Files
percona-toolkit/bin/pt-online-schema-change
2012-05-30 14:27:07 -03:00

7648 lines
242 KiB
Perl
Executable File

#!/usr/bin/env perl
# This program is part of Percona Toolkit: http://www.percona.com/software/
# See "COPYRIGHT, LICENSE, AND WARRANTY" at the end of this file for legal
# notices and disclaimers.
use strict;
use warnings FATAL => 'all';
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
# ###########################################################################
# OptionParser package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/OptionParser.pm
# t/lib/OptionParser.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package OptionParser;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use List::Util qw(max);
use Getopt::Long;
my $POD_link_re = '[LC]<"?([^">]+)"?>';
sub new {
my ( $class, %args ) = @_;
my @required_args = qw();
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($program_name) = $PROGRAM_NAME =~ m/([.A-Za-z-]+)$/;
$program_name ||= $PROGRAM_NAME;
my $home = $ENV{HOME} || $ENV{HOMEPATH} || $ENV{USERPROFILE} || '.';
my %attributes = (
'type' => 1,
'short form' => 1,
'group' => 1,
'default' => 1,
'cumulative' => 1,
'negatable' => 1,
);
my $self = {
head1 => 'OPTIONS', # These args are used internally
skip_rules => 0, # to instantiate another Option-
item => '--(.*)', # Parser obj that parses the
attributes => \%attributes, # DSN OPTIONS section. Tools
parse_attributes => \&_parse_attribs, # don't tinker with these args.
%args,
strict => 1, # disabled by a special rule
program_name => $program_name,
opts => {},
got_opts => 0,
short_opts => {},
defaults => {},
groups => {},
allowed_groups => {},
errors => [],
rules => [], # desc of rules for --help
mutex => [], # rule: opts are mutually exclusive
atleast1 => [], # rule: at least one opt is required
disables => {}, # rule: opt disables other opts
defaults_to => {}, # rule: opt defaults to value of other opt
DSNParser => undef,
default_files => [
"/etc/percona-toolkit/percona-toolkit.conf",
"/etc/percona-toolkit/$program_name.conf",
"$home/.percona-toolkit.conf",
"$home/.$program_name.conf",
],
types => {
string => 's', # standard Getopt type
int => 'i', # standard Getopt type
float => 'f', # standard Getopt type
Hash => 'H', # hash, formed from a comma-separated list
hash => 'h', # hash as above, but only if a value is given
Array => 'A', # array, similar to Hash
array => 'a', # array, similar to hash
DSN => 'd', # DSN
size => 'z', # size with kMG suffix (powers of 2^10)
time => 'm', # time, with an optional suffix of s/h/m/d
},
};
return bless $self, $class;
}
sub get_specs {
my ( $self, $file ) = @_;
$file ||= $self->{file} || __FILE__;
my @specs = $self->_pod_to_specs($file);
$self->_parse_specs(@specs);
open my $fh, "<", $file or die "Cannot open $file: $OS_ERROR";
my $contents = do { local $/ = undef; <$fh> };
close $fh;
if ( $contents =~ m/^=head1 DSN OPTIONS/m ) {
PTDEBUG && _d('Parsing DSN OPTIONS');
my $dsn_attribs = {
dsn => 1,
copy => 1,
};
my $parse_dsn_attribs = sub {
my ( $self, $option, $attribs ) = @_;
map {
my $val = $attribs->{$_};
if ( $val ) {
$val = $val eq 'yes' ? 1
: $val eq 'no' ? 0
: $val;
$attribs->{$_} = $val;
}
} keys %$attribs;
return {
key => $option,
%$attribs,
};
};
my $dsn_o = new OptionParser(
description => 'DSN OPTIONS',
head1 => 'DSN OPTIONS',
dsn => 0, # XXX don't infinitely recurse!
item => '\* (.)', # key opts are a single character
skip_rules => 1, # no rules before opts
attributes => $dsn_attribs,
parse_attributes => $parse_dsn_attribs,
);
my @dsn_opts = map {
my $opts = {
key => $_->{spec}->{key},
dsn => $_->{spec}->{dsn},
copy => $_->{spec}->{copy},
desc => $_->{desc},
};
$opts;
} $dsn_o->_pod_to_specs($file);
$self->{DSNParser} = DSNParser->new(opts => \@dsn_opts);
}
if ( $contents =~ m/^=head1 VERSION\n\n^(.+)$/m ) {
$self->{version} = $1;
PTDEBUG && _d($self->{version});
}
return;
}
sub DSNParser {
my ( $self ) = @_;
return $self->{DSNParser};
};
sub get_defaults_files {
my ( $self ) = @_;
return @{$self->{default_files}};
}
sub _pod_to_specs {
my ( $self, $file ) = @_;
$file ||= $self->{file} || __FILE__;
open my $fh, '<', $file or die "Cannot open $file: $OS_ERROR";
my @specs = ();
my @rules = ();
my $para;
local $INPUT_RECORD_SEPARATOR = '';
while ( $para = <$fh> ) {
next unless $para =~ m/^=head1 $self->{head1}/;
last;
}
while ( $para = <$fh> ) {
last if $para =~ m/^=over/;
next if $self->{skip_rules};
chomp $para;
$para =~ s/\s+/ /g;
$para =~ s/$POD_link_re/$1/go;
PTDEBUG && _d('Option rule:', $para);
push @rules, $para;
}
die "POD has no $self->{head1} section" unless $para;
do {
if ( my ($option) = $para =~ m/^=item $self->{item}/ ) {
chomp $para;
PTDEBUG && _d($para);
my %attribs;
$para = <$fh>; # read next paragraph, possibly attributes
if ( $para =~ m/: / ) { # attributes
$para =~ s/\s+\Z//g;
%attribs = map {
my ( $attrib, $val) = split(/: /, $_);
die "Unrecognized attribute for --$option: $attrib"
unless $self->{attributes}->{$attrib};
($attrib, $val);
} split(/; /, $para);
if ( $attribs{'short form'} ) {
$attribs{'short form'} =~ s/-//;
}
$para = <$fh>; # read next paragraph, probably short help desc
}
else {
PTDEBUG && _d('Option has no attributes');
}
$para =~ s/\s+\Z//g;
$para =~ s/\s+/ /g;
$para =~ s/$POD_link_re/$1/go;
$para =~ s/\.(?:\n.*| [A-Z].*|\Z)//s;
PTDEBUG && _d('Short help:', $para);
die "No description after option spec $option" if $para =~ m/^=item/;
if ( my ($base_option) = $option =~ m/^\[no\](.*)/ ) {
$option = $base_option;
$attribs{'negatable'} = 1;
}
push @specs, {
spec => $self->{parse_attributes}->($self, $option, \%attribs),
desc => $para
. (defined $attribs{default} ? " (default $attribs{default})" : ''),
group => ($attribs{'group'} ? $attribs{'group'} : 'default'),
};
}
while ( $para = <$fh> ) {
last unless $para;
if ( $para =~ m/^=head1/ ) {
$para = undef; # Can't 'last' out of a do {} block.
last;
}
last if $para =~ m/^=item /;
}
} while ( $para );
die "No valid specs in $self->{head1}" unless @specs;
close $fh;
return @specs, @rules;
}
sub _parse_specs {
my ( $self, @specs ) = @_;
my %disables; # special rule that requires deferred checking
foreach my $opt ( @specs ) {
if ( ref $opt ) { # It's an option spec, not a rule.
