--- loncom/loncnew 2003/06/25 01:54:44 1.11
+++ loncom/loncnew 2004/05/11 19:55:35 1.45
@@ -2,13 +2,12 @@
# The LearningOnline Network with CAPA
# lonc maintains the connections to remote computers
#
-# $Id: loncnew,v 1.11 2003/06/25 01:54:44 foxr Exp $
+# $Id: loncnew,v 1.45 2004/05/11 19:55:35 albertel Exp $
#
# Copyright Michigan State University Board of Trustees
#
# 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
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
@@ -27,7 +26,7 @@
# 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.
# Development iterations:
# - Setup basic event loop. (done)
@@ -36,41 +35,19 @@
# - Add ability to create/negotiate lond connections (done).
# - Add general logic for dispatching requests and timeouts. (done).
# - Add support for the lonc/lond requests. (done).
-# - Add logging/status monitoring.
-# - Add Signal handling - HUP restarts. USR1 status report.
+# - Add logging/status monitoring. (done)
+# - Add Signal handling - HUP restarts. USR1 status report. (done)
# - Add Configuration file I/O (done).
-# - Add management/status request interface.
+# - Add management/status request interface. (done)
# - Add deferred request capability. (done)
-# - Detect transmission timeouts.
-#
-
-# Change log:
-# $Log: loncnew,v $
-# 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 6the 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.
-#
+# - Detect transmission timeouts. (done)
#
+use strict;
use lib "/home/httpd/lib/perl/";
-use lib "/home/foxr/newloncapa/types";
use Event qw(:DEFAULT );
use POSIX qw(:signal_h);
+use POSIX;
use IO::Socket;
use IO::Socket::INET;
use IO::Socket::UNIX;
@@ -86,17 +63,6 @@ use LONCAPA::Configuration;
use LONCAPA::HashIterator;
-#
-# 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
# normally set in apache modules:
@@ -107,19 +73,24 @@ my %perlvar = %{$perlvarref};
# parent and shared variables.
my %ChildHash; # by pid -> host.
+my %HostToPid; # By host -> pid.
+my %HostHash; # by loncapaname -> IP.
my $MaxConnectionCount = 10; # Will get from config later.
my $ClientConnection = 0; # Uniquifier for client events.
my $DebugLevel = 0;
+my $NextDebugLevel= 2; # So Sigint can toggle this.
my $IdleTimeout= 3600; # Wait an hour before pruning connections.
+my $LogTransactions = 0; # When True, all transactions/replies get logged.
+
#
# The variables below are only used by the child processes.
#
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 %ActiveConnections; # Connections to the remote lond.
my %ActiveTransactions; # LondTransactions in flight.
@@ -128,8 +99,10 @@ my $WorkQueue = Queue->new(); # Qu
my $ConnectionCount = 0;
my $IdleSeconds = 0; # Number of seconds idle.
my $Status = ""; # Current status string.
-my $ConnectionRetries=5; # Number of connection retries allowed.
-my $ConnectionRetriesLeft=5; # Number of connection retries remaining.
+my $RecentLogEntry = "";
+my $ConnectionRetries=2; # Number of connection retries allowed.
+my $ConnectionRetriesLeft=2; # Number of connection retries remaining.
+my $LondVersion = "unknown"; # Version of lond we talk with.
