Thu Mar 10 18:03:12 PST 2005
- Previous message: [Slony1-commit] By smsimms: Removed a "my" keyword that limits the scope of Slony sets
- Next message: [Slony1-commit] By smsimms: Document the existence of store_node.pl.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Log Message: ----------- New script to add a node to an existing cluster. Based on init_cluster.pl, minus the "INIT CLUSTER" command, and only adding paths and listens for the relevant node. Added Files: ----------- slony1-engine/tools/altperl: store_node.pl (r1.1) -------------- next part -------------- --- /dev/null +++ tools/altperl/store_node.pl @@ -0,0 +1,233 @@ +#!@@PERL@@ +# $Id: store_node.pl,v 1.1 2005/03/10 18:03:10 smsimms Exp $ +# Author: Steve Simms +# Copyright 2005 PostgreSQL Global Development Group + +use Getopt::Long; + +# Defaults +my $CONFIG_FILE = '@@SYSCONFDIR@@/slon_tools.conf'; +my $SHOW_USAGE = 0; + +# Read command-line options +GetOptions("config=s" => \$CONFIG_FILE, + "help" => \$SHOW_USAGE); + +my $USAGE = +"Usage: store_node [--config file] node# + + Generates the slonik commands necessary to add a node to a + cluster. Also displays a report showing the relationships between + the various nodes. + +"; + +if ($SHOW_USAGE) { + print $USAGE; + exit 0; +} + +require '@@PGLIBDIR@@/slon-tools.pm'; +require $CONFIG_FILE; + +$node = $ARGV[0]; + +# Node can be passed either as "node1" or just "1" +if ($node =~ /^(?:node)?(\d+)$/) { + $node = $1; +} else { + die $USAGE; +} + +my $FILE="/tmp/store_node.$$"; + +open(SLONIK, ">", $FILE); +print SLONIK genheader(); + +# STORE NODE +print SLONIK "\n# STORE NODE\n"; +my ($dbname, $dbhost) = ($DBNAME[$node], $HOST[$node]); +print SLONIK " store node (id = $node, event node = $MASTERNODE, comment = 'Node $node - $dbname\@$dbhost');\n"; +print SLONIK " echo 'Set up replication nodes';\n"; + +# STORE PATH +print SLONIK "\n# STORE PATH\n"; + +my @COST; +my @VIA; +my @PATH; +generate_listen_paths(); + +print SLONIK " echo 'Next: configure paths for each node/origin';\n"; +foreach my $nodea (@NODES) { + my $dsna = $DSN[$nodea]; + foreach my $nodeb (@NODES) { + if ($nodea != $nodeb) { + next unless ($node == $nodea or $node == $nodeb); + my $dsnb = $DSN[$nodeb]; + my $providerba = $VIA[$nodea][$nodeb]; + my $providerab = $VIA[$nodeb][$nodea]; + if (!$printed[$nodea][$nodeb] and $providerab == $nodea) { + print SLONIK " store path (server = $nodea, client = $nodeb, conninfo = '$dsna');\n"; + $printed[$nodea][$nodeb] = "done"; + } + if (!$printed[$nodeb][$nodea] and $providerba == $nodea) { + print SLONIK " store path (server = $nodeb, client = $nodea, conninfo = '$dsnb');\n"; + $printed[$nodeb][$nodea] = "done"; + } + } + } +} + +# STORE LISTEN +print SLONIK "\n# STORE LISTEN\n"; +foreach my $origin (@NODES) { + my $dsna = $DSN[$origin]; + foreach my $receiver (@NODES) { + if ($origin != $receiver) { + my $provider = $VIA[$origin][$receiver]; + next unless ($node == $origin or + $node == $receiver or + $node == $provider); + print SLONIK " store listen (origin = $origin, receiver = $receiver, provider = $provider);\n"; + } + } +} +print SLONIK " echo 'Replication nodes prepared';\n"; +print SLONIK " echo 'Please start a slon replication daemon for each node';\n"; +close SLONIK; +run_slonik_script($FILE); +report_on_paths(); + +sub generate_listen_paths { + my $infinity = 10000000; # Initial costs are all infinite + foreach my $node1 (@NODES) { + foreach my $node2 (@NODES) { + $COST[$node1][$node2] = $infinity; + } + } + + # Initialize paths between parents and children, and based on them, + # generate initial seeding of listener paths, @VIA + + foreach my $node1 (@NODES) { + $COST[$node1][$node1] = 0; + $VIA[$node1][$node1] = 0; + foreach my $node2 (@NODES) { + if ($node2 != $node1) { + if ((not ($PARENT[$node1] or $PARENT[$node2])) or + ($PARENT[$node1] and $PARENT[$node1] == $node2) or + ($PARENT[$node2] and $PARENT[$node2] == $node1)) { + $PATH[$node1][$node2] = 1; + $PATH[$node2][$node1] = 1; + # Set up a cost 1 path between them + # Parent to child + $COST[$node1][$node2] = 1; + $VIA[$node1][$node2] = $node1; + + # Child to parent + $COST[$node2][$node1] = 1; + $VIA[$node2][$node1] = $node2; + } + } + } + } + + # Now, update the listener paths... + # 4 level nested iteration: + # 1 while not done, do + # 2 for each node, node1 + # 3 for each node, node2, where node2 <> node1, where we don't + # yet have a listener path + # 4 for each node node3 (<> node1 or node2), + # consider introducing the listener path: + # node1 to node2 then node2 to node3 + # In concept, it's an O(n^4) algorithm; since the number of nodes, n, + # is not likely to get particularly large, it's not worth tuning + # further. + $didwork = "yes"; + while ($didwork eq "yes") { + $didwork = "no"; + foreach my $node1 (@NODES) { + foreach my $node3 (@NODES) { + if (($VIA[$node3][$node1] == 0) && ($node3 != $node1)) { + foreach my $node2 (@NODES) { + if ($PATH[$node1][$node2] && ($VIA[$node2][$node3] != 0) && ($node2 != $node3) && ($node2 != $node1)) { + # Consider introducing a path from n1 to n2 then n2 to n3 + # as a cheaper alternative to going direct from n1 to n3 + my $oldcost = $COST[$node3][$node1]; + my $newcost = $COST[$node1][$node2] + $COST[$node2][$node3]; + if ($newcost < $oldcost) { + $didwork = "yes"; + # So we go via node 2 + $VIA[$node3][$node1] = $node2; + $COST[$node3][$node1] = $newcost; + + $VIA[$node1][$node3] = $node2; + $COST[$node1][$node3] = $newcost; + } + } + } + } + } + } + } +} + +sub report_on_paths { + print "# Configuration Summary:\n"; + print "#\n"; + print "# COST\n"; + print "# "; + foreach my $node2 (@NODES) { + printf "| %3d ", $node2; + } + print "|\n# "; + print ("-----+" x (scalar(@NODES) + 1)); + print "\n"; + foreach my $node1 (@NODES) { + printf "# %3d ", $node1; + foreach my $node2 (@NODES) { + if ($COST[$node2][$node1] == $infinity) { + printf "| inf "; + } else { + printf "|%4d ", $COST[$node2][$node1]; + } + } + print "|\n"; + } + print "# \n"; + print "# VIA\n"; + print "# "; + foreach my $node2 (@NODES) { + printf "| %3d ", $node2; + } + print "|\n# "; + print ("-----+" x (scalar(@NODES) + 1)); + print "\n"; + foreach my $node1 (@NODES) { + printf "# %3d ", $node1; + foreach my $node2 (@NODES) { + printf "|%4d ", $VIA[$node2][$node1]; + } + print "|\n"; + } + + print "# \n"; + print "# PATHS\n"; + print "# "; + foreach my $node2 (@NODES) { + printf "| %3d ", $node2; + } + print "|\n# "; + print ("-----+" x (scalar(@NODES) + 1)); + print "\n"; + foreach my $node1 (@NODES) { + printf "# %3d ", $node1; + foreach my $node2 (@NODES) { + printf "|%4d ", $PATH[$node2][$node1]; + } + print "|\n"; + } + print "\n"; +}
- Previous message: [Slony1-commit] By smsimms: Removed a "my" keyword that limits the scope of Slony sets
- Next message: [Slony1-commit] By smsimms: Document the existence of store_node.pl.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Slony1-commit mailing list