Line 1668 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 1634 Optional parameter:	Line 1680 Optional parameter:
=cut	=cut

sub ChildProcess {	sub ChildProcess {
# If we are in DieWhenIdle mode, we've inherited all the	# We've inherited all the
# events of our parent and those have to be cancelled or else	# events of our parent and those have to be cancelled or else
# all holy bloody chaos will result.. trust me, I already made	# all holy bloody chaos will result.. trust me, I already made
# >that< mistake.	# >that< mistake.
Line 1650 sub ChildProcess {	Line 1696 sub ChildProcess {
Debug(5, "Killing watcher for $listener");	Debug(5, "Killing watcher for $listener");

$watcher->cancel();	$watcher->cancel();
undef $parent_dispatchers{$listener};	delete($parent_dispatchers{$listener});

}	}

	# kill off the parent's signal handlers too!
	#

	for my $handler (keys %parent_handlers) {
	my $watcher = $parent_handlers{$handler};
	$watcher->cancel();
	delete($parent_handlers{$handler});
	}

$I_am_child = 1; # Seems like in spite of it all I may still getting	$I_am_child = 1; # Seems like in spite of it all I may still getting
# parent event dispatches.. flag I'm a child.	# parent event dispatches.. flag I'm a child.

Line 1692 sub ChildProcess {	Line 1748 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 1700 sub ChildProcess {	Line 1756 sub ChildProcess {

# &MakeLondConnection(); // let first work request do it.	# &MakeLondConnection(); // let first work request do it.

# If We are in diwhenidle, need to accept the connection since the	# need to accept the connection since the event may not fire.
# event may not fire.

if ($DieWhenIdle) {	&accept_client($socket);
&accept_client($socket);
}

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 1717 sub ChildProcess {	Line 1770 sub ChildProcess {
# Create a new child for host passed in:	# Create a new child for host passed in:

sub CreateChild {	sub CreateChild {
my ($host, $socket) = @_;	my ($host, $hostid) = @_;

my $sigset = POSIX::SigSet->new(SIGINT);	my $sigset = POSIX::SigSet->new(SIGINT);
sigprocmask(SIG_BLOCK, $sigset);	sigprocmask(SIG_BLOCK, $sigset);
Line 1726 sub CreateChild {	Line 1779 sub CreateChild {
my $pid = fork;	my $pid = fork;
if($pid) { # Parent	if($pid) { # Parent
$RemoteHost = "Parent";	$RemoteHost = "Parent";
$ChildHash{$pid} = $host;	$ChildPid{$pid} = $host;
$HostToPid{$host}= $pid;
sigprocmask(SIG_UNBLOCK, $sigset);	sigprocmask(SIG_UNBLOCK, $sigset);
	undef(@all_host_ids);
} else { # child.	} else { # child.
	$RemoteHostId = $hostid;
ShowStatus("Connected to ".$RemoteHost);	ShowStatus("Connected to ".$RemoteHost);
$SIG{INT} = 'DEFAULT';	$SIG{INT} = 'DEFAULT';
sigprocmask(SIG_UNBLOCK, $sigset);	sigprocmask(SIG_UNBLOCK, $sigset);
if(defined $socket) {	&ChildProcess(); # Does not return.
&ChildProcess($socket);
} else {
ChildProcess; # Does not return.
}
}	}
}	}

Line 1766 sub parent_client_connection {	Line 1815 sub parent_client_connection {
my ($event) = @_;	my ($event) = @_;
my $watcher = $event->w;	my $watcher = $event->w;
my $socket = $watcher->fd;	my $socket = $watcher->fd;
	my $connection = $socket->accept(); # Accept the client connection.
# Lookup the host associated with this socket:	Event->io(cb => \&get_remote_hostname,
	poll => 'r',
my $host = $listening_to{$socket};	data => "",
	fd => $connection);
# 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();
}	}
}	}

	sub get_remote_hostname {
	my ($event) = @_;
	my $watcher = $event->w;
	my $socket = $watcher->fd;

	my $thisread;
	my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 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(5,"Creating child for $data (parent_client_connection)");
	(my $hostname,my $lonid,@all_host_ids) = split(':',$data);
	$ChildHost{$hostname}++;
	if ($ChildHost{$hostname} == 1) {
	&CreateChild($hostname,$lonid);
	} else {
	&Log('WARNING',"Request for a second child on $hostname");
	}
	# Clean up the listen since now the child takes over until it exits.
	$watcher->cancel(); # Nolonger listening to this event
	$socket->send("done\n");
	$socket->close();
	}