#
# The hash below gives the HTML format for log messages
@@ -137,10 +110,10 @@ my $ConnectionRetriesLeft=5; # Number of
#
my %LogFormats;
-$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{"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 ";
@@ -204,6 +177,7 @@ sub Log {
my $execdir = $perlvar{'lonDaemons'};
my $fh = IO::File->new(">>$execdir/logs/lonc.log");
my $msg = sprintf($finalformat, $message);
+ $RecentLogEntry = $msg;
print $fh $msg;
@@ -226,7 +200,7 @@ sub GetPeername {
my $peerip;
if($AdrFamily == AF_INET) {
($peerport, $peerip) = sockaddr_in($peer);
- my $peername = gethostbyaddr($iaddr, $AdrFamily);
+ my $peername = gethostbyaddr($peerip, $AdrFamily);
return $peername;
} elsif ($AdrFamily == AF_UNIX) {
my $peerfile;
@@ -234,7 +208,6 @@ sub GetPeername {
return $peerfile;
}
}
-#----------------------------- Timer management ------------------------
=pod
=head2 Debug
@@ -247,7 +220,7 @@ sub Debug {
my $level = shift;
my $message = shift;
if ($level <= $DebugLevel) {
- print $message." host = ".$RemoteHost."\n";
+ Log("INFO", "-Debug- $message host = $RemoteHost");
}
}
@@ -277,6 +250,33 @@ sub ShowStatus {
=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;
+ Log("WARNING", "A socket timeout was detected");
+ Debug(0, " SocketTimeout called: ");
+ $Socket->Dump();
+ if(exists($ActiveTransactions{$Socket})) {
+ FailTransaction($ActiveTransactions{$Socket});
+ }
+ KillSocket($Socket); # A transaction timeout also counts as
+ # a connection failure:
+ $ConnectionRetriesLeft--;
+ if($ConnectionRetriesLeft <= 0) {
+ Log("CRITICAL", "Host marked dead: ".GetServerHost());
+ }
+
+}
+#----------------------------- Timer management ------------------------
+
+=pod
+
=head2 Tick
Invoked each timer tick.
@@ -286,11 +286,11 @@ Invoked each timer tick.
sub Tick {
my $client;
- ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount);
- Debug(10,"Tick");
- Debug(10," Current connection count: ".$ConnectionCount);
- foreach $client (keys %ActiveClients) {
- Debug(10," Have client: with id: ".$ActiveClients{$client});
+ if($ConnectionRetriesLeft > 0) {
+ ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
+ ." Retries remaining: ".$ConnectionRetriesLeft);
+ } else {
+ ShowStatus(GetServerHost()." >> DEAD <<");
}
# Is it time to prune connection count:
@@ -299,13 +299,23 @@ sub Tick {
($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
$IdleSeconds++;
if($IdleSeconds > $IdleTimeout) { # Prune a connection...
- $Socket = $IdleConnections->pop();
+ 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 %ActiveConnections) {
+ my $State = $ActiveConnections{$item}->data->GetState();
+ if ($State ne 'Idle') {
+ Debug(5,"Ticking Socket $State $item");
+ $ActiveConnections{$item}->data->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.
#
@@ -315,13 +325,21 @@ sub Tick {
if ($ConnectionRetriesLeft > 0) {
my $Connections = ($Requests <= $MaxConnectionCount) ?
$Requests : $MaxConnectionCount;
- Debug(1,"Work but no connections, start ".$Connections." of them");
- for ($i =0; $i < $Connections; $i++) {
- MakeLondConnection();
+ Debug(5,"Work but no connections, start ".$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.
+ CloseAllLondConnections(); # Should all be closed but....
}
} else {
- Debug(1,"Work in queue, but gave up on connections..flushing\n");
+ ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
+ Debug(5,"Work in queue, but gave up on connections..flushing\n");
EmptyQueue(); # Connections can't be established.
+ CloseAllLondConnections(); # Should all already be closed but...
}
}
@@ -345,7 +363,7 @@ Trigger disconnections of idle sockets.
sub SetupTimer {
Debug(6, "SetupTimer");
- Event->timer(interval => 1, debug => 1, cb => \&Tick );
+ Event->timer(interval => 1, cb => \&Tick );
}
=pod
@@ -367,19 +385,19 @@ sub ServerToIdle {
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:
- $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
- unless($reqdata eq undef) {
- Debug(9, "Queue gave request data: ".$reqdata->getRequest());
+ my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
+ if ($reqdata ne undef) {
+ Debug(5, "Queue gave request data: ".$reqdata->getRequest());
&StartRequest($Socket, $reqdata);
} else {
# 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");
$IdleConnections->push($Socket);
}
}
@@ -425,7 +443,7 @@ sub ClientWritable {
# request.
&Debug(9,"Send result is ".$result." Defined: ".defined($result));
- if(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");
@@ -443,6 +461,12 @@ sub ClientWritable {
} 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...
