--- loncom/loncnew 2003/09/16 09:46:42 1.24 +++ loncom/loncnew 2005/03/15 01:12:20 1.68 @@ -2,7 +2,7 @@ # The LearningOnline Network with CAPA # lonc maintains the connections to remote computers # -# $Id: loncnew,v 1.24 2003/09/16 09:46:42 foxr Exp $ +# $Id: loncnew,v 1.68 2005/03/15 01:12:20 albertel Exp $ # # Copyright Michigan State University Board of Trustees # @@ -35,94 +35,16 @@ # - 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. +# - Detect transmission timeouts. (done) # -# Change log: -# $Log: loncnew,v $ -# Revision 1.24 2003/09/16 09:46:42 foxr -# Added skeletal infrastructure to support SIGUSR2 update hosts request. -# -# Revision 1.23 2003/09/15 09:24:49 foxr -# Add use strict and fix all the fallout from that. -# -# Revision 1.22 2003/09/02 10:34:47 foxr -# - Fix errors in host dead detection logic (too many cases where the -# retries left were not getting incremented or just not checked). -# - Added some additional status to the ps axuww display: -# o Remaining retries on a host. -# o >>> DEAD <<< indicator if I've given up on a host. -# - Tested the SIGHUP will reset the retries remaining count (thanks to -# the above status stuff, and get allow the loncnew to re-try again -# on the host (thanks to the log). -# -# Revision 1.21 2003/08/26 09:19:51 foxr -# How embarrassing... put in the SocketTimeout function in loncnew and forgot -# to actually hook it into the LondTransaction. Added this to MakeLondConnection -# where it belongs... hopefully transactions (not just connection attempts) will -# timeout more speedily than the socket errors will catch it. -# -# Revision 1.20 2003/08/25 18:48:11 albertel -# - fixing a forgotten ; -# -# Revision 1.19 2003/08/19 09:31:46 foxr -# Get socket directory from configuration rather than the old hard coded test -# way that I forgot to un-hard code. -# -# Revision 1.18 2003/08/06 09:52:29 foxr -# Also needed to remember to fail in-flight transactions if their sends fail. -# -# Revision 1.17 2003/08/03 00:44:31 foxr -# 1. Correct handling of connection failure: Assume it means the host is -# unreachable and fail all of the queued transactions. Note that the -# inflight transactions should fail on their own time due either to timeout -# or send/receive failures. -# 2. Correct handling of logs for forced death signals. Pull the signal -# from the event watcher. -# -# Revision 1.16 2003/07/29 02:33:05 foxr -# Add SIGINT processing to child processes to toggle annoying trace mode -# on/off.. will try to use this to isolate the compute boud process issue. -# -# Revision 1.15 2003/07/15 02:07:05 foxr -# Added code for lonc/lond transaction timeouts. Who knows if it works right. -# The intent is for a timeout to fail any transaction in progress and kill -# off the sockt that timed out. -# -# Revision 1.14 2003/07/03 02:10:18 foxr -# Get all of the signals to work correctly. -# -# Revision 1.13 2003/07/02 01:31:55 foxr -# Added kill -HUP logic (restart). -# -# Revision 1.11 2003/06/25 01:54:44 foxr -# Fix more problems with transaction failure. -# -# Revision 1.10 2003/06/24 02:46:04 foxr -# Put a limit on the number of times we'll retry a connection. -# Start getting the signal stuff put in as well...note that need to get signals -# going or else the client will permanently give up on dead servers. -# -# Revision 1.9 2003/06/13 02:38:43 foxr -# Add logging in 'expected format' -# -# Revision 1.8 2003/06/11 02:04:35 foxr -# Support delayed transactions... this is done uniformly by encapsulating -# transactions in an object ... a LondTransaction that is implemented by -# LondTransaction.pm -# -# Revision 1.7 2003/06/03 01:59:39 foxr -# complete coding to support deferred transactions. -# -# use strict; use lib "/home/httpd/lib/perl/"; -use lib "/home/foxr/newloncapa/types"; use Event qw(:DEFAULT ); use POSIX qw(:signal_h); use POSIX; @@ -139,17 +61,7 @@ use LONCAPA::LondConnection; use LONCAPA::LondTransaction; 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; +use Fcntl qw(:flock); # Read the httpd configuration file to get perl variables @@ -162,14 +74,23 @@ my %perlvar = %{$perlvarref}; # parent and shared variables. my %ChildHash; # by pid -> host. +my %HostToPid; # By