PTDEBUG && _d('Parsing opt spec:',
map { ($_, '=>', $opt->{$_}) } keys %$opt);
my ( $long, $short ) = $opt->{spec} =~ m/^([\w-]+)(?:\|([^!+=]*))?/;
if ( !$long ) {
die "Cannot parse long option from spec $opt->{spec}";
}
$opt->{long} = $long;
die "Duplicate long option --$long" if exists $self->{opts}->{$long};
$self->{opts}->{$long} = $opt;
if ( length $long == 1 ) {
PTDEBUG && _d('Long opt', $long, 'looks like short opt');
$self->{short_opts}->{$long} = $long;
}
if ( $short ) {
die "Duplicate short option -$short"
if exists $self->{short_opts}->{$short};
$self->{short_opts}->{$short} = $long;
$opt->{short} = $short;
}
else {
$opt->{short} = undef;
}
$opt->{is_negatable} = $opt->{spec} =~ m/!/ ? 1 : 0;
$opt->{is_cumulative} = $opt->{spec} =~ m/\+/ ? 1 : 0;
$opt->{is_required} = $opt->{desc} =~ m/required/ ? 1 : 0;
$opt->{group} ||= 'default';
$self->{groups}->{ $opt->{group} }->{$long} = 1;
$opt->{value} = undef;
$opt->{got} = 0;
my ( $type ) = $opt->{spec} =~ m/=(.)/;
$opt->{type} = $type;
PTDEBUG && _d($long, 'type:', $type);
$opt->{spec} =~ s/=./=s/ if ( $type && $type =~ m/[HhAadzm]/ );
if ( (my ($def) = $opt->{desc} =~ m/default\b(?: ([^)]+))?/) ) {
$self->{defaults}->{$long} = defined $def ? $def : 1;
PTDEBUG && _d($long, 'default:', $def);
}
if ( $long eq 'config' ) {
$self->{defaults}->{$long} = join(',', $self->get_defaults_files());
}
if ( (my ($dis) = $opt->{desc} =~ m/(disables .*)/) ) {
$disables{$long} = $dis;
PTDEBUG && _d('Deferring check of disables rule for', $opt, $dis);
}
$self->{opts}->{$long} = $opt;
}
else { # It's an option rule, not a spec.
PTDEBUG && _d('Parsing rule:', $opt);
push @{$self->{rules}}, $opt;
my @participants = $self->_get_participants($opt);
my $rule_ok = 0;
if ( $opt =~ m/mutually exclusive|one and only one/ ) {
$rule_ok = 1;
push @{$self->{mutex}}, \@participants;
PTDEBUG && _d(@participants, 'are mutually exclusive');
}
if ( $opt =~ m/at least one|one and only one/ ) {
$rule_ok = 1;
push @{$self->{atleast1}}, \@participants;
PTDEBUG && _d(@participants, 'require at least one');
}
if ( $opt =~ m/default to/ ) {
$rule_ok = 1;
$self->{defaults_to}->{$participants[0]} = $participants[1];
PTDEBUG && _d($participants[0], 'defaults to', $participants[1]);
}
if ( $opt =~ m/restricted to option groups/ ) {
$rule_ok = 1;
my ($groups) = $opt =~ m/groups ([\w\s\,]+)/;
my @groups = split(',', $groups);
%{$self->{allowed_groups}->{$participants[0]}} = map {
s/\s+//;
$_ => 1;
} @groups;
}
if( $opt =~ m/accepts additional command-line arguments/ ) {
$rule_ok = 1;
$self->{strict} = 0;
PTDEBUG && _d("Strict mode disabled by rule");
}
die "Unrecognized option rule: $opt" unless $rule_ok;
}
}
foreach my $long ( keys %disables ) {
my @participants = $self->_get_participants($disables{$long});
$self->{disables}->{$long} = \@participants;
PTDEBUG && _d('Option', $long, 'disables', @participants);
}
return;
}
sub _get_participants {
my ( $self, $str ) = @_;
my @participants;
foreach my $long ( $str =~ m/--(?:\[no\])?([\w-]+)/g ) {
die "Option --$long does not exist while processing rule $str"
unless exists $self->{opts}->{$long};
push @participants, $long;
}
PTDEBUG && _d('Participants for', $str, ':', @participants);
return @participants;
}
sub opts {
my ( $self ) = @_;
my %opts = %{$self->{opts}};
return %opts;
}
sub short_opts {
my ( $self ) = @_;
my %short_opts = %{$self->{short_opts}};
return %short_opts;
}
sub set_defaults {
my ( $self, %defaults ) = @_;
$self->{defaults} = {};
foreach my $long ( keys %defaults ) {
die "Cannot set default for nonexistent option $long"
unless exists $self->{opts}->{$long};
$self->{defaults}->{$long} = $defaults{$long};
PTDEBUG && _d('Default val for', $long, ':', $defaults{$long});
}
return;
}
sub get_defaults {
my ( $self ) = @_;
return $self->{defaults};
}
sub get_groups {
my ( $self ) = @_;
return $self->{groups};
}
sub _set_option {
my ( $self, $opt, $val ) = @_;
my $long = exists $self->{opts}->{$opt} ? $opt
: exists $self->{short_opts}->{$opt} ? $self->{short_opts}->{$opt}
: die "Getopt::Long gave a nonexistent option: $opt";
$opt = $self->{opts}->{$long};
if ( $opt->{is_cumulative} ) {
$opt->{value}++;
}
else {
$opt->{value} = $val;
}
$opt->{got} = 1;
PTDEBUG && _d('Got option', $long, '=', $val);
}
sub get_opts {
my ( $self ) = @_;
foreach my $long ( keys %{$self->{opts}} ) {
$self->{opts}->{$long}->{got} = 0;
$self->{opts}->{$long}->{value}
= exists $self->{defaults}->{$long} ? $self->{defaults}->{$long}
: $self->{opts}->{$long}->{is_cumulative} ? 0
: undef;
}
$self->{got_opts} = 0;
$self->{errors} = [];
if ( @ARGV && $ARGV[0] eq "--config" ) {
shift @ARGV;
$self->_set_option('config', shift @ARGV);
}
if ( $self->has('config') ) {
my @extra_args;
foreach my $filename ( split(',', $self->get('config')) ) {
eval {
push @extra_args, $self->_read_config_file($filename);
};
if ( $EVAL_ERROR ) {
if ( $self->got('config') ) {
die $EVAL_ERROR;
}
elsif ( PTDEBUG ) {
_d($EVAL_ERROR);
}
}
}
unshift @ARGV, @extra_args;
}
Getopt::Long::Configure('no_ignore_case', 'bundling');
GetOptions(
map { $_->{spec} => sub { $self->_set_option(@_); } }
grep { $_->{long} ne 'config' } # --config is handled specially above.