@@ -490,20 +514,25 @@ The transaction that is being completed.
=cut
sub CompleteTransaction {
- &Debug(6,"Complete transaction");
+ &Debug(5,"Complete transaction");
my $Socket = shift;
my $Transaction = shift;
if (!$Transaction->isDeferred()) { # Normal transaction
my $data = $Socket->GetReply(); # Data to send.
+ if($LogTransactions) {
+ Log("SUCCESS", "Reply from lond: '$data'");
+ }
StartClientReply($Transaction, $data);
} else { # Delete deferred transaction file.
Log("SUCCESS", "A delayed transaction was completed");
- LogPerm("S:$Client:".$Transaction->getRequest());
+ LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
unlink $Transaction->getFile();
}
}
+
=pod
+
=head1 StartClientReply
Initiates a reply to a client where the reply data is a parameter.
@@ -519,16 +548,17 @@ sub CompleteTransaction {
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);
my $Serial = $ActiveClients{$Client};
my $desc = sprintf("Connection to lonc client %d",
-
$Serial);
Event->io(fd => $Client,
poll => "w",
@@ -536,7 +566,9 @@ sub StartClientReply {
cb => \&ClientWritable,
data => $data);
}
+
=pod
+
=head2 FailTransaction
Finishes a transaction with failure because the associated lond socket
@@ -546,8 +578,7 @@ sub StartClientReply {
- 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.
+Deleting the transaction means killing it from the %ActiveTransactions hash.
Parameters:
@@ -555,10 +586,12 @@ Parameters:
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();
@@ -576,6 +609,7 @@ sub FailTransaction {
=cut
sub EmptyQueue {
+ $ConnectionRetriesLeft--; # Counts as connection failure too.
while($WorkQueue->Count()) {
my $request = $WorkQueue->dequeue(); # This is a transaction
FailTransaction($request);
@@ -590,8 +624,11 @@ Close all connections open on lond prior
=cut
sub CloseAllLondConnections {
- foreach $Socket (keys %ActiveConnections) {
- KillSocket($Socket);
+ foreach my $Socket (keys %ActiveConnections) {
+ if(exists($ActiveTransactions{$Socket})) {
+ FailTransaction($ActiveTransactions{$Socket});
+ }
+ KillSocket($Socket);
}
}
=cut
@@ -621,6 +658,7 @@ nonzero if we are allowed to create a ne
sub KillSocket {
my $Socket = shift;
+ Log("WARNING", "Shutting down a socket");
$Socket->Shutdown();
# If the socket came from the active connection set,
@@ -634,14 +672,15 @@ sub KillSocket {
}
if(exists($ActiveConnections{$Socket})) {
delete($ActiveConnections{$Socket});
+ $ConnectionCount--;
+ if ($ConnectionCount < 0) { $ConnectionCount = 0; }
}
- $ConnectionCount--;
-
# If the connection count has gone to zero and there is work in the
# work queue, the work all gets failed with con_lost.
#
if($ConnectionCount == 0) {
- EmptyQueue;
+ EmptyQueue();
+ CloseAllLondConnections; # Should all already be closed but...
}
}
@@ -667,6 +706,17 @@ The connection must echo the challenge b
The challenge has been replied to. The we are receiveing the
'ok' from the partner.
+=head3 State=ReadingVersionString
+
+We have requested the lond version and are reading the
+version back. Upon completion, we'll store the version away
+for future use(?).
+
+=head3 State=HostSet
+
+We have selected the domain name of our peer (multhomed hosts)
+and are getting the reply (presumably ok) back.
+
=head3 State=RequestingKey
The ok has been received and we need to send the request for
@@ -710,23 +760,28 @@ sub LondReadable {
my $Socket = $Watcher->data;
my $client = undef;
- &Debug(6,"LondReadable called state = ".$State);
+ &Debug(6,"LondReadable called state = ".$Socket->GetState());
my $State = $Socket->GetState(); # All action depends on the state.