# parent_listen:	# parent_listen:
# Opens a socket and starts a listen for the parent process on a client UNIX	# Opens a socket and starts a listen for the parent process on a client UNIX
# domain socket.	# domain socket.
Line 1806 sub parent_listen {	Line 1880 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 ($socket,$file);
	if (!$loncapa_host) {
	$loncapa_host = 'common_parent';
	$file = $perlvar{'lonSockCreate'};
	} else {
	$file = &GetLoncSocketPath($loncapa_host);
	}
	$socket = &SetupLoncListener($loncapa_host,$file);

$listening_to{$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";
}	}

my $lock_file = &GetLoncSocketPath($loncapa_host).".lock";	my $lock_file = $file.".lock";
unlink($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(1, "parent_clean_up: $loncapa_host");

# listen_on_all_unix_sockets:	my $socket_file = &GetLoncSocketPath($loncapa_host);
# This sub initiates a listen on all unix domain lonc client sockets.	unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
# This will be called in the case where we are trimming idle processes.	my $lock_file = $socket_file.".lock";
# When idle processes are trimmed, loncnew starts up with no children,	unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
# and only spawns off children when a connection request occurs on the	}
# client unix socket. The spawned child continues to run until it has
# been idle a while at which point it eventually exits and once more
# the parent picks up the listen.
	# This sub initiates a listen on the common unix domain lonc client socket.
	# loncnew starts up with no children, and only spawns off children when a
	# connection request occurs on the common client unix socket. The spawned
	# child continues to run until it has been idle a while at which point it
	# eventually exits and once more the parent picks up the listen.
#	#
# Parameters:	# Parameters:
# NONE	# NONE
Line 1840 sub parent_listen {	Line 1932 sub parent_listen {
# Returns:	# Returns:
# NONE	# NONE
#	#
sub listen_on_all_unix_sockets {	sub listen_on_common_socket {
Debug(5, "listen_on_all_unix_sockets");	Debug(5, "listen_on_common_socket");
my $host_iterator = &LondConnection::GetHostIterator();	&parent_listen();
while (!$host_iterator->end()) {
my $host_entry_ref = $host_iterator->get();
my $host_name = $host_entry_ref->[0];
Debug(9, "Listen for $host_name");
&parent_listen($host_name);
$host_iterator->next();
}
}	}

# server_died is called whenever a child process exits.	# server_died is called whenever a child process exits.
Line 1871 sub server_died {	Line 1956 sub server_died {
}	}
# need the host to restart:	# need the host to restart:

my $host = $ChildHash{$pid};	my $host = $ChildPid{$pid};
if($host) { # It's for real...	if($host) { # It's for real...
&Debug(9, "Caught sigchild for $host");	&Debug(9, "Caught sigchild for $host");
delete($ChildHash{$pid});	delete($ChildPid{$pid});
delete($HostToPid{$host});	delete($ChildHost{$host});
&parent_listen($host);	&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 1933 ShowStatus("Forking node servers");	Line 2018 ShowStatus("Forking node servers");
Log("CRITICAL", "--------------- Starting children ---------------");	Log("CRITICAL", "--------------- Starting children ---------------");

LondConnection::ReadConfig; # Read standard config files.	LondConnection::ReadConfig; # Read standard config files.
my $HostIterator = LondConnection::GetHostIterator;

if ($DieWhenIdle) {	$RemoteHost = "[parent]";
$RemoteHost = "[parent]";	&listen_on_common_socket();
&listen_on_all_unix_sockets();
} else {

while (! $HostIterator->end()) {

my $hostentryref = $HostIterator->get();
CreateChild($hostentryref->[0]);
$HostHash{$hostentryref->[0]} = $hostentryref->[4];
$HostIterator->next();
}
}

$RemoteHost = "Parent Server";	$RemoteHost = "Parent Server";

Line 1956 $RemoteHost = "Parent Server";	Line 2029 $RemoteHost = "Parent Server";
ShowStatus("Parent keeping the flock");	ShowStatus("Parent keeping the flock");


if ($DieWhenIdle) {	# We need to setup a SIGChild event to handle the exit (natural or otherwise)
# We need to setup a SIGChild event to handle the exit (natural or otherwise)	# of the children.
# of the children.