values %{$self->{opts}}
) or $self->save_error('Error parsing options');
if ( exists $self->{opts}->{version} && $self->{opts}->{version}->{got} ) {
if ( $self->{version} ) {
print $self->{version}, "\n";
}
else {
print "Error parsing version. See the VERSION section of the tool's documentation.\n";
}
exit 0;
}
if ( @ARGV && $self->{strict} ) {
$self->save_error("Unrecognized command-line options @ARGV");
}
foreach my $mutex ( @{$self->{mutex}} ) {
my @set = grep { $self->{opts}->{$_}->{got} } @$mutex;
if ( @set > 1 ) {
my $err = join(', ', map { "--$self->{opts}->{$_}->{long}" }
@{$mutex}[ 0 .. scalar(@$mutex) - 2] )
. ' and --'.$self->{opts}->{$mutex->[-1]}->{long}
. ' are mutually exclusive.';
$self->save_error($err);
}
}
foreach my $required ( @{$self->{atleast1}} ) {
my @set = grep { $self->{opts}->{$_}->{got} } @$required;
if ( @set == 0 ) {
my $err = join(', ', map { "--$self->{opts}->{$_}->{long}" }
@{$required}[ 0 .. scalar(@$required) - 2] )
.' or --'.$self->{opts}->{$required->[-1]}->{long};
$self->save_error("Specify at least one of $err");
}
}
$self->_check_opts( keys %{$self->{opts}} );
$self->{got_opts} = 1;
return;
}
sub _check_opts {
my ( $self, @long ) = @_;
my $long_last = scalar @long;
while ( @long ) {
foreach my $i ( 0..$#long ) {
my $long = $long[$i];
next unless $long;
my $opt = $self->{opts}->{$long};
if ( $opt->{got} ) {
if ( exists $self->{disables}->{$long} ) {
my @disable_opts = @{$self->{disables}->{$long}};
map { $self->{opts}->{$_}->{value} = undef; } @disable_opts;
PTDEBUG && _d('Unset options', @disable_opts,
'because', $long,'disables them');
}
if ( exists $self->{allowed_groups}->{$long} ) {
my @restricted_groups = grep {
!exists $self->{allowed_groups}->{$long}->{$_}
} keys %{$self->{groups}};
my @restricted_opts;
foreach my $restricted_group ( @restricted_groups ) {
RESTRICTED_OPT:
foreach my $restricted_opt (
keys %{$self->{groups}->{$restricted_group}} )
{
next RESTRICTED_OPT if $restricted_opt eq $long;
push @restricted_opts, $restricted_opt
if $self->{opts}->{$restricted_opt}->{got};
}
}
if ( @restricted_opts ) {
my $err;
if ( @restricted_opts == 1 ) {
$err = "--$restricted_opts[0]";
}
else {
$err = join(', ',
map { "--$self->{opts}->{$_}->{long}" }
grep { $_ }
@restricted_opts[0..scalar(@restricted_opts) - 2]
)
. ' or --'.$self->{opts}->{$restricted_opts[-1]}->{long};
}
$self->save_error("--$long is not allowed with $err");
}
}
}
elsif ( $opt->{is_required} ) {
$self->save_error("Required option --$long must be specified");
}
$self->_validate_type($opt);
if ( $opt->{parsed} ) {
delete $long[$i];
}
else {
PTDEBUG && _d('Temporarily failed to parse', $long);
}
}
die "Failed to parse options, possibly due to circular dependencies"
if @long == $long_last;
$long_last = @long;
}
return;
}
sub _validate_type {
my ( $self, $opt ) = @_;
return unless $opt;
if ( !$opt->{type} ) {
$opt->{parsed} = 1;
return;
}
my $val = $opt->{value};
if ( $val && $opt->{type} eq 'm' ) { # type time
PTDEBUG && _d('Parsing option', $opt->{long}, 'as a time value');
my ( $prefix, $num, $suffix ) = $val =~ m/([+-]?)(\d+)([a-z])?$/;
if ( !$suffix ) {
my ( $s ) = $opt->{desc} =~ m/\(suffix (.)\)/;
$suffix = $s || 's';
PTDEBUG && _d('No suffix given; using', $suffix, 'for',
$opt->{long}, '(value:', $val, ')');
}
if ( $suffix =~ m/[smhd]/ ) {
$val = $suffix eq 's' ? $num # Seconds
: $suffix eq 'm' ? $num * 60 # Minutes
: $suffix eq 'h' ? $num * 3600 # Hours
: $num * 86400; # Days
$opt->{value} = ($prefix || '') . $val;
PTDEBUG && _d('Setting option', $opt->{long}, 'to', $val);
}
else {
$self->save_error("Invalid time suffix for --$opt->{long}");
}
}
elsif ( $val && $opt->{type} eq 'd' ) { # type DSN
PTDEBUG && _d('Parsing option', $opt->{long}, 'as a DSN');
my $prev = {};
my $from_key = $self->{defaults_to}->{ $opt->{long} };
if ( $from_key ) {
PTDEBUG && _d($opt->{long}, 'DSN copies from', $from_key, 'DSN');
if ( $self->{opts}->{$from_key}->{parsed} ) {
$prev = $self->{opts}->{$from_key}->{value};
}
else {
PTDEBUG && _d('Cannot parse', $opt->{long}, 'until',
$from_key, 'parsed');
return;
}
}
my $defaults = $self->{DSNParser}->parse_options($self);
$opt->{value} = $self->{DSNParser}->parse($val, $prev, $defaults);
}
elsif ( $val && $opt->{type} eq 'z' ) { # type size
PTDEBUG && _d('Parsing option', $opt->{long}, 'as a size value');
$self->_parse_size($opt, $val);
}
elsif ( $opt->{type} eq 'H' || (defined $val && $opt->{type} eq 'h') ) {
$opt->{value} = { map { $_ => 1 } split(/(?<!\\),\s*/, ($val || '')) };
}
elsif ( $opt->{type} eq 'A' || (defined $val && $opt->{type} eq 'a') ) {