SocketDump(6, $Socket);
+ my $status = $Socket->Readable();
+
+ &Debug(2, "Socket->Readable returned: $status");
- if($Socket->Readable() != 0) {
- # bad return from socket read. Currently this means that
+ 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{$Socket})) {
- Debug(3,"Lond connection lost failing transaction");
FailTransaction($ActiveTransactions{$Socket});
}
$Watcher->cancel();
KillSocket($Socket);
+ $ConnectionRetriesLeft--; # Counts as connection failure
return;
}
SocketDump(6,$Socket);
@@ -734,19 +789,34 @@ sub LondReadable {
$State = $Socket->GetState(); # Update in case of transition.
&Debug(6, "After read, state is ".$State);
- if($State eq "Initialized") {
+ if($State eq "Initialized") {
} elsif ($State eq "ChallengeReceived") {
# The challenge must be echoed back; The state machine
# in the connection takes care of setting that up. Just
# need to transition to writable:
-
+
$Watcher->cb(\&LondWritable);
$Watcher->poll("w");
} elsif ($State eq "ChallengeReplied") {
+ } elsif ($State eq "RequestingVersion") {
+ # Need to ask for the version... that is writiability:
+
+ $Watcher->cb(\&LondWritable);
+ $Watcher->poll("w");
+
+ } elsif ($State eq "ReadingVersionString") {
+ # Read the rest of the version string...
+ } elsif ($State eq "SetHost") {
+ # Need to request the actual domain get set...
+
+ $Watcher->cb(\&LondWritable);
+ $Watcher->poll("w");
+ } elsif ($State eq "HostSet") {
+ # Reading the 'ok' from the peer.
} elsif ($State eq "RequestingKey") {
# The ok was received. Now we need to request the key
@@ -758,11 +828,23 @@ sub LondReadable {
} elsif ($State eq "ReceivingKey") {
} elsif ($State eq "Idle") {
+
+ # This is as good a spot as any to get the peer version
+ # string:
+
+ if($LondVersion eq "unknown") {
+ $LondVersion = $Socket->PeerVersion();
+ Log("INFO", "Connected to lond version: $LondVersion");
+ }
# If necessary, complete a transaction and then go into the
# 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{$Socket})) {
- Debug(8,"Completing transaction!!");
+ Debug(5,"Completing transaction!!");
CompleteTransaction($Socket,
$ActiveTransactions{$Socket});
} else {
@@ -783,7 +865,7 @@ sub LondReadable {
} else {
- # Invalid state.
+ # Invalid state.
Debug(4, "Invalid state in LondReadable");
}
}
@@ -865,20 +947,36 @@ sub LondWritable {
SocketDump(6,$Socket);
+ # If the socket is writable, we must always write.
+ # Only by writing will we undergo state transitions.
+ # Old logic wrote in state specific code below, however
+ # That forces us at least through another invocation of
+ # this function after writability is possible again.
+ # This logic also factors out common code for handling
+ # write failures... in all cases, write failures
+ # Kill the socket.
+ # This logic makes the branches of the >big< if below
+ # so that the writing states are actually NO-OPs.
+
+ if ($Socket->Writable() != 0) {
+ # The write resulted in an error.
+ # We'll treat this as if the socket got disconnected:
+ Log("WARNING", "Connection to ".$RemoteHost.
+ " has been disconnected");
+ if(exists($ActiveTransactions{$Socket})) {
+ FailTransaction($ActiveTransactions{$Socket});
+ }
+ $Watcher->cancel();
+ KillSocket($Socket);
+ return;
+ }
+
+
+
if ($State eq "Connected") {
- if ($Socket->Writable() != 0) {
- # The write resulted in an error.
- # We'll treat this as if the socket got disconnected:
- Log("WARNING", "Connection to ".$RemoteHost.