host -> pid. +my %HostHash; # by loncapaname -> IP. +my %listening_to; # Socket->host table for who the parent + # is listening to. +my %parent_dispatchers; # host-> listener watcher events. +my %parent_handlers; # Parent signal handlers... my $MaxConnectionCount = 10; # Will get from config later. my $ClientConnection = 0; # Uniquifier for client events. my $DebugLevel = 0; -my $NextDebugLevel= 10; # So Sigint can toggle this. -my $IdleTimeout= 3600; # Wait an hour before pruning connections. +my $NextDebugLevel= 2; # So Sigint can toggle this. +my $IdleTimeout= 600; # Wait 10 minutes before pruning connections. + +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. @@ -185,8 +106,16 @@ my $ConnectionCount = 0; my $IdleSeconds = 0; # Number of seconds idle. my $Status = ""; # Current status string. my $RecentLogEntry = ""; -my $ConnectionRetries=5; # Number of connection retries allowed. -my $ConnectionRetriesLeft=5; # Number of connection retries remaining. +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. +my $KeyMode = ""; # e.g. ssl, local, insecure from last connect. +my $LondConnecting = 0; # True when a connection is being built. + + + +my $DieWhenIdle = 1; # When true children die when trimmed -> 0. +my $I_am_child = 0; # True if this is the child process. # # The hash below gives the HTML format for log messages @@ -194,13 +123,27 @@ my $ConnectionRetriesLeft=5; # Number of # my %LogFormats; -$LogFormats{"CRITICAL"} = "CRITICAL: %s"; -$LogFormats{"SUCCESS"} = "SUCCESS: %s"; -$LogFormats{"INFO"} = "INFO: %s"; -$LogFormats{"WARNING"} = "WARNING: %s"; +$LogFormats{"CRITICAL"} = "CRITICAL: %s"; +$LogFormats{"SUCCESS"} = "SUCCESS: %s"; +$LogFormats{"INFO"} = "INFO: %s"; +$LogFormats{"WARNING"} = "WARNING: %s"; $LogFormats{"DEFAULT"} = " %s "; +# UpdateStatus; +# Update the idle status display to show how many connections +# are left, retries and other stuff. +# +sub UpdateStatus { + if ($ConnectionRetriesLeft > 0) { + ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount + ." Retries remaining: ".$ConnectionRetriesLeft + ." ($KeyMode)"); + } else { + ShowStatus(GetServerHost()." >> DEAD <<"); + } +} + =pod @@ -240,9 +183,9 @@ host and the time will be formatted into =cut sub Log { - my $severity = shift; - my $message = shift; - + + my ($severity, $message) = @_; + if(!$LogFormats{$severity}) { $severity = "DEFAULT"; } @@ -277,8 +220,10 @@ Returns the name of the host that a sock =cut sub GetPeername { - my $connection = shift; - my $AdrFamily = shift; + + + my ($connection, $AdrFamily) = @_; + my $peer = $connection->peername(); my $peerport; my $peerip; @@ -292,7 +237,6 @@ sub GetPeername { return $peerfile; } } -#----------------------------- Timer management ------------------------ =pod =head2 Debug @@ -302,18 +246,20 @@ Invoked to issue a debug message. =cut sub Debug { - my $level = shift; - my $message = shift; + + my ($level, $message) = @_; + if ($level <= $DebugLevel) { Log("INFO", "-Debug- $message host = $RemoteHost"); } } sub SocketDump { - my $level = shift; - my $socket= shift; + + my ($level, $socket) = @_; + if($level <= $DebugLevel) { - $socket->Dump(); + $socket->Dump(-1); # Ensure it will get dumped. } } @@ -344,11 +290,60 @@ sub ShowStatus { =cut sub SocketTimeout { my $Socket = shift; - + Log("WARNING", "A socket timeout was detected"); + Debug(5, " SocketTimeout called: "); + $Socket->Dump(0); + if(exists($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()); + $LondConnecting = 0; + } + } +# +# This function should be called by the child in all cases where it must +# exit. If the child process is running with the DieWhenIdle turned on +# it must create a lock file for the AF_UNIX socket in order to prevent +# connection requests from lonnet in the time between process exit +# and the parent picking up the listen again. +# Parameters: +# exit_code - Exit status value, however see the next parameter. +# message - If this optional parameter is supplied, the exit +# is via a die with this message. +# +sub child_exit { + my ($exit_code, $message) = @_; + + # Regardless of how we exit, we may need to do the lock thing: + + if($DieWhenIdle) { + # + # Create a lock file since there will be a