$opt->{value} = [ split(/(?<!\\),\s*/, ($val || '')) ];
}
else {
PTDEBUG && _d('Nothing to validate for option',
$opt->{long}, 'type', $opt->{type}, 'value', $val);
}
$opt->{parsed} = 1;
return;
}
sub get {
my ( $self, $opt ) = @_;
my $long = (length $opt == 1 ? $self->{short_opts}->{$opt} : $opt);
die "Option $opt does not exist"
unless $long && exists $self->{opts}->{$long};
return $self->{opts}->{$long}->{value};
}
sub got {
my ( $self, $opt ) = @_;
my $long = (length $opt == 1 ? $self->{short_opts}->{$opt} : $opt);
die "Option $opt does not exist"
unless $long && exists $self->{opts}->{$long};
return $self->{opts}->{$long}->{got};
}
sub has {
my ( $self, $opt ) = @_;
my $long = (length $opt == 1 ? $self->{short_opts}->{$opt} : $opt);
return defined $long ? exists $self->{opts}->{$long} : 0;
}
sub set {
my ( $self, $opt, $val ) = @_;
my $long = (length $opt == 1 ? $self->{short_opts}->{$opt} : $opt);
die "Option $opt does not exist"
unless $long && exists $self->{opts}->{$long};
$self->{opts}->{$long}->{value} = $val;
return;
}
sub save_error {
my ( $self, $error ) = @_;
push @{$self->{errors}}, $error;
return;
}
sub errors {
my ( $self ) = @_;
return $self->{errors};
}
sub usage {
my ( $self ) = @_;
warn "No usage string is set" unless $self->{usage}; # XXX
return "Usage: " . ($self->{usage} || '') . "\n";
}
sub descr {
my ( $self ) = @_;
warn "No description string is set" unless $self->{description}; # XXX
my $descr = ($self->{description} || $self->{program_name} || '')
. " For more details, please use the --help option, "
. "or try 'perldoc $PROGRAM_NAME' "
. "for complete documentation.";
$descr = join("\n", $descr =~ m/(.{0,80})(?:\s+|$)/g)
unless $ENV{DONT_BREAK_LINES};
$descr =~ s/ +$//mg;
return $descr;
}
sub usage_or_errors {
my ( $self, $file, $return ) = @_;
$file ||= $self->{file} || __FILE__;
if ( !$self->{description} || !$self->{usage} ) {
PTDEBUG && _d("Getting description and usage from SYNOPSIS in", $file);
my %synop = $self->_parse_synopsis($file);
$self->{description} ||= $synop{description};
$self->{usage} ||= $synop{usage};
PTDEBUG && _d("Description:", $self->{description},
"\nUsage:", $self->{usage});
}
if ( $self->{opts}->{help}->{got} ) {
print $self->print_usage() or die "Cannot print usage: $OS_ERROR";
exit 0 unless $return;
}
elsif ( scalar @{$self->{errors}} ) {
print $self->print_errors() or die "Cannot print errors: $OS_ERROR";
exit 0 unless $return;
}
return;
}
sub print_errors {
my ( $self ) = @_;
my $usage = $self->usage() . "\n";
if ( (my @errors = @{$self->{errors}}) ) {
$usage .= join("\n * ", 'Errors in command-line arguments:', @errors)
. "\n";
}
return $usage . "\n" . $self->descr();
}
sub print_usage {
my ( $self ) = @_;
die "Run get_opts() before print_usage()" unless $self->{got_opts};
my @opts = values %{$self->{opts}};
my $maxl = max(
map {
length($_->{long}) # option long name
+ ($_->{is_negatable} ? 4 : 0) # "[no]" if opt is negatable
+ ($_->{type} ? 2 : 0) # "=x" where x is the opt type
}
@opts);
my $maxs = max(0,
map {
length($_)
+ ($self->{opts}->{$_}->{is_negatable} ? 4 : 0)
+ ($self->{opts}->{$_}->{type} ? 2 : 0)
}
values %{$self->{short_opts}});
my $lcol = max($maxl, ($maxs + 3));
my $rcol = 80 - $lcol - 6;
my $rpad = ' ' x ( 80 - $rcol );
$maxs = max($lcol - 3, $maxs);
my $usage = $self->descr() . "\n" . $self->usage();
my @groups = reverse sort grep { $_ ne 'default'; } keys %{$self->{groups}};
push @groups, 'default';
foreach my $group ( reverse @groups ) {
$usage .= "\n".($group eq 'default' ? 'Options' : $group).":\n\n";
foreach my $opt (
sort { $a->{long} cmp $b->{long} }
grep { $_->{group} eq $group }
@opts )
{
my $long = $opt->{is_negatable} ? "[no]$opt->{long}" : $opt->{long};
my $short = $opt->{short};
my $desc = $opt->{desc};
$long .= $opt->{type} ? "=$opt->{type}" : "";
if ( $opt->{type} && $opt->{type} eq 'm' ) {
my ($s) = $desc =~ m/\(suffix (.)\)/;
$s ||= 's';
$desc =~ s/\s+\(suffix .\)//;
$desc .= ". Optional suffix s=seconds, m=minutes, h=hours, "
. "d=days; if no suffix, $s is used.";
}
$desc = join("\n$rpad", grep { $_ } $desc =~ m/(.{0,$rcol})(?:\s+|$)/g);
$desc =~ s/ +$//mg;
if ( $short ) {
$usage .= sprintf(" --%-${maxs}s -%s %s\n", $long, $short, $desc);
}
else {
$usage .= sprintf(" --%-${lcol}s %s\n", $long, $desc);
}
}
}
$usage .= "\nOption types: s=string, i=integer, f=float, h/H/a/A=comma-separated list, d=DSN, z=size, m=time\n";
if ( (my @rules = @{$self->{rules}}) ) {
$usage .= "\nRules:\n\n";
$usage .= join("\n", map { " $_" } @rules) . "\n";
}
