percona/percona-toolkit
1
0
Fork 0
mirror of https://github.com/percona/percona-toolkit.git synced 2025-09-09 18:30:16 +00:00
Code Issues Packages Projects Releases Wiki Activity

Rewrite and enhance lib/Percona/XtraDB/Cluster.pm and t/pt-table-checksum/pxc.t. Change how ptc handles various cluster issues. Change lib/Sandbox.pm subs like start_sandbox() and start_cluster(). PXC docs in ptc are a work in progress.

Browse Source
This commit is contained in:
Daniel Nichter
2012-11-18 22:05:30 -07:00
parent 07bee85a00
commit e0f0ea0cdb
5 changed files with 841 additions and 407 deletions
Download Patch File Download Diff File Expand all files Collapse all files

290
bin/pt-table-checksum
View File

@@ -3375,87 +3375,66 @@ sub _d {
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Percona::XtraDB::Cluster;
use Mo;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Mo;
use Data::Dumper;
sub get_cluster_name {
my ($self, $cxn) = @_;
my $sql = "SHOW VARIABLES LIKE 'wsrep\_cluster\_name'";
PTDEBUG && _d($cxn->name, $sql);
my (undef, $cluster_name) = $cxn->dbh->selectrow_array($sql);
return $cluster_name;
}
sub is_cluster_node {
my ($self, $cxn) = @_;
return $self->{is_cluster_node}->{$cxn} if defined $self->{is_cluster_node}->{$cxn};
my $sql = "SHOW VARIABLES LIKE 'wsrep_on'";
PTDEBUG && _d($sql);
my $sql = "SHOW VARIABLES LIKE 'wsrep\_on'";
PTDEBUG && _d($cxn->name, $sql);
my $row = $cxn->dbh->selectrow_arrayref($sql);
PTDEBUG && _d(defined $row ? @$row : 'undef');
$self->{is_cluster_node}->{$cxn} = $row && $row->[1]
? ($row->[1] eq 'ON' || $row->[1] eq '1')
: 0;
PTDEBUG && _d(Dumper($row));
return unless $row && $row->[1] && ($row->[1] eq 'ON' || $row->[1] eq '1');
return $self->{is_cluster_node}->{$cxn};
my $cluster_name = $self->get_cluster_name($cxn);
return $cluster_name;
}
sub same_node {
my ($self, $cxn1, $cxn2) = @_;
my $sql = "SHOW VARIABLES LIKE 'wsrep\_sst\_receive\_address'";
PTDEBUG && _d($cxn1->name, $sql);
my (undef, $val1) = $cxn1->dbh->selectrow_array($sql);
PTDEBUG && _d($cxn2->name, $sql);
my (undef, $val2) = $cxn2->dbh->selectrow_array($sql);
return ($val1 || '') eq ($val2 || '');
}
sub same_cluster {
my ($self, $cxn1, $cxn2) = @_;
return unless $self->is_cluster_node($cxn1) && $self->is_cluster_node($cxn2);
return if $self->is_master_of($cxn1, $cxn2) || $self->is_master_of($cxn2, $cxn1);
my $sql = q{SHOW VARIABLES LIKE 'wsrep_cluster_name'};
PTDEBUG && _d($sql);
my (undef, $row) = $cxn1->dbh->selectrow_array($sql);
my (undef, $cxn2_row) = $cxn2->dbh->selectrow_array($sql);
return 0 if !$self->is_cluster_node($cxn1) || !$self->is_cluster_node($cxn2);
return unless $row eq $cxn2_row;
my $cluster1 = $self->get_cluster_name($cxn1);
my $cluster2 = $self->get_cluster_name($cxn2);
$sql = q{SHOW VARIABLES LIKE 'wsrep_cluster_address'};
PTDEBUG && _d($sql);
my (undef, $addr) = $cxn1->dbh->selectrow_array($sql);
my (undef, $cxn2_addr) = $cxn2->dbh->selectrow_array($sql);
return if $addr eq 'gcomm://' && $cxn2_addr eq 'gcomm://';
if ( $addr eq 'gcomm://' ) {
$addr = $self->_find_full_gcomm_addr($cxn1->dbh);
}
elsif ( $cxn2_addr eq 'gcomm://' ) {
$cxn2_addr = $self->_find_full_gcomm_addr($cxn2->dbh);
}
return 1 if lc($addr) eq lc($cxn2_addr);
return 1;
return ($cluster1 || '') eq ($cluster2 || '');
}
sub is_master_of {
my ($self, $cxn1, $cxn2) = @_;
my $cxn2_dbh = $cxn2->dbh;
my $sql = q{SHOW SLAVE STATUS};
PTDEBUG && _d($sql);
local $cxn2_dbh->{FetchHashKeyName} = 'NAME_lc';
my $slave_status = $cxn2_dbh->selectrow_hashref($sql);
return unless ref($slave_status) eq 'HASH';
my $port = $cxn1->dsn->{P};
return unless $slave_status->{master_port} eq $port;
return 1 if $cxn1->dsn->{h} eq $slave_status->{master_host};
my $host = scalar gethostbyname($cxn1->dsn->{h});
my $master_host = scalar gethostbyname($slave_status->{master_host});
return 1 if $master_host eq $host;
return;
}
sub _find_full_gcomm_addr {
my ($self, $dbh) = @_;
my $sql = q{SHOW VARIABLES LIKE 'wsrep_provider_options'};
PTDEBUG && _d($sql);
my (undef, $provider_opts) = $dbh->selectrow_array($sql);
my ($prov_addr) = $provider_opts =~ m{\Qgmcast.listen_addr\E\s*=\s*tcp://([^:]+:[0-9]+)\s*;}i;
my $full_gcomm = "gcomm://$prov_addr";
PTDEBUG && _d("gcomm address: ", $full_gcomm);
return $full_gcomm;
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
@@ -8658,9 +8637,13 @@ sub main {
$have_time = sub { return 1; };