time window + # between our exit and the parent's picking up the listen + # during which no listens will be done on the + # lonnet client socket. + # + my $lock_file = GetLoncSocketPath().".lock"; + open(LOCK,">$lock_file"); + print LOCK "Contents not important"; + close(LOCK); + + exit(0); + } + # Now figure out how we exit: + + if($message) { + die $message; + } else { + exit($exit_code); + } +} +#----------------------------- Timer management ------------------------ =pod @@ -360,13 +355,12 @@ Invoked each timer tick. sub Tick { + my ($Event) = @_; + my $clock_watcher = $Event->w; + my $client; - if($ConnectionRetriesLeft > 0) { - ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount - ." Retries remaining: ".$ConnectionRetriesLeft); - } else { - ShowStatus(GetServerHost()." >> DEAD <<"); - } + UpdateStatus(); + # Is it time to prune connection count: @@ -376,6 +370,12 @@ sub Tick { if($IdleSeconds > $IdleTimeout) { # Prune a connection... my $Socket = $IdleConnections->pop(); KillSocket($Socket); + $IdleSeconds = 0; # Otherwise all connections get trimmed to fast. + UpdateStatus(); + if(($ConnectionCount == 0) && $DieWhenIdle) { + &child_exit(0); + + } } } else { $IdleSeconds = 0; # Reset idle count if not idle. @@ -383,35 +383,44 @@ sub Tick { # # For each inflight transaction, tick down its timeout counter. # - foreach my $item (keys %ActiveTransactions) { - my $Socket = $ActiveTransactions{$item}->getServer(); - $Socket->Tick(); + + 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. # - + # Note this code is dead now... + # my $Requests = $WorkQueue->Count(); - if (($ConnectionCount == 0) && ($Requests > 0)) { + if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) { if ($ConnectionRetriesLeft > 0) { - my $Connections = ($Requests <= $MaxConnectionCount) ? - $Requests : $MaxConnectionCount; - Debug(1,"Work but no connections, start ".$Connections." of them"); - my $successCount = 0; - for (my $i =0; $i < $Connections; $i++) { - $successCount += MakeLondConnection(); - } - if($successCount == 0) { # All connections failed: - Debug(1,"Work in queue failed to make any connectiouns\n"); + Debug(5,"Work but no connections, Make a new one"); + my $success; + $success = &MakeLondConnection; + if($success == 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 { + $LondConnecting = 0; ShowStatus(GetServerHost()." >>> DEAD!!! <<<"); - Debug(1,"Work in queue, but gave up on connections..flushing\n"); + Debug(5,"Work in queue, but gave up on connections..flushing\n"); EmptyQueue(); # Connections can't be established. + CloseAllLondConnections(); # Should all already be closed but... } } + if ($ConnectionCount == 0) { + $KeyMode = ""; + $clock_watcher->cancel(); + } + &UpdateStatus(); } =pod @@ -432,7 +441,7 @@ 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 @@ -452,21 +461,22 @@ long enough, it will be shut down and re sub ServerToIdle { my $Socket = shift; # Get the socket. + $KeyMode = $Socket->{AuthenticationMode}; delete($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: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction - unless($reqdata eq undef) { - Debug(9, "Queue gave request data: ".$reqdata->getRequest()); + if ($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); } } @@ -512,7 +522,7 @@ 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"); @@ -531,7 +541,7 @@ 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!"); + # Log("INFO", "lonc pipe client hung up on us!"); $Watcher->cancel; $Socket->shutdown(2); $Socket->close(); @@ -583,12 +593,15 @@ The transaction that is being completed. =cut sub CompleteTransaction { - &Debug(6,"Complete transaction"); - my $Socket = shift; - my $Transaction = shift; + &Debug(5,"Complete transaction"); + + my ($Socket, $Transaction) = @_; if (!