if ( $self->{DSNParser} ) {
$usage .= "\n" . $self->{DSNParser}->usage();
}
$usage .= "\nOptions and values after processing arguments:\n\n";
foreach my $opt ( sort { $a->{long} cmp $b->{long} } @opts ) {
my $val = $opt->{value};
my $type = $opt->{type} || '';
my $bool = $opt->{spec} =~ m/^[\w-]+(?:\|[\w-])?!?$/;
$val = $bool ? ( $val ? 'TRUE' : 'FALSE' )
: !defined $val ? '(No value)'
: $type eq 'd' ? $self->{DSNParser}->as_string($val)
: $type =~ m/H|h/ ? join(',', sort keys %$val)
: $type =~ m/A|a/ ? join(',', @$val)
: $val;
$usage .= sprintf(" --%-${lcol}s %s\n", $opt->{long}, $val);
}
return $usage;
}
sub prompt_noecho {
shift @_ if ref $_[0] eq __PACKAGE__;
my ( $prompt ) = @_;
local $OUTPUT_AUTOFLUSH = 1;
print $prompt
or die "Cannot print: $OS_ERROR";
my $response;
eval {
require Term::ReadKey;
Term::ReadKey::ReadMode('noecho');
chomp($response = <STDIN>);
Term::ReadKey::ReadMode('normal');
print "\n"
or die "Cannot print: $OS_ERROR";
};
if ( $EVAL_ERROR ) {
die "Cannot read response; is Term::ReadKey installed? $EVAL_ERROR";
}
return $response;
}
sub _read_config_file {
my ( $self, $filename ) = @_;
open my $fh, "<", $filename or die "Cannot open $filename: $OS_ERROR\n";
my @args;
my $prefix = '--';
my $parse = 1;
LINE:
while ( my $line = <$fh> ) {
chomp $line;
next LINE if $line =~ m/^\s*(?:\#|\;|$)/;
$line =~ s/\s+#.*$//g;
$line =~ s/^\s+|\s+$//g;
if ( $line eq '--' ) {
$prefix = '';
$parse = 0;
next LINE;
}
if ( $parse
&& (my($opt, $arg) = $line =~ m/^\s*([^=\s]+?)(?:\s*=\s*(.*?)\s*)?$/)
) {
push @args, grep { defined $_ } ("$prefix$opt", $arg);
}
elsif ( $line =~ m/./ ) {
push @args, $line;
}
else {
die "Syntax error in file $filename at line $INPUT_LINE_NUMBER";
}
}
close $fh;
return @args;
}
sub read_para_after {
my ( $self, $file, $regex ) = @_;
open my $fh, "<", $file or die "Can't open $file: $OS_ERROR";
local $INPUT_RECORD_SEPARATOR = '';
my $para;
while ( $para = <$fh> ) {
next unless $para =~ m/^=pod$/m;
last;
}
while ( $para = <$fh> ) {
next unless $para =~ m/$regex/;
last;
}
$para = <$fh>;
chomp($para);
close $fh or die "Can't close $file: $OS_ERROR";
return $para;
}
sub clone {
my ( $self ) = @_;
my %clone = map {
my $hashref = $self->{$_};
my $val_copy = {};
foreach my $key ( keys %$hashref ) {
my $ref = ref $hashref->{$key};
$val_copy->{$key} = !$ref ? $hashref->{$key}
: $ref eq 'HASH' ? { %{$hashref->{$key}} }
: $ref eq 'ARRAY' ? [ @{$hashref->{$key}} ]
: $hashref->{$key};
}
$_ => $val_copy;
} qw(opts short_opts defaults);
foreach my $scalar ( qw(got_opts) ) {
$clone{$scalar} = $self->{$scalar};
}
return bless \%clone;
}
sub _parse_size {
my ( $self, $opt, $val ) = @_;
if ( lc($val || '') eq 'null' ) {
PTDEBUG && _d('NULL size for', $opt->{long});
$opt->{value} = 'null';
return;
}
my %factor_for = (k => 1_024, M => 1_048_576, G => 1_073_741_824);
my ($pre, $num, $factor) = $val =~ m/^([+-])?(\d+)([kMG])?$/;
if ( defined $num ) {
if ( $factor ) {
$num *= $factor_for{$factor};
PTDEBUG && _d('Setting option', $opt->{y},
'to num', $num, '* factor', $factor);
}
$opt->{value} = ($pre || '') . $num;
}
else {
$self->save_error("Invalid size for --$opt->{long}: $val");
}
return;
}
sub _parse_attribs {
my ( $self, $option, $attribs ) = @_;
my $types = $self->{types};
return $option
. ($attribs->{'short form'} ? '|' . $attribs->{'short form'} : '' )
. ($attribs->{'negatable'} ? '!' : '' )
. ($attribs->{'cumulative'} ? '+' : '' )
. ($attribs->{'type'} ? '=' . $types->{$attribs->{type}} : '' );
}
sub _parse_synopsis {
my ( $self, $file ) = @_;
$file ||= $self->{file} || __FILE__;
PTDEBUG && _d("Parsing SYNOPSIS in", $file);
local $INPUT_RECORD_SEPARATOR = ''; # read paragraphs
open my $fh, "<", $file or die "Cannot open $file: $OS_ERROR";
my $para;
1 while defined($para = <$fh>) && $para !~ m/^=head1 SYNOPSIS/;
die "$file does not contain a SYNOPSIS section" unless $para;
my @synop;
for ( 1..2 ) { # 1 for the usage, 2 for the description
my $para = <$fh>;
push @synop, $para;
}
close $fh;
PTDEBUG && _d("Raw SYNOPSIS text:", @synop);
my ($usage, $desc) = @synop;
die "The SYNOPSIS section in $file is not formatted properly"
unless $usage && $desc;
$usage =~ s/^\s*Usage:\s+(.+)/$1/;
chomp $usage;
$desc =~ s/\n/ /g;
$desc =~ s/\s{2,}/ /g;
$desc =~ s/\. ([A-Z][a-z])/. $1/g;
$desc =~ s/\s+$//;
return (
description => $desc,
usage => $usage,
);
};
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
if ( PTDEBUG ) {
print '# ', $^X, ' ', $], "\n";
if ( my $uname = `uname -a` ) {
$uname =~ s/\s+/ /g;
print "# $uname\n";
}
print '# Arguments: ',