Event->signal(cb => \&server_died,
desc => "Child exit handler",
signal => "CHLD");


$Event::DebugLevel = $DebugLevel;
Debug(9, "Parent entering event loop");
my $ret = Event::loop();
die "Main Event loop exited: $ret";


} else {
#
# Set up parent signals:
#

$SIG{INT} = \&Terminate;
$SIG{TERM} = \&Terminate;
$SIG{HUP} = \&Restart;
$SIG{USR1} = \&CheckKids;
$SIG{USR2} = \&UpdateKids; # LonManage update request.

while(1) {
my $deadchild = wait();
if(exists $ChildHash{$deadchild}) { # need to restart.
my $deadhost = $ChildHash{$deadchild};
delete($HostToPid{$deadhost});
delete($ChildHash{$deadchild});
Log("WARNING","Lost child pid= ".$deadchild.
"Connected to host ".$deadhost);
Log("INFO", "Restarting child procesing ".$deadhost);
CreateChild($deadhost);
}
}
}

	Event->signal(cb => \&server_died,
	desc => "Child exit handler",
	signal => "CHLD");


	# Set up all the other signals we set up.

	$parent_handlers{INT} = Event->signal(cb => \&Terminate,
	desc => "Parent INT handler",
	signal => "INT");
	$parent_handlers{TERM} = Event->signal(cb => \&Terminate,
	desc => "Parent TERM handler",
	signal => "TERM");
	$parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
	desc => "Parent HUP handler.",
	signal => "HUP");
	$parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
	desc => "Parent USR1 handler",
	signal => "USR1");
	$parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
	desc => "Parent USR2 handler.",
	signal => "USR2");

	# Start procdesing events.

	$Event::DebugLevel = $DebugLevel;
	Debug(9, "Parent entering event loop");
	my $ret = Event::loop();
	die "Main Event loop exited: $ret";

=pod	=pod

Line 2016 sub CheckKids {	Line 2081 sub CheckKids {
my $now=time;	my $now=time;
my $local=localtime($now);	my $local=localtime($now);
print $fh "LONC status $local - parent $$ \n\n";	print $fh "LONC status $local - parent $$ \n\n";
foreach my $pid (keys %ChildHash) {	foreach my $host (keys %parent_dispatchers) {
	print $fh "LONC Parent process listening for $host\n";
	}
	foreach my $pid (keys %ChildPid) {
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.
}	}

}	}

=pod	=pod
Line 2053 sub UpdateKids {	Line 2121 sub UpdateKids {

Log("INFO", "Updating connections via SIGUSR2");	Log("INFO", "Updating connections via SIGUSR2");

# Just in case we need to kill our own lonc, we wait a few seconds to	# I'm not sure what I was thinking in the first implementation.
# give it a chance to receive and relay lond's response to the	# someone will have to work hard to convince me the effect is any
# re-init command.	# different than Restart, especially now that we don't start up
#	# per host servers automatically, may as well just restart.
	# The down side is transactions that are in flight will get timed out
sleep(2); # Wait a couple of seconds.	# (lost unless they are critical).