$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"); @@ -596,7 +609,9 @@ sub CompleteTransaction { unlink $Transaction->getFile(); } } + =pod + =head1 StartClientReply Initiates a reply to a client where the reply data is a parameter. @@ -612,17 +627,16 @@ sub CompleteTransaction { The data to send to apached client. =cut + sub StartClientReply { - my $Transaction = shift; - my $data = shift; + my ($Transaction, $data) = @_; 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", @@ -630,7 +644,9 @@ sub StartClientReply { cb => \&ClientWritable, data => $data); } + =pod + =head2 FailTransaction Finishes a transaction with failure because the associated lond socket @@ -640,8 +656,7 @@ 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: @@ -649,20 +664,24 @@ Parameters: The LondTransaction we are failing. + =cut sub FailTransaction { my $transaction = shift; - Log("WARNING", "Failing transaction ".$transaction->getRequest()); + + # If the socket is dead, that's already logged. + + if ($ConnectionRetriesLeft > 0) { + Log("WARNING", "Failing transaction " + .$transaction->getRequest()); + } Debug(1, "Failing transaction: ".$transaction->getRequest()); if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it. my $client = $transaction->getClient(); Debug(1," Replying con_lost to ".$transaction->getRequest()); StartClientReply($transaction, "con_lost\n"); } - if($ConnectionRetriesLeft <= 0) { - Log("CRITICAL", "Host marked dead: ".GetServerHost()); - } } @@ -690,7 +709,10 @@ Close all connections open on lond prior =cut sub CloseAllLondConnections { foreach my $Socket (keys %ActiveConnections) { - KillSocket($Socket); + if(exists($ActiveTransactions{$Socket})) { + FailTransaction($ActiveTransactions{$Socket}); + } + KillSocket($Socket); } } =cut @@ -734,14 +756,15 @@ sub KillSocket { } if(exists($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(); + CloseAllLondConnections; # Should all already be closed but... } } @@ -767,6 +790,17 @@ 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 @@ -821,13 +855,18 @@ sub LondReadable { &Debug(2, "Socket->Readable returned: $status"); if($status != 0) { - # bad return from socket read. Currently this means that + # bad return from socket read. Currently this means that # The socket has become disconnected. We fail the transaction. Log("WARNING", "Lond connection lost."); if(exists($ActiveTransactions{$Socket})) { FailTransaction($ActiveTransactions{$Socket}); + } else { + # Socket is connecting and failed... need to mark + # no longer connecting. + + $LondConnecting = 0; } $Watcher->cancel(); KillSocket($Socket); @@ -839,19 +878,34 @@ 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 ($State eq "RequestingKey") { # The ok was received. Now we need to request the key @@ -863,6 +917,14 @@ sub LondReadable { } elsif ($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 @@ -871,7 +933,7 @@ sub LondReadable { $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 { @@ -879,6 +941,10 @@ sub LondReadable { .$RemoteHost." now ready for action"); } ServerToIdle($Socket); # Next work unit or idle. + + # + $LondConnecting = 0; # Best spot I can think of for this. + # } elsif ($State eq "SendingRequest") { # We need to be writable for this and probably don't belong @@ -892,7 +958,7 @@ sub LondReadable { } else { - # Invalid state. + # Invalid state. Debug(4, "Invalid state in LondReadable"); } } @@ -974,21 +1040,40 @@ sub LondWritable { SocketDump(6,$Socket); - if ($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"); + # 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; + } else { + # In the process of conneting, so need to turn that off. + + $LondConnecting = 0; } - # "init" is being sent... + $Watcher->cancel(); + KillSocket($Socket); + return; + } - + + + if ($State eq "Connected") { + + # "init" is being sent... + } elsif ($State eq "Initialized") { # Now that init was sent, we switch @@ -996,40 +1081,43 @@ 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: @@ -1038,20 +1126,10 @@ 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 @@ -1118,7 +1196,7 @@ 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. @@ -1130,8 +1208,8 @@ sub MakeLondConnection { # exchange underway. # my $Socket = $Connection->GetSocket(); - if($Socket == undef) { - die "did not get a socket from the connection"; + if($Socket eq undef) { + &child_exit(-1, "did not get a socket from the connection"); } else { &Debug(9,"MakeLondConnection got socket: ".