join(' ', map { my $a = "_[$_]_"; $a =~ s/\n/\n# /g; $a; } @ARGV), "\n";
}
1;
}
# ###########################################################################
# End OptionParser package
# ###########################################################################
# ###########################################################################
# VersionParser package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/VersionParser.pm
# t/lib/VersionParser.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package VersionParser;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class ) = @_;
bless {}, $class;
}
sub parse {
my ( $self, $str ) = @_;
my $result = sprintf('%03d%03d%03d', $str =~ m/(\d+)/g);
PTDEBUG && _d($str, 'parses to', $result);
return $result;
}
sub version_ge {
my ( $self, $dbh, $target ) = @_;
if ( !$self->{$dbh} ) {
$self->{$dbh} = $self->parse(
$dbh->selectrow_array('SELECT VERSION()'));
}
my $result = $self->{$dbh} ge $self->parse($target) ? 1 : 0;
PTDEBUG && _d($self->{$dbh}, 'ge', $target, ':', $result);
return $result;
}
sub innodb_version {
my ( $self, $dbh ) = @_;
return unless $dbh;
my $innodb_version = "NO";
my ($innodb) =
grep { $_->{engine} =~ m/InnoDB/i }
map {
my %hash;
@hash{ map { lc $_ } keys %$_ } = values %$_;
\%hash;
}
@{ $dbh->selectall_arrayref("SHOW ENGINES", {Slice=>{}}) };
if ( $innodb ) {
PTDEBUG && _d("InnoDB support:", $innodb->{support});
if ( $innodb->{support} =~ m/YES|DEFAULT/i ) {
my $vars = $dbh->selectrow_hashref(
"SHOW VARIABLES LIKE 'innodb_version'");
$innodb_version = !$vars ? "BUILTIN"
: ($vars->{Value} || $vars->{value});
}
else {
$innodb_version = $innodb->{support}; # probably DISABLED or NO
}
}
PTDEBUG && _d("InnoDB version:", $innodb_version);
return $innodb_version;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End VersionParser package
# ###########################################################################
# ###########################################################################
# DSNParser package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/DSNParser.pm
# t/lib/DSNParser.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package DSNParser;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Data::Dumper;
$Data::Dumper::Indent = 0;
$Data::Dumper::Quotekeys = 0;
my $dsn_sep = qr/(?<!\\),/;
eval {
require DBI;
};
my $have_dbi = $EVAL_ERROR ? 0 : 1;
sub new {
my ( $class, %args ) = @_;
foreach my $arg ( qw(opts) ) {
die "I need a $arg argument" unless $args{$arg};
}
my $self = {
opts => {} # h, P, u, etc. Should come from DSN OPTIONS section in POD.
};
foreach my $opt ( @{$args{opts}} ) {
if ( !$opt->{key} || !$opt->{desc} ) {
die "Invalid DSN option: ", Dumper($opt);
}
PTDEBUG && _d('DSN option:',
join(', ',
map { "$_=" . (defined $opt->{$_} ? ($opt->{$_} || '') : 'undef') }
keys %$opt
)
);
$self->{opts}->{$opt->{key}} = {
dsn => $opt->{dsn},
desc => $opt->{desc},
copy => $opt->{copy} || 0,
};
}
return bless $self, $class;
}
sub prop {
my ( $self, $prop, $value ) = @_;
if ( @_ > 2 ) {
PTDEBUG && _d('Setting', $prop, 'property');
$self->{$prop} = $value;
}
return $self->{$prop};
}
sub parse {
my ( $self, $dsn, $prev, $defaults ) = @_;
if ( !$dsn ) {
PTDEBUG && _d('No DSN to parse');
return;
}
PTDEBUG && _d('Parsing', $dsn);
$prev ||= {};
$defaults ||= {};
my %given_props;
my %final_props;
my $opts = $self->{opts};
foreach my $dsn_part ( split($dsn_sep, $dsn) ) {
$dsn_part =~ s/\\,/,/g;
if ( my ($prop_key, $prop_val) = $dsn_part =~ m/^(.)=(.*)$/ ) {
$given_props{$prop_key} = $prop_val;
}
else {
PTDEBUG && _d('Interpreting', $dsn_part, 'as h=', $dsn_part);
$given_props{h} = $dsn_part;
}
}
foreach my $key ( keys %$opts ) {
PTDEBUG && _d('Finding value for', $key);
$final_props{$key} = $given_props{$key};
if ( !defined $final_props{$key}
&& defined $prev->{$key} && $opts->{$key}->{copy} )
{
$final_props{$key} = $prev->{$key};
PTDEBUG && _d('Copying value for', $key, 'from previous DSN');
}
if ( !defined $final_props{$key} ) {
$final_props{$key} = $defaults->{$key};
PTDEBUG && _d('Copying value for', $key, 'from defaults');
}
}
foreach my $key ( keys %given_props ) {
die "Unknown DSN option '$key' in '$dsn'. For more details, "
. "please use the --help option, or try 'perldoc $PROGRAM_NAME' "
. "for complete documentation."
unless exists $opts->{$key};
}
if ( (my $required = $self->prop('required')) ) {
foreach my $key ( keys %$required ) {
die "Missing required DSN option '$key' in '$dsn'. For more details, "
. "please use the --help option, or try 'perldoc $PROGRAM_NAME' "
. "for complete documentation."