- " has been disconnected");
- $Watcher->cancel();
- KillSocket($Socket);
- return;
- }
# "init" is being sent...
-
-
+
} elsif ($State eq "Initialized") {
# Now that init was sent, we switch
@@ -886,40 +984,43 @@ sub LondWritable {
$Watcher->cb(\&LondReadable);
$Watcher->poll("r");
-
+
} elsif ($State eq "ChallengeReceived") {
# We received the challenge, now we
# are echoing it back. This is a no-op,
# we're waiting for the state to change
- if($Socket->Writable() != 0) {
-
- $Watcher->cancel();
- KillSocket($Socket);
- return;
- }
-
} elsif ($State eq "ChallengeReplied") {
# The echo was sent back, so we switch
# to watching readability.
$Watcher->cb(\&LondReadable);
$Watcher->poll("r");
+ } elsif ($State eq "RequestingVersion") {
+ # Sending the peer a version request...
+
+ } elsif ($State eq "ReadingVersionString") {
+ # Transition to read since we have sent the
+ # version command and now just need to read the
+ # version string from the peer:
+
+ $Watcher->cb(\&LondReadable);
+ $Watcher->poll("r");
+
+ } elsif ($State eq "SetHost") {
+ # Setting the remote domain...
+
+ } elsif ($State eq "HostSet") {
+ # Back to readable to get the ok.
+
+ $Watcher->cb(\&LondReadable);
+ $Watcher->poll("r");
+
} elsif ($State eq "RequestingKey") {
# At this time we're requesting the key.
# again, this is essentially a no-op.
- # we'll write the next chunk until the
- # state changes.
-
- if($Socket->Writable() != 0) {
- # Write resulted in an error.
- $Watcher->cancel();
- KillSocket($Socket);
- return;
-
- }
} elsif ($State eq "ReceivingKey") {
# Now we need to wait for the key
# to come back from the peer:
@@ -928,20 +1029,10 @@ sub LondWritable {
$Watcher->poll("r");
} elsif ($State eq "SendingRequest") {
+
# At this time we are sending a request to the
# peer... write the next chunk:
- if($Socket->Writable() != 0) {
-
- if(exists($ActiveTransactions{$Socket})) {
- Debug(3, "Lond connection lost, failing transactions");
- FailTransaction($ActiveTransactions{$Socket});
- }
- $Watcher->cancel();
- KillSocket($Socket);
- return;
-
- }
} elsif ($State eq "ReceivingReply") {
# The send has completed. Wait for the
@@ -971,8 +1062,7 @@ sub QueueDelayed {
Debug(4, "Delayed path: ".$path);
opendir(DIRHANDLE, $path);
- @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
- Debug(4, "Got ".$alldelayed." delayed files");
+ my @alldelayed = grep /\.$RemoteHost$/, readdir DIRHANDLE;
closedir(DIRHANDLE);
my $dfname;
my $reqfile;
@@ -982,7 +1072,7 @@ sub QueueDelayed {
my $Handle = IO::File->new($reqfile);
my $cmd = <$Handle>;
chomp $cmd; # There may or may not be a newline...
- $cmd = $cmd."\ny"; # now for sure there's exactly one newline.
+ $cmd = $cmd."\n"; # now for sure there's exactly one newline.
my $Transaction = LondTransaction->new($cmd);
$Transaction->SetDeferred($reqfile);
QueueTransaction($Transaction);
@@ -1009,26 +1099,27 @@ sub MakeLondConnection {
my $Connection = LondConnection->new(&GetServerHost(),
&GetServerPort());
- if($Connection == undef) { # Needs to be more robust later.
+ if($Connection eq undef) { # Needs to be more robust later.
Log("CRITICAL","Failed to make a connection with lond.");
$ConnectionRetriesLeft--;
return 0; # Failure.