$Socket); } @@ -1144,7 +1222,9 @@ sub MakeLondConnection { data => $Connection, desc => 'Connection to lond server'); $ActiveConnections{$Connection} = $event; - + if ($ConnectionCount == 0) { + &SetupTimer; # Need to handle timeouts with connections... + } $ConnectionCount++; Debug(4, "Connection count = ".$ConnectionCount); if($ConnectionCount == 1) { # First Connection: @@ -1152,6 +1232,7 @@ sub MakeLondConnection { } Log("SUCESS", "Created connection ".$ConnectionCount ." to host ".GetServerHost()); + $LondConnecting = 1; # Connection in progress. return 1; # Return success. } @@ -1183,8 +1264,8 @@ The text of the request to send. =cut sub StartRequest { - my $Lond = shift; - my $Request = shift; # This is a LondTransaction. + + my ($Lond, $Request) = @_; Debug(6, "StartRequest: ".$Request->getRequest()); @@ -1233,17 +1314,22 @@ 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(5, "Queue Transaction startnew $ConnectionCount $LondConnecting"); + if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) { + if($ConnectionRetriesLeft > 0) { - Debug(4,"Starting additional lond connection"); - if(MakeLondConnection() == 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 !!!! <<<"); + $LondConnecting = 0; EmptyQueue(); # It's worse than that ... he's dead Jim. + CloseAllLondConnections; # Should all be closed but.. } } } else { # Can start the request: @@ -1275,7 +1361,7 @@ 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); @@ -1286,7 +1372,7 @@ 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"); @@ -1294,6 +1380,9 @@ 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); @@ -1302,6 +1391,40 @@ sub ClientRequest { } +# +# Accept a connection request for a client (lonc child) and +# start up an event watcher to keep an eye on input from that +# Event. This can be called both from NewClient and from +# ChildProcess if we are started in DieWhenIdle mode. +# Parameters: +# $socket - The listener socket. +# Returns: +# NONE +# Side Effects: +# An event is made to watch the accepted connection. +# Active clients hash is updated to reflect the new connection. +# The client connection count is incremented. +# +sub accept_client { + my ($socket) = @_; + + Debug(8, "Entering accept for lonc UNIX socket\n"); + my $connection = $socket->accept(); # Accept the client connection. + Debug(8,"Connection request accepted from " + .GetPeername($connection, AF_UNIX)); + + + my $description = sprintf("Connection to lonc client %d", + $ClientConnection); + Debug(9, "Creating event named: ".$description); + Event->io(cb => \&ClientRequest, + poll => 'r', + desc => $description, + data => "", + fd => $connection); + $ActiveClients{$connection} = $ClientConnection; + $ClientConnection++; +} =pod @@ -1320,21 +1443,8 @@ sub NewClient { my $event = shift; # Get the event parameters. my $watcher = $event->w; my $socket = $watcher->fd; # Get the event' socket. - my $connection = $socket->accept(); # Accept the client connection. - Debug(8,"Connection request accepted from " - .GetPeername($connection, AF_UNIX)); - - my $description = sprintf("Connection to lonc client %d", - $ClientConnection); - Debug(9, "Creating event named: ".$description); - Event->io(cb => \&ClientRequest, - poll => 'r', - desc => $description, - data => "", - fd => $connection); - $ActiveClients{$connection} = $ClientConnection; - $ClientConnection++; + &accept_client($socket); } =pod @@ -1344,10 +1454,20 @@ sub NewClient { Returns the name of the UNIX socket on which to listen for client connections. +=head2 Parameters: + + host (optional) - Name of the host socket to return.. defaults to + the return from GetServerHost(). + =cut sub GetLoncSocketPath { - return $UnixSocketDir."/".GetServerHost(); + + my $host = GetServerHost(); # Default host. + if (@_) { + ($host) = @_; # Override if supplied. + } + return $UnixSocketDir."/".