unless $final_props{$key};
}
}
return \%final_props;
}
sub parse_options {
my ( $self, $o ) = @_;
die 'I need an OptionParser object' unless ref $o eq 'OptionParser';
my $dsn_string
= join(',',
map { "$_=".$o->get($_); }
grep { $o->has($_) && $o->get($_) }
keys %{$self->{opts}}
);
PTDEBUG && _d('DSN string made from options:', $dsn_string);
return $self->parse($dsn_string);
}
sub as_string {
my ( $self, $dsn, $props ) = @_;
return $dsn unless ref $dsn;
my @keys = $props ? @$props : sort keys %$dsn;
return join(',',
map { "$_=" . ($_ eq 'p' ? '...' : $dsn->{$_}) }
grep {
exists $self->{opts}->{$_}
&& exists $dsn->{$_}
&& defined $dsn->{$_}
} @keys);
}
sub usage {
my ( $self ) = @_;
my $usage
= "DSN syntax is key=value[,key=value...] Allowable DSN keys:\n\n"
. " KEY COPY MEANING\n"
. " === ==== =============================================\n";
my %opts = %{$self->{opts}};
foreach my $key ( sort keys %opts ) {
$usage .= " $key "
. ($opts{$key}->{copy} ? 'yes ' : 'no ')
. ($opts{$key}->{desc} || '[No description]')
. "\n";
}
$usage .= "\n If the DSN is a bareword, the word is treated as the 'h' key.\n";
return $usage;
}
sub get_cxn_params {
my ( $self, $info ) = @_;
my $dsn;
my %opts = %{$self->{opts}};
my $driver = $self->prop('dbidriver') || '';
if ( $driver eq 'Pg' ) {
$dsn = 'DBI:Pg:dbname=' . ( $info->{D} || '' ) . ';'
. join(';', map { "$opts{$_}->{dsn}=$info->{$_}" }
grep { defined $info->{$_} }
qw(h P));
}
else {
$dsn = 'DBI:mysql:' . ( $info->{D} || '' ) . ';'
. join(';', map { "$opts{$_}->{dsn}=$info->{$_}" }
grep { defined $info->{$_} }
qw(F h P S A))
. ';mysql_read_default_group=client';
}
PTDEBUG && _d($dsn);
return ($dsn, $info->{u}, $info->{p});
}
sub fill_in_dsn {
my ( $self, $dbh, $dsn ) = @_;
my $vars = $dbh->selectall_hashref('SHOW VARIABLES', 'Variable_name');
my ($user, $db) = $dbh->selectrow_array('SELECT USER(), DATABASE()');
$user =~ s/@.*//;
$dsn->{h} ||= $vars->{hostname}->{Value};
$dsn->{S} ||= $vars->{'socket'}->{Value};
$dsn->{P} ||= $vars->{port}->{Value};
$dsn->{u} ||= $user;
$dsn->{D} ||= $db;
}
sub get_dbh {
my ( $self, $cxn_string, $user, $pass, $opts ) = @_;
$opts ||= {};
my $defaults = {
AutoCommit => 0,
RaiseError => 1,
PrintError => 0,
ShowErrorStatement => 1,
mysql_enable_utf8 => ($cxn_string =~ m/charset=utf8/i ? 1 : 0),
};
@{$defaults}{ keys %$opts } = values %$opts;
if ( $opts->{mysql_use_result} ) {
$defaults->{mysql_use_result} = 1;
}
if ( !$have_dbi ) {
die "Cannot connect to MySQL because the Perl DBI module is not "
. "installed or not found. Run 'perl -MDBI' to see the directories "
. "that Perl searches for DBI. If DBI is not installed, try:\n"
. " Debian/Ubuntu apt-get install libdbi-perl\n"
. " RHEL/CentOS yum install perl-DBI\n"
. " OpenSolaris pgk install pkg:/SUNWpmdbi\n";
}
my $dbh;
my $tries = 2;
while ( !$dbh && $tries-- ) {
PTDEBUG && _d($cxn_string, ' ', $user, ' ', $pass,
join(', ', map { "$_=>$defaults->{$_}" } keys %$defaults ));
eval {
$dbh = DBI->connect($cxn_string, $user, $pass, $defaults);
if ( $cxn_string =~ m/mysql/i ) {
my $sql;
$sql = 'SELECT @@SQL_MODE';
PTDEBUG && _d($dbh, $sql);
my ($sql_mode) = $dbh->selectrow_array($sql);
$sql = 'SET @@SQL_QUOTE_SHOW_CREATE = 1'
. '/*!40101, @@SQL_MODE=\'NO_AUTO_VALUE_ON_ZERO'
. ($sql_mode ? ",$sql_mode" : '')
. '\'*/';
PTDEBUG && _d($dbh, $sql);
$dbh->do($sql);
if ( my ($charset) = $cxn_string =~ m/charset=(\w+)/ ) {
$sql = "/*!40101 SET NAMES $charset*/";
PTDEBUG && _d($dbh, ':', $sql);
$dbh->do($sql);
PTDEBUG && _d('Enabling charset for STDOUT');
if ( $charset eq 'utf8' ) {
binmode(STDOUT, ':utf8')
or die "Can't binmode(STDOUT, ':utf8'): $OS_ERROR";
}
else {
binmode(STDOUT) or die "Can't binmode(STDOUT): $OS_ERROR";
}
}
if ( $self->prop('set-vars') ) {
$sql = "SET " . $self->prop('set-vars');
PTDEBUG && _d($dbh, ':', $sql);
$dbh->do($sql);
}
}
};
if ( !$dbh && $EVAL_ERROR ) {
PTDEBUG && _d($EVAL_ERROR);
if ( $EVAL_ERROR =~ m/not a compiled character set|character set utf8/ ) {
PTDEBUG && _d('Going to try again without utf8 support');
delete $defaults->{mysql_enable_utf8};
}
elsif ( $EVAL_ERROR =~ m/locate DBD\/mysql/i ) {
die "Cannot connect to MySQL because the Perl DBD::mysql module is "
. "not installed or not found. Run 'perl -MDBD::mysql' to see "
. "the directories that Perl searches for DBD::mysql. If "
. "DBD::mysql is not installed, try:\n"
. " Debian/Ubuntu apt-get install libdbd-mysql-perl\n"
. " RHEL/CentOS yum install perl-DBD-MySQL\n"
. " OpenSolaris pgk install pkg:/SUNWapu13dbd-mysql\n";
}
if ( !$tries ) {
die $EVAL_ERROR;
}
}
}
PTDEBUG && _d('DBH info: ',
$dbh,
Dumper($dbh->selectrow_hashref(
'SELECT DATABASE(), CONNECTION_ID(), VERSION()/*!50038 , @@hostname*/')),
'Connection info:', $dbh->{mysql_hostinfo},
'Character set info:', Dumper($dbh->selectall_arrayref(
'SHOW VARIABLES LIKE "character_set%"', { Slice => {}})),
'$DBD::mysql::VERSION:', $DBD::mysql::VERSION,
'$DBI::VERSION:', $DBI::VERSION,
);
return $dbh;
}
sub get_hostname {
my ( $self, $dbh ) = @_;
if ( my ($host) = ($dbh->{mysql_hostinfo} || '') =~ m/^(\w+) via/ ) {
return $host;
}
my ( $hostname, $one ) = $dbh->selectrow_array(
'SELECT /*!50038 @@hostname, */ 1');
return $hostname;
}
sub disconnect {
my ( $self, $dbh ) = @_;
PTDEBUG && $self->print_active_handles($dbh);
$dbh->disconnect;
}
sub print_active_handles {
my ( $self, $thing, $level ) = @_;
$level ||= 0;
printf("# Active %sh: %s %s %s\n", ($thing->{Type} || 'undef'), "\t" x $level,
$thing, (($thing->{Type} || '') eq 'st' ? $thing->{Statement} || '' : ''))