} else {
- $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
+
# 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) {
+ if($Socket eq undef) {
die "did not get a socket from the connection";
} else {
&Debug(9,"MakeLondConnection got socket: ".$Socket);
}
-
- $event = Event->io(fd => $Socket,
+ $Connection->SetTimeoutCallback(\&SocketTimeout);
+
+ my $event = Event->io(fd => $Socket,
poll => 'w',
cb => \&LondWritable,
data => $Connection,
@@ -1084,7 +1175,7 @@ sub StartRequest {
$ActiveTransactions{$Lond} = $Request;
$Lond->InitiateTransaction($Request->getRequest());
- $event = Event->io(fd => $Socket,
+ my $event = Event->io(fd => $Socket,
poll => "w",
cb => \&LondWritable,
data => $Lond,
@@ -1123,11 +1214,20 @@ sub QueueTransaction {
my $LondSocket = $IdleConnections->pop();
if(!defined $LondSocket) { # Need to queue request.
- Debug(8,"Must queue...");
+ Debug(5,"Must queue...");
$WorkQueue->enqueue($requestData);
if($ConnectionCount < $MaxConnectionCount) {
- Debug(4,"Starting additional lond connection");
- MakeLondConnection();
+ if($ConnectionRetriesLeft > 0) {
+ Debug(5,"Starting additional lond connection");
+ if(MakeLondConnection() == 0) {
+ EmptyQueue(); # Fail transactions, can't make connection.
+ CloseAllLondConnections; # Should all be closed but...
+ }
+ } else {
+ ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
+ EmptyQueue(); # It's worse than that ... he's dead Jim.
+ CloseAllLondConnections; # Should all be closed but..
+ }
}
} else { # Can start the request:
Debug(8,"Can start...");
@@ -1158,7 +1258,7 @@ sub ClientRequest {
my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
Debug(8, "rcv: data length = ".length($thisread)
." read =".$thisread);
- unless (defined $rv && length($thisread)) {
+ unless (defined $rv && length($thisread)) {
# Likely eof on socket.
Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
close($socket);
@@ -1169,7 +1269,7 @@ sub ClientRequest {
Debug(8,"Data: ".$data." this read: ".$thisread);
$data = $data.$thisread; # Append new 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");
@@ -1177,6 +1277,9 @@ sub ClientRequest {
exit;
}
Debug(8, "Complete transaction received: ".$data);
+ if($LogTransactions) {
+ Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
+ }
my $Transaction = LondTransaction->new($data);
$Transaction->SetClient($socket);
QueueTransaction($Transaction);
@@ -1285,6 +1388,60 @@ sub SetupLoncListener {
fd => $socket);
}
+#
+# Toggle transaction logging.
+# Implicit inputs:
+# LogTransactions
+# Implicit Outputs:
+# LogTransactions
+sub ToggleTransactionLogging {
+ print STDERR "Toggle transaction logging...\n";
+ if(!$LogTransactions) {
+ $LogTransactions = 1;
+ } else {
+ $LogTransactions = 0;
+ }
+
+
+ Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
+}
+
+=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 $$."\t".$RemoteHost."\t".$Status."\t".
+ $RecentLogEntry."\n";
+ #
+ # Write out information about each of the connections:
+ #
+ print $fh "Active connection statuses: \n";
+ my $i = 1;
+ print STDERR "================================= Socket Status Dump:\n";
+ foreach my $item (keys %ActiveConnections) {
+ my $Socket = $ActiveConnections{$item}->data;
+ my $state = $Socket->GetState();
+ print $fh "Connection $i State: $state\n";
+ print STDERR "---------------------- Connection $i \n";
+ $Socket->Dump();
+ $i++;
+ }
+ $ConnectionRetriesLeft = $ConnectionRetries;
+}
+
=pod
=head2 SignalledToDeath
@@ -1293,18 +1450,37 @@ Called in response to a signal that caus
=cut
-=pod
sub SignalledToDeath {
- my ($signal) = @_;
+ my $event = shift;
+ my $watcher= $event->w;
+
+ Debug(2,"Signalled to death! via ".$watcher->data);
+ my ($signal) = $watcher->data;
chomp($signal);
Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
."died through "."\"$signal\"");
LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
."\"$signal\"");
- die("Signal abnormal end");
+ 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
This sub implements a child process for a single lonc daemon.