$host; } =pod @@ -1384,22 +1504,50 @@ connection. The event handler establish (creating a communcations channel), that int turn will establish another event handler to subess requests. +=head2 Parameters: + + host (optional) Name of the host to set up a unix socket to. + =cut sub SetupLoncListener { + my $host = GetServerHost(); # Default host. + if (@_) { + ($host) = @_ # Override host with parameter. + } + my $socket; - my $SocketName = GetLoncSocketPath(); + my $SocketName = GetLoncSocketPath($host); unlink($SocketName); unless ($socket =IO::Socket::UNIX->new(Local => $SocketName, - Listen => 10, + Listen => 250, Type => SOCK_STREAM)) { - die "Failed to create a lonc listner socket"; + if($I_am_child) { + &child_exit(-1, "Failed to create a lonc listener socket"); + } else { + die "Failed to create a lonc listner socket"; + } } - Event->io(cb => \&NewClient, - poll => 'r', - desc => 'Lonc listener Unix Socket', - fd => $socket); + return $socket; +} + +# +# Toggle transaction logging. +# Implicit inputs: +# LogTransactions +# Implicit Outputs: +# LogTransactions +sub ToggleTransactionLogging { + print STDERR "Toggle transaction logging...\n"; + if(!$LogTransactions) { + $LogTransactions = 1; + } else { + $LogTransactions = 0; + } + + + Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n"); } =pod @@ -1412,15 +1560,36 @@ 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 $$."\t".$RemoteHost."\t".$Status."\t". + + open(LOG,">>$docdir/lon-status/loncstatus.txt"); + flock(LOG,LOCK_EX); + print LOG $$."\t".$RemoteHost."\t".$Status."\t". $RecentLogEntry."\n"; + # + # Write out information about each of the connections: + # + if ($DebugLevel > 2) { + print LOG "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 LOG "Connection $i State: $state\n"; + print STDERR "---------------------- Connection $i \n"; + $Socket->Dump(-1); # Ensure it gets dumped.. + $i++; + } + } + flock(LOG,LOCK_UN); + close(LOG); $ConnectionRetriesLeft = $ConnectionRetries; } @@ -1442,8 +1611,8 @@ sub SignalledToDeath { chomp($signal); Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost " ."died through "."\"$signal\""); - LogPerm("F:lonc: $$ on $RemoteHost signalled to death: " - ."\"$signal\""); + #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: " +# ."\"$signal\""); exit 0; } @@ -1466,15 +1635,49 @@ sub ToggleDebug { =head2 ChildProcess This sub implements a child process for a single lonc daemon. +Optional parameter: + $socket - if provided, this is a socket already open for listen + on the client socket. Otherwise, a new listen is set up. =cut sub ChildProcess { + # If we are in DieWhenIdle mode, we've inherited all the + # events of our parent and those have to be cancelled or else + # all holy bloody chaos will result.. trust me, I already made + # >that< mistake. + + my $host = GetServerHost(); + foreach my $listener (keys %parent_dispatchers) { + my $watcher = $parent_dispatchers{$listener}; + my $s = $watcher->fd; + if ($listener ne $host) { # Close everyone but me. + Debug(5, "Closing listen socket for $listener"); + $s->close(); + } + Debug(5, "Killing watcher for $listener"); + + $watcher->cancel(); + 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 + # parent event dispatches.. flag I'm a child. # # Signals must be handled by the Event framework... -# + # Event->signal(signal => "QUIT", cb => \&SignalledToDeath, @@ -1485,13 +1688,27 @@ sub ChildProcess { Event->signal(signal => "USR1", cb => \&ChildStatus, data => "USR1"); + Event->signal(signal => "USR2", + cb => \&ToggleTransactionLogging); Event->signal(signal => "INT", cb => \&ToggleDebug, data => "INT"); - SetupTimer(); - - SetupLoncListener(); + # Figure out if we got passed a socket or need to open one to listen for + # client requests. + + my ($socket) = @_; + if (!$socket) { + + $socket = SetupLoncListener(); + } + # Establish an event to listen for client connection requests. + + + Event->io(cb => \&NewClient, + poll => 'r', + desc => 'Lonc Listener Unix Socket', + fd => $socket); $Event::Debuglevel = $DebugLevel; @@ -1499,38 +1716,193 @@ sub ChildProcess { # Setup the initial server connection: - # &MakeLondConnection(); // let first work requirest do it. + # &MakeLondConnection(); // let first work request do it. + # If We are in diwhenidle, need to accept the connection since the + # event may not fire. + + if ($DieWhenIdle) { + &accept_client($socket); + } Debug(9,"Entering event loop"); my $ret = Event::loop(); # Start the main event loop. - die "Main event loop exited!!!"; + &child_exit (-1,"Main event loop exited!!!"); } # Create a new child for host passed in: sub CreateChild { + my ($host, $socket) = @_; + my $sigset = POSIX::SigSet->new(SIGINT); sigprocmask(SIG_BLOCK, $sigset); - my $host = shift; $RemoteHost = $host; Log("CRITICAL", "Forking server for ".