--- loncom/loncnew 2007/03/28 21:44:05 1.82 +++ loncom/loncnew 2007/06/18 22:49:52 1.87 @@ -2,7 +2,7 @@ # The LearningOnline Network with CAPA # lonc maintains the connections to remote computers # -# $Id: loncnew,v 1.82 2007/03/28 21:44:05 albertel Exp $ +# $Id: loncnew,v 1.87 2007/06/18 22:49:52 albertel Exp $ # # Copyright Michigan State University Board of Trustees # @@ -72,7 +72,8 @@ my %perlvar = %{$perlvarref}; # # parent and shared variables. -my %ChildHash; # by pid -> host. +my %ChildPid; # by pid -> host. +my %ChildHost; # by host. my %listening_to; # Socket->host table for who the parent # is listening to. my %parent_dispatchers; # host-> listener watcher events. @@ -157,6 +158,7 @@ sub LogPerm { my $now=time; my $local=localtime($now); my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log"); + chomp($message); print $fh "$now:$message:$local\n"; } @@ -501,6 +503,9 @@ the data and Event->w->fd is the socket sub ClientWritable { my $Event = shift; my $Watcher = $Event->w; + if (!defined($Watcher)) { + &child_exit(-1,'No watcher for event in ClientWritable'); + } my $Data = $Watcher->data; my $Socket = $Watcher->fd; @@ -564,6 +569,7 @@ sub ClientWritable { } } else { $Watcher->cancel(); # A delayed request...just cancel. + return; } } @@ -603,8 +609,8 @@ sub CompleteTransaction { StartClientReply($Transaction, $data); } else { # Delete deferred transaction file. Log("SUCCESS", "A delayed transaction was completed"); - LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest()); - unlink $Transaction->getFile(); + LogPerm("S:".$Transaction->getClient().":".$Transaction->getRequest()); + unlink($Transaction->getFile()); } } @@ -1377,14 +1383,23 @@ sub ClientRequest { $data = $data.$thisread; # Append new data. $watcher->data($data); if($data =~ /\n$/) { # Request entirely read. - if($data eq "close_connection_exit\n") { + if ($data eq "close_connection_exit\n") { Log("CRITICAL", "Request Close Connection ... exiting"); CloseAllLondConnections(); 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); - if($LogTransactions) { + if ($LogTransactions) { Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n. } my $Transaction = LondTransaction->new($data); @@ -1762,7 +1777,7 @@ sub CreateChild { my $pid = fork; if($pid) { # Parent $RemoteHost = "Parent"; - $ChildHash{$pid} = $host; + $ChildPid{$pid} = $host; sigprocmask(SIG_UNBLOCK, $sigset); undef(@all_host_ids); } else { # child. @@ -1832,8 +1847,12 @@ sub get_remote_hostname { &Debug(5,"Creating child for $data (parent_client_connection)"); (my $hostname,my $lonid,@all_host_ids) = split(':',$data); - &CreateChild($hostname,$lonid); - + $ChildHost{$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 $socket->send("done\n"); @@ -1888,7 +1907,7 @@ sub parent_listen { sub parent_clean_up { my ($loncapa_host) = @_; - Debug(5, "parent_clean_up: $loncapa_host"); + Debug(1, "parent_clean_up: $loncapa_host"); my $socket_file = &GetLoncSocketPath($loncapa_host); unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.] @@ -1897,14 +1916,12 @@ sub parent_clean_up { } -# 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. + +# 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: # NONE @@ -1913,18 +1930,6 @@ sub parent_clean_up { # 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->[3]; - Debug(9, "Listen for $host_name"); - &parent_listen($host_name); - $host_iterator->next(); - } -} - sub listen_on_common_socket { Debug(5, "listen_on_common_socket"); &parent_listen(); @@ -1949,10 +1954,11 @@ sub server_died { } # need the host to restart: - my $host = $ChildHash{$pid}; + my $host = $ChildPid{$pid}; if($host) { # It's for real... &Debug(9, "Caught sigchild for $host"); - delete($ChildHash{$pid}); + delete($ChildPid{$pid}); + delete($ChildHost{$host}); &parent_clean_up($host); } else { @@ -2076,7 +2082,7 @@ sub CheckKids { foreach my $host (keys %parent_dispatchers) { print $fh "LONC Parent process listening for $host\n"; } - foreach my $pid (keys %ChildHash) { + foreach my $pid (keys %ChildPid) { Debug(2, "Sending USR1 -> $pid"); kill 'USR1' => $pid; # Tell Child to report status. } @@ -2153,13 +2159,19 @@ SIGHUP. Responds to sigint and sigterm. sub KillThemAll { Debug(2, "Kill them all!!"); - local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die. - foreach my $pid (keys %ChildHash) { - my $serving = $ChildHash{$pid}; + + #local($SIG{CHLD}) = 'IGNORE'; + # 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"); Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid"); kill 'QUIT' => $pid; } + ShowStatus("Finished killing child processes off."); } @@ -2171,12 +2183,12 @@ sub really_kill_them_all_dammit { Debug(2, "Kill them all Dammit"); local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them. - foreach my $pid (keys %ChildHash) { - my $serving = $ChildHash{$pid}; + foreach my $pid (keys %ChildPid) { + my $serving = $ChildPid{$pid}; &ShowStatus("Nastily killing lonc for $serving pid = $pid"); Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid"); kill 'KILL' => $pid; - delete($ChildHash{$pid}); + delete($ChildPid{$pid}); my $execdir = $perlvar{'lonDaemons'}; unlink("$execdir/logs/lonc.pid"); }