@@ -1314,14 +1490,24 @@ This sub implements a child process for
sub ChildProcess {
- # For now turn off signals.
-
- $SIG{QUIT} = \&SignalledToDeath;
- $SIG{HUP} = IGNORE;
- $SIG{USR1} = IGNORE;
- $SIG{INT} = IGNORE;
- $SIG{CHLD} = IGNORE;
- $SIG{__DIE__} = \&SignalledToDeath;
+ #
+ # Signals must be handled by the Event framework...
+#
+
+ Event->signal(signal => "QUIT",
+ cb => \&SignalledToDeath,
+ data => "QUIT");
+ Event->signal(signal => "HUP",
+ cb => \&ChildStatus,
+ data => "HUP");
+ Event->signal(signal => "USR1",
+ cb => \&ChildStatus,
+ data => "USR1");
+ Event->signal(signal => "USR2",
+ cb => \&ToggleTransactionLogging);
+ Event->signal(signal => "INT",
+ cb => \&ToggleDebug,
+ data => "INT");
SetupTimer();
@@ -1333,7 +1519,7 @@ sub ChildProcess {
# Setup the initial server connection:
- # &MakeLondConnection(); // let first work requirest do it.
+ # &MakeLondConnection(); // let first work requirest do� it.
Debug(9,"Entering event loop");
@@ -1346,15 +1532,23 @@ sub ChildProcess {
# Create a new child for host passed in:
sub CreateChild {
+ my $sigset = POSIX::SigSet->new(SIGINT);
+ sigprocmask(SIG_BLOCK, $sigset);
my $host = shift;
$RemoteHost = $host;
Log("CRITICAL", "Forking server for ".$host);
- $pid = fork;
+ my $pid = fork;
if($pid) { # Parent
- $ChildHash{$pid} = $RemoteHost;
+ $RemoteHost = "Parent";
+ $ChildHash{$pid} = $host;
+ $HostToPid{$host}= $pid;
+ sigprocmask(SIG_UNBLOCK, $sigset);
+
} else { # child.
ShowStatus("Connected to ".$RemoteHost);
- ChildProcess;
+ $SIG{INT} = 'DEFAULT';
+ sigprocmask(SIG_UNBLOCK, $sigset);
+ ChildProcess; # Does not return.
}
}
@@ -1389,11 +1583,13 @@ if ($childpid != 0) {
#
ShowStatus("Parent writing pid file:");
-$execdir = $perlvar{'lonDaemons'};
+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;
@@ -1403,13 +1599,16 @@ ShowStatus("Forking node servers");
Log("CRITICAL", "--------------- Starting children ---------------");
+LondConnection::ReadConfig; # Read standard config files.
my $HostIterator = LondConnection::GetHostIterator;
while (! $HostIterator->end()) {
- $hostentryref = $HostIterator->get();
+ my $hostentryref = $HostIterator->get();
CreateChild($hostentryref->[0]);
+ $HostHash{$hostentryref->[0]} = $hostentryref->[4];
$HostIterator->next();
}
+$RemoteHost = "Parent Server";
# Maintain the population:
@@ -1418,13 +1617,18 @@ ShowStatus("Parent keeping the flock");
#
# Set up parent signals:
#
-$SIG{INT} = &KillThemAll;
-$SIG{TERM} = &KillThemAll;
+
+$SIG{INT} = \&Terminate;
+$SIG{TERM} = \&Terminate;
+$SIG{HUP} = \&Restart;
+$SIG{USR1} = \&CheckKids;
+$SIG{USR2} = \&UpdateKids; # LonManage update request.
while(1) {
- $deadchild = wait();
+ my $deadchild = wait();
if(exists $ChildHash{$deadchild}) { # need to restart.
- $deadhost = $ChildHash{$deadchild};
+ my $deadhost = $ChildHash{$deadchild};
+ delete($HostToPid{$deadhost});
delete($ChildHash{$deadchild});
Log("WARNING","Lost child pid= ".$deadchild.
"Connected to host ".$deadhost);
@@ -1432,8 +1636,201 @@ while(1) {
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