$host); my $pid = fork; if($pid) { # Parent $RemoteHost = "Parent"; - $ChildHash{$pid} = $RemoteHost; + $ChildHash{$pid} = $host; + $HostToPid{$host}= $pid; sigprocmask(SIG_UNBLOCK, $sigset); } else { # child. ShowStatus("Connected to ".$RemoteHost); $SIG{INT} = 'DEFAULT'; sigprocmask(SIG_UNBLOCK, $sigset); - ChildProcess; # Does not return. + if(defined $socket) { + &ChildProcess($socket); + } else { + ChildProcess; # Does not return. + } + } +} + +# parent_client_connection: +# Event handler that processes client connections for the parent process. +# This sub is called when the parent is listening on a socket and +# a connection request arrives. We must: +# Start a child process to accept the connection request. +# Kill our listen on the socket. +# Parameter: +# event - The event object that was created to monitor this socket. +# event->w->fd is the socket. +# Returns: +# NONE +# +sub parent_client_connection { + if ($I_am_child) { + # Should not get here, but seem to anyway: + &Debug(5," Child caught parent client connection event!!"); + my ($event) = @_; + my $watcher = $event->w; + $watcher->cancel(); # Try to kill it off again!! + } else { + &Debug(9, "parent_client_connection"); + my ($event) = @_; + my $watcher = $event->w; + my $socket = $watcher->fd; + + # Lookup the host associated with this socket: + + my $host = $listening_to{$socket}; + + # 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(); + } +} + +# parent_listen: +# Opens a socket and starts a listen for the parent process on a client UNIX +# domain socket. +# +# This involves: +# Creating a socket for listen. +# Removing any socket lock file +# Adding an event handler for this socket becoming readable +# To the parent's event dispatcher. +# Parameters: +# loncapa_host - LonCAPA cluster name of the host represented by the client +# socket. +# Returns: +# NONE +# +sub parent_listen { + my ($loncapa_host) = @_; + Debug(5, "parent_listen: $loncapa_host"); + + my $socket = &SetupLoncListener($loncapa_host); + $listening_to{$socket} = $loncapa_host; + if (!$socket) { + die "Unable to create a listen socket for $loncapa_host"; + } + + my $lock_file = &GetLoncSocketPath($loncapa_host).".lock"; + unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.] + + my $watcher = Event->io(cb => \&parent_client_connection, + poll => 'r', + desc => "Parent listener unix socket ($loncapa_host)", + fd => $socket); + $parent_dispatchers{$loncapa_host} = $watcher; + +} + + +# listen_on_all_unix_sockets: +# This sub initiates a listen on all unix domain lonc client sockets. +# This will be called in the case where we are trimming idle processes. +# When idle processes are trimmed, loncnew starts up with no children, +# 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. +# +# Parameters: +# NONE +# Implicit Inputs: +# The configuration file that has been read in by LondConnection. +# Returns: +# NONE +# +sub listen_on_all_unix_sockets { + Debug(5, "listen_on_all_unix_sockets"); + my $host_iterator = &LondConnection::GetHostIterator(); + 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. +# Since this is dispatched via a signal, we must process all +# dead children until there are no more left. The action +# is to: +# - Remove the child from the bookeeping hashes +# - Re-establish a listen on the unix domain socket associated +# with that host. +# Parameters: +# The event, but we don't actually care about it. +sub server_died { + &Debug(9, "server_died called..."); + + while(1) { # Loop until waitpid nowait fails. + my $pid = waitpid(-1, WNOHANG); + if($pid <= 0) { + return; # Nothing left to wait for. + } + # need the host to restart: + + my $host = $ChildHash{$pid}; + if($host) { # It's for real... + &Debug(9, "Caught sigchild for $host"); + delete($ChildHash{$pid}); + delete($HostToPid{$host}); + &parent_listen($host); + + } else { + &Debug(5, "Caught sigchild for pid not in hosts hash: $pid"); + } } } + # # Parent process logic pass 1: # For each entry in the hosts table, we will @@ -1578,43 +1950,93 @@ ShowStatus("Forking node