}
# PXC helper class
# ########################################################################
# Set up PXC stuff.
# ########################################################################
my $cluster = Percona::XtraDB::Cluster->new();
my %cluster_name_for;
$cluster_name_for{$master_cxn} = $cluster->is_cluster_node($master_cxn);
# ########################################################################
# If this is not a dry run (--explain was not specified), then we're
# going to checksum the tables, so do the necessary preparations and
@@ -8678,7 +8661,7 @@ sub main {
my $slaves = []; # all slaves (that we can find)
my $slave_lag_cxns; # slaves whose lag we'll check
my $replica_lag; # ReplicaLagWaiter object
my $replica_lag_pr; # Progress for ReplicaLagWaiter
my $sys_load; # MySQLStatusWaiter object
@@ -8697,9 +8680,29 @@ sub main {
dbh => $master_dbh,
dsn => $master_dsn,
make_cxn => sub {
return $make_cxn->(@_, prev_dsn => $master_cxn->dsn());
my $cxn = $make_cxn->(@_, prev_dsn => $master_cxn->dsn());
$cluster_name_for{$cxn} = $cluster->is_cluster_node($cxn);
return $cxn;
},
);
# If the "master" is a cluster node, then a DSN table should have been
# used, and it may have all nodes' DSNs so the user can run the tool
# on any node, in which case it has the "master" node, the DSN given
# on the command line. So detect and remove this dupe.
if ( $cluster_name_for{$master_cxn} ) {
@$slaves = grep {
my $slave_cxn = $_;
if ( $cluster->same_node($master_cxn, $slave_cxn) ) {
PTDEBUG && _d('Removing ', $slave_cxn->name, 'from slaves',
'because it is the master');
0;
}
else {
$slave_cxn;
}
} @$slaves;
}
PTDEBUG && _d(scalar @$slaves, 'slaves found');
# https://bugs.launchpad.net/percona-toolkit/+bug/938068
@@ -8711,8 +8714,10 @@ sub main {
}
my $err = '';
for my $slave_cxn (@$slaves) {
next if $cluster->is_cluster_node($slave_cxn);
for my $slave_cxn ( @$slaves ) {
# https://bugs.launchpad.net/percona-toolkit/+bug/1080385
next if $cluster_name_for{$slave_cxn};
my $slave_binlog = 'STATEMENT';
if ( VersionParser->new($slave_cxn->dbh) >= '5.1.5' ) {
($slave_binlog) = $slave_cxn->dbh->selectrow_array(
@@ -8731,42 +8736,91 @@ sub main {
die $err if $err;
}
if ( $cluster->is_cluster_node($master_cxn) ) {
if ( $cluster_name_for{$master_cxn} ) {
if ( !@$slaves ) {
die $master_cxn->name() . " is a cluster node but no other nodes "
. "or regular replicas were found. Use --recursion-method=dsn "
. "to specify the other nodes in the cluster.\n";
}
else {
my $err = '';
for my $slave (@$slaves) {
if ( $cluster->is_cluster_node($slave)
&& !$cluster->same_cluster($slave, $master_cxn) ) {
$err .= $slave->name() . " is a cluster node, but doesn't "
. "belong to the same cluster as " . $master_cxn->name()
. ". This is not currently supported; You can try "
. "using --recursion-method=dsn to specify all nodes "
. "in the slave cluster.\n"
}
# Make sure the master and all node are in the same cluster.
my @other_cluster;
foreach my $slave ( @$slaves ) {
next unless $cluster_name_for{$slave};
if ( $cluster_name_for{$master_cxn} ne $cluster_name_for{$slave}) {
push @other_cluster, $slave;
}
warn $err if $err;
}
if ( @other_cluster ) {
die $master_cxn->name . " is in cluster "
. $cluster_name_for{$master_cxn} . " but these nodes are "
. "in other clusters:\n"
. join("\n",
map {' ' . $_->name . " is in cluster $cluster_name_for{$_}"}
@other_cluster) . "\n"
. "All nodes must be in the same cluster. "
. "For more information, please read the Percona XtraDB "
. "Cluster section of the tool's documentation.\n";
}
}
elsif ( @$slaves ) {
my $err = '';
for my $slave (@$slaves) {
if ( $cluster->is_cluster_node($slave) ) {
$err .= $slave->name() . " is a cluster node, but "
. $master_cxn->name() . " is not. This is not currently "
. "supported; You can try to specify "
. "all nodes in the cluster with "
. "--recursion-method=dsn if you want them checksummed.\n"