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};	&KillThemAll();
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.
}
}
}	}


Line 2146 sub Restart {	Line 2147 sub Restart {
Log("CRITICAL", "Restarting");	Log("CRITICAL", "Restarting");
my $execdir = $perlvar{'lonDaemons'};	my $execdir = $perlvar{'lonDaemons'};
unlink("$execdir/logs/lonc.pid");	unlink("$execdir/logs/lonc.pid");
exec("$execdir/loncnew");	exec("$executable");
}	}

=pod	=pod
Line 2160 SIGHUP. Responds to sigint and sigterm.	Line 2161 SIGHUP. Responds to sigint and sigterm.

sub KillThemAll {	sub KillThemAll {
Debug(2, "Kill them all!!");	Debug(2, "Kill them all!!");
local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
foreach my $pid (keys %ChildHash) {	#local($SIG{CHLD}) = 'IGNORE';
my $serving = $ChildHash{$pid};	# Our children >will< die.
	# but we need to catch their death and cleanup after them in case this is
	# a restart set of kills
	my @allpids = keys(%ChildPid);
	foreach my $pid (@allpids) {
	my $serving = $ChildPid{$pid};
ShowStatus("Nicely Killing lonc for $serving pid = $pid");	ShowStatus("Nicely Killing lonc for $serving pid = $pid");
Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");	Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
kill 'QUIT' => $pid;	kill 'QUIT' => $pid;
}	}
	ShowStatus("Finished killing child processes off.");

}	}


Line 2180 sub really_kill_them_all_dammit	Line 2185 sub really_kill_them_all_dammit
{	{
Debug(2, "Kill them all Dammit");	Debug(2, "Kill them all Dammit");
local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.	local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
foreach my $pid (keys %ChildHash) {	foreach my $pid (keys %ChildPid) {
my $serving = $ChildHash{$pid};	my $serving = $ChildPid{$pid};
&ShowStatus("Nastily killing lonc for $serving pid = $pid");	&ShowStatus("Nastily killing lonc for $serving pid = $pid");
Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");	Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
kill 'KILL' => $pid;	kill 'KILL' => $pid;
delete($ChildHash{$pid});	delete($ChildPid{$pid});
my $execdir = $perlvar{'lonDaemons'};	my $execdir = $perlvar{'lonDaemons'};
unlink("$execdir/logs/lonc.pid");	unlink("$execdir/logs/lonc.pid");
}	}
}	}

=pod	=pod

=head1 Terminate	=head1 Terminate
Line 2212 sub Terminate {	Line 2218 sub Terminate {
exit 0;	exit 0;

}	}

	sub my_hostname {
	use Sys::Hostname;
	my $name = &hostname();
	&Debug(9,"Name is $name");
	return $name;
	}

=pod	=pod

=head1 Theory	=head1 Theory
Line 2253 A hash of lond connections that have no	Line 2267 A hash of lond connections that have no
can be closed if they are idle for a long enough time.	can be closed if they are idle for a long enough time.

=cut	=cut

	=pod

	=head1 Log messages

	The following is a list of log messages that can appear in the
	lonc.log file. Each log file has a severity and a message.

	=over 2

	=item Warning A socket timeout was detected

	If there are pending transactions in the socket's queue,
	they are failed (saved if critical). If the connection
	retry count gets exceeded by this, the
	remote host is marked as dead.
	Called when timeouts occured during the connection and
	connection dialog with a remote host.

	=item Critical Host makred DEAD <hostname>

	The numer of retry counts for contacting a host was
	exceeded. The host is marked dead an no
	further attempts will be made by that child.

	=item Info lonc pipe client hung up on us

	Write to the client pipe indicated no data transferred
	Socket to remote host is shut down. Reply to the client
	is discarded. Note: This is commented out in &ClientWriteable

	=item Success Reply from lond: <data>

	Can be enabled for debugging by setting LogTransactions to nonzero.
	Indicates a successful transaction with lond, <data> is the data received
	from the remote lond.

	=item Success A delayed transaction was completed

	A transaction that must be reliable was executed and completed
	as lonc restarted. This is followed by a mesage of the form

	S: client-name : request

	=item WARNING Failing transaction <cmd>:<subcmd>

	Transaction failed on a socket, but the failure retry count for the remote
	node has not yet been exhausted (the node is not yet marked dead).
	cmd is the command, subcmd is the subcommand. This results from a con_lost
	when communicating with lond.

	=item WARNING Shutting down a socket

	Called when a socket is being closed to lond. This is emitted both when
	idle pruning is being done and when the socket has been disconnected by the remote.

	=item WARNING Lond connection lost.