servers"); Log("CRITICAL", "--------------- Starting children ---------------"); +LondConnection::ReadConfig; # Read standard config files. my $HostIterator = LondConnection::GetHostIterator; -while (! $HostIterator->end()) { - my $hostentryref = $HostIterator->get(); - CreateChild($hostentryref->[0]); - $HostIterator->next(); +if ($DieWhenIdle) { + $RemoteHost = "[parent]"; + &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"; # Maintain the population: ShowStatus("Parent keeping the flock"); -# -# Set up parent signals: -# -$SIG{INT} = \&Terminate; -$SIG{TERM} = \&Terminate; -$SIG{HUP} = \&Restart; -$SIG{USR1} = \&CheckKids; -$SIG{USR2} = \&UpdateKids; # LonManage update request. - -while(1) { - my $deadchild = wait(); - if(exists $ChildHash{$deadchild}) { # need to restart. - my $deadhost = $ChildHash{$deadchild}; - delete($ChildHash{$deadchild}); - Log("WARNING","Lost child pid= ".$deadchild. - "Connected to host ".$deadhost); - Log("INFO", "Restarting child procesing ".$deadhost); - CreateChild($deadhost); +if ($DieWhenIdle) { + # We need to setup a SIGChild event to handle the exit (natural or otherwise) + # of the children. + + Event->signal(cb => \&server_died, + desc => "Child exit handler", + signal => "CHLD"); + + + # Set up all the other signals we set up. We'll vector them off to the + # same subs as we would for DieWhenIdle false and, if necessary, conditionalize + # the code there. + + $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 => \&Restart, + 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"; + + +} 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); + } } } - =pod =head1 CheckKids @@ -1634,18 +2056,21 @@ sub CheckKids { my $now=time; my $local=localtime($now); print $fh "LONC status $local - parent $$ \n\n"; + foreach my $host (keys %parent_dispatchers) { + print $fh "LONC Parent process listening for $host\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 SIGUSR1 handler. This handler: +parent's SIGUSR2 handler. This handler: =item @@ -1657,7 +2082,7 @@ Kills off (via sigint) children for host =item -HUP's children for hosts that already exist (this just forces a status display +QUITs children for hosts that already exist (this just forces a status display and resets the connection retry count for that host. =item @@ -1668,6 +2093,18 @@ the start of the master program and main =cut sub UpdateKids { + + Log("INFO", "Updating connections via SIGUSR2"); + + # I'm not sure what I was thinking in the first implementation. + # someone will have to work hard to convince me the effect is any + # 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 + # (lost unless they are critical). + + &Restart(); + } @@ -1686,7 +2123,7 @@ sub Restart { Log("CRITICAL", "Restarting"); my $execdir = $perlvar{'lonDaemons'}; unlink("$execdir/logs/lonc.pid"); - exec("$execdir/lonc"); + exec("$executable"); } =pod @@ -1703,17 +2140,33 @@ sub KillThemAll { 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"); + ShowStatus("Nicely Killing lonc for $serving pid = $pid"); + Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid"); kill 'QUIT' => $pid; - delete($ChildHash{$pid}); } - my $execdir = $perlvar{'lonDaemons'}; - unlink("$execdir/logs/lonc.pid"); + } + +# +# Kill all children via KILL. Just in case the +# first shot didn't get them. + +sub really_kill_them_all_dammit +{ + Debug(2, "Kill them all Dammit"); + local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them. + foreach my $pid (keys %ChildHash) { + my $serving = $ChildHash{$pid}; + &ShowStatus("Nastily killing lonc for $serving pid = $pid"); + Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid"); + kill 'KILL' => $pid; + delete($ChildHash{$pid}); + my $execdir = $perlvar{'lonDaemons'}; + unlink("$execdir/logs/lonc.pid"); + } +} =pod =head1 Terminate @@ -1723,7 +2176,15 @@ Terminate the system. =cut sub Terminate { - KillThemAll; + &Log("CRITICAL", "Asked to kill children.. first be nice..."); + &KillThemAll; + # + # By now they really should all be dead.. but just in case + # send them all SIGKILL's after a bit of waiting: + + sleep(4); + &Log("CRITICAL", "Now kill children nasty"); + &really_kill_them_all_dammit; Log("CRITICAL","Master process exiting"); exit 0;