# master is not a cluster node, but what about the slaves?
my $direct_slave; # master -> direct_slave
my @slaves; # traditional slaves
my @nodes; # cluster nodes
foreach my $slave ( @$slaves ) {
if ( !$cluster_name_for{$slave} ) {
push @slaves, $slave;
next;
}
my $is_master_of = eval {
$ms->is_master_of($master_cxn->dbh, $slave->dbh);
};
if ( $EVAL_ERROR && $EVAL_ERROR =~ m/is not a slave/ ) {
push @nodes, $slave;
}
elsif ( $is_master_of ) {
$direct_slave = $slave;
}
else {
# Another error could have happened but we don't really
# care. We know for sure the slave is a node, so just
# presume that and carry on.
push @nodes, $slave;
}
}
my $err = '';
if ( @nodes ) {
if ( $direct_slave ) {
warn "Diffs will only be detected if the cluster is "
. "consistent with " . $direct_slave->name . " because "
. $master_cxn->name . " is a traditional replication master "
. " but these replicas are cluster nodes:\n"
. join("\n", map { ' ' . $_->name } @nodes) . "\n"
. "For more information, please read the Percona XtraDB "
. "Cluster section of the tool's documentation.\n";
}
else {
warn "Diffs may not be detected on these cluster nodes "
. "because the direct replica of " . $master_cxn->name
. " was not found or specified:\n"
. join("\n", map { ' ' . $_->name } @nodes) . "\n"
. "For more information, please read the Percona XtraDB "
. "Cluster section of the tool's documentation.\n";
}
if ( @slaves ) {
warn "Diffs will only be detected on these replicas if "
. "they replicate from " . $master_cxn->name . ":\n"
. join("\n", map { ' ' . $_->name } @slaves) . "\n"
. "For more information, please read the Percona XtraDB "
. "Cluster section of the tool's documentation.\n";
}
}
warn $err if $err;
}
if ( $o->get('check-slave-lag') ) {
PTDEBUG && _d('Will use --check-slave-lag to check for slave lag');
my $cxn = $make_cxn->(
@@ -8786,8 +8840,7 @@ sub main {
# to appear should be sufficient.
@$slave_lag_cxns = grep {
my $slave_cxn = $_;
if ( $cluster->is_cluster_node($slave_cxn)
&& $cluster->same_cluster($master_cxn, $slave_cxn) ) {
if ( $cluster_name_for{$slave_cxn} ) {
warn "Not checking replica lag on " . $slave_cxn->name()
. " because it is a cluster node.\n";
0;
@@ -10689,6 +10742,35 @@ can try something like the following:
SET boundaries = COALESCE(CONCAT('id BETWEEN ', lower_boundary,
' AND ', upper_boundary), '1=1');
=head1 Percona XtraDB Cluster
pt-table-checksum works with Percona XtraDB Cluster (PXC) 5.5.27-23.6 and newer,
but only the following PXC setups are supported, all of which require that
you use the C<dsn> method for L<"--recursion-method"> to specify cluster nodes.
Also, the lag check (see L<"REPLICA CHECKS">) is not performed for cluster
nodes.
=over
=item Single cluster
The simplest PXC setup is a single cluster: all servers are cluster nodes,
and there are no regular replicas. If all nodes are specified in the
DSN table (see L<"--recursion-method">), then you can run the tool on any
node and any diffs on any other nodes will be detected.
=item Single cluster with a replica
Any cluster node can also be a regular master and replicate to a regular
replica. The tool can only detect diffs on a replica if you run it on
the "master node"
Mixed replication setups are not currently supported. For example, the tool
does not work completely if the master host is replicating to a cluster,
or if the cluster is replicating to another cluster. In short, the only
supported setup is a single cluster with nodes optionally having traditional
replication slaves.
=head1 OUTPUT
The tool prints tabular results, one line per table:
@@ -11669,24 +11751,10 @@ past any replicas using row-based replication that are masters for
further replicas.
The tool automatically checks the C<binlog_format> on all servers.
See L<"--[no]check-binlog-format">
See L<"--[no]check-binlog-format"> .
(L<Bug 899415|https://bugs.launchpad.net/percona-toolkit/+bug/899415>)
=item Percona XtraDB Cluster
pt-table-checksum works with Percona XtraDB Cluster 5.5.27-23.6 and newer.
The C<dsn> method for L<"--recursion-method"> must be used to specify cluster
nodes and regular replicas because nodes are not regular replicas so they
cannot be detected automatically. The lag check (see L<"REPLICA CHECKS">)
is not performed for cluster nodes.
Mixed replication setups are not currently supported. For example, the tool
does not work completely if the master host is replicating to a cluster,
or if the cluster is replicating to another cluster. In short, the only
supported setup is a single cluster with nodes optionally having traditional
replication slaves.
=back
=head1 BUGS