	Called when a read from lond's socket failed indicating lond has closed the
	connection or died. This should be followed by one or more

	"WARNING Failing transaction..." msgs for each in-flight or queued transaction.

	=item INFO Connected to lond version: <version>

	When connection negotiation is complete, the lond version is requested and logged here.

	=item SUCCESS Connection n to host now ready for action

	Emitted when connection has been completed with lond. n is then number of
	concurrent connections and host, the host to which the connection has just
	been established.

	=item WARNING Connection to host has been disconnected

	Write to a lond resulted in failure status. Connection to lond is dropped.

	=item SUCCESS Created connection n to host host

	Initial connection request to host..(before negotiation).

	=item CRITICAL Request Close Connection ... exiting

	Client has sent "close_connection_exit" The loncnew server is exiting.

	=item INFO Resetting Connection Retries

	Client has sent "reset_retries" The lond connection retries are reset to zero for the
	corresponding lond.

	=item SUCCESS Transaction <data>

	Only emitted if the global variable $LogTransactions was set to true.
	A client has requested a lond transaction <data> is the contents of the request.

	=item SUCCESS Toggled transaction logging <LogTransactions>

	The state of the $LogTransactions global has been toggled, and its current value
	(after being toggled) is displayed. When non zero additional logging of transactions
	is enabled for debugging purposes. Transaction logging is toggled on receipt of a USR2
	signal.

	=item CRITICAL Abnormal exit. Child <pid> for <host> died thorugh signal.

	QUIT signal received. lonc child process is exiting.

	=item SUCCESS New debugging level for <RemoteHost> now <DebugLevel>

	Debugging toggled for the host loncnew is talking with.
	Currently debugging is a level based scheme with higher number
	conveying more information. The daemon starts out at
	DebugLevel 0 and can toggle back and forth between that and
	DebugLevel 2 These are controlled by
	the global variables $DebugLevel and $NextDebugLevel
	The debug level can go up to 9.
	SIGINT toggles the debug level. The higher the debug level the
	more debugging information is spewed. See the Debug
	sub in loncnew.

	=item CRITICAL Forking server for host

	A child is being created to service requests for the specified host.


	=item WARNING Request for a second child on hostname

	Somehow loncnew was asked to start a second child on a host that already had a child
	servicing it. This request is not honored, but themessage is emitted. This could happen
	due to a race condition. When a client attempts to contact loncnew for a new host, a child
	is forked off to handle the requests for that server. The parent then backs off the Unix
	domain socket leaving it for the child to service all requests. If in the time between
	creating the child, and backing off, a new connection request comes in to the unix domain
	socket, this could trigger (unlikely but remotely possible),.

	=item CRITICAL ------ Starting Children ----

	This message should probably be changed to "Entering event loop" as the loncnew only starts
	children as needed. This message is emitted as new events are established and
	the event processing loop is entered.

	=item INFO Updating connections via SIGUSR2

	SIGUSR2 received. The original code would kill all clients, re-read the host file,
	then restart children for each host. Now that childrean aree started on demand, this
	just kills all child processes and lets requests start them as needed again.


	=item CRITICAL Restarting

	SigHUP received. all the children are killed and the script exec's itself to start again.

	=item CRITICAL Nicely killing lonc for host pid = <pid>

	Attempting to kill the child that is serving the specified host (pid given) cleanly via
	SIGQUIT The child should handle that, clean up nicely and exit.

	=item CRITICAL Nastily killing lonc for host pid = <pid>

	The child specified did not die when requested via SIGQUIT. Therefore it is killed
	via SIGKILL.

	=item CRITICAL Asked to kill children.. first be nice..

	In the parent's INT handler. INT kills the child processes. This inidicate loncnew
	is about to attempt to kill all known children via SIGQUIT. This message should be followed
	by one "Nicely killing" message for each extant child.

	=item CRITICAL Now kill children nasty

	In the parent's INT handler. remaining children are about to be killed via
	SIGKILL. Should be followed by a Nastily killing... for each lonc child that
	refused to die.

	=item CRITICAL Master process exiting

	In the parent's INT handler. just prior to the exit 0 call.

	=back

	=cut

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

Removed from v.1.64
changed lines
	Added in v.1.90