or die "Cannot print: $OS_ERROR";
foreach my $handle ( grep {defined} @{ $thing->{ChildHandles} } ) {
$self->print_active_handles( $handle, $level + 1 );
}
}
sub copy {
my ( $self, $dsn_1, $dsn_2, %args ) = @_;
die 'I need a dsn_1 argument' unless $dsn_1;
die 'I need a dsn_2 argument' unless $dsn_2;
my %new_dsn = map {
my $key = $_;
my $val;
if ( $args{overwrite} ) {
$val = defined $dsn_1->{$key} ? $dsn_1->{$key} : $dsn_2->{$key};
}
else {
$val = defined $dsn_2->{$key} ? $dsn_2->{$key} : $dsn_1->{$key};
}
$key => $val;
} keys %{$self->{opts}};
return \%new_dsn;
}
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;
}
# ###########################################################################
# End DSNParser package
# ###########################################################################
# ###########################################################################
# Daemon package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Daemon.pm
# t/lib/Daemon.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Daemon;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use POSIX qw(setsid);
sub new {
my ( $class, %args ) = @_;
foreach my $arg ( qw(o) ) {
die "I need a $arg argument" unless $args{$arg};
}
my $o = $args{o};
my $self = {
o => $o,
log_file => $o->has('log') ? $o->get('log') : undef,
PID_file => $o->has('pid') ? $o->get('pid') : undef,
};
check_PID_file(undef, $self->{PID_file});
PTDEBUG && _d('Daemonized child will log to', $self->{log_file});
return bless $self, $class;
}
sub daemonize {
my ( $self ) = @_;
PTDEBUG && _d('About to fork and daemonize');
defined (my $pid = fork()) or die "Cannot fork: $OS_ERROR";
if ( $pid ) {
PTDEBUG && _d('Parent PID', $PID, 'exiting after forking child PID',$pid);
exit;
}
PTDEBUG && _d('Daemonizing child PID', $PID);
$self->{PID_owner} = $PID;
$self->{child} = 1;
POSIX::setsid() or die "Cannot start a new session: $OS_ERROR";
chdir '/' or die "Cannot chdir to /: $OS_ERROR";
$self->_make_PID_file();
$OUTPUT_AUTOFLUSH = 1;
PTDEBUG && _d('Redirecting STDIN to /dev/null');
close STDIN;
open STDIN, '/dev/null'
or die "Cannot reopen STDIN to /dev/null: $OS_ERROR";
if ( $self->{log_file} ) {
PTDEBUG && _d('Redirecting STDOUT and STDERR to', $self->{log_file});
close STDOUT;
open STDOUT, '>>', $self->{log_file}
or die "Cannot open log file $self->{log_file}: $OS_ERROR";
close STDERR;
open STDERR, ">&STDOUT"
or die "Cannot dupe STDERR to STDOUT: $OS_ERROR";
}
else {
if ( -t STDOUT ) {
PTDEBUG && _d('No log file and STDOUT is a terminal;',
'redirecting to /dev/null');
close STDOUT;
open STDOUT, '>', '/dev/null'
or die "Cannot reopen STDOUT to /dev/null: $OS_ERROR";
}
if ( -t STDERR ) {
PTDEBUG && _d('No log file and STDERR is a terminal;',
'redirecting to /dev/null');
close STDERR;
open STDERR, '>', '/dev/null'
or die "Cannot reopen STDERR to /dev/null: $OS_ERROR";
}
}
return;
}
sub check_PID_file {
my ( $self, $file ) = @_;
my $PID_file = $self ? $self->{PID_file} : $file;
PTDEBUG && _d('Checking PID file', $PID_file);
if ( $PID_file && -f $PID_file ) {
my $pid;
eval { chomp($pid = `cat $PID_file`); };
die "Cannot cat $PID_file: $OS_ERROR" if $EVAL_ERROR;
PTDEBUG && _d('PID file exists; it contains PID', $pid);
if ( $pid ) {
my $pid_is_alive = kill 0, $pid;
if ( $pid_is_alive ) {
die "The PID file $PID_file already exists "
. " and the PID that it contains, $pid, is running";
}
else {
warn "Overwriting PID file $PID_file because the PID that it "
. "contains, $pid, is not running";
}
}
else {
die "The PID file $PID_file already exists but it does not "
. "contain a PID";
}
}
else {
PTDEBUG && _d('No PID file');
}
return;
}
sub make_PID_file {
my ( $self ) = @_;
if ( exists $self->{child} ) {
die "Do not call Daemon::make_PID_file() for daemonized scripts";
}
$self->_make_PID_file();
$self->{PID_owner} = $PID;
return;
}
sub _make_PID_file {
my ( $self ) = @_;
my $PID_file = $self->{PID_file};
if ( !$PID_file ) {
PTDEBUG && _d('No PID file to create');
return;
}
$self->check_PID_file();
open my $PID_FH, '>', $PID_file
or die "Cannot open PID file $PID_file: $OS_ERROR";
print $PID_FH $PID
or die "Cannot print to PID file $PID_file: $OS_ERROR";
close $PID_FH
or die "Cannot close PID file $PID_file: $OS_ERROR";
PTDEBUG && _d('Created PID file:', $self->{PID_file});
return;
}
sub _remove_PID_file {
my ( $self ) = @_;
if ( $self->{PID_file} && -f $self->{PID_file} ) {
unlink $self->{PID_file}
or warn "Cannot remove PID file $self->{PID_file}: $OS_ERROR";
PTDEBUG && _d('Removed PID file');
}
else {
PTDEBUG && _d('No PID to remove');
}
return;
}
sub DESTROY {
my ( $self ) = @_;
$self->_remove_PID_file() if ($self->{PID_owner} || 0) == $PID;
return;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End Daemon package
# ###########################################################################
# ###########################################################################
# Quoter package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Quoter.pm
# t/lib/Quoter.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Quoter;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
return bless {}, $class;
}
sub quote {
my ( $self, @vals ) = @_;
foreach my $val ( @vals ) {
$val =~ s/`/``/g;
}
return join('.', map { '`' . $_ . '`' } @vals);
}
sub quote_val {
my ( $self, $val ) = @_;
return 'NULL' unless defined $val; # undef = NULL
return "''" if $val eq ''; # blank string = ''
return $val if $val =~ m/^0x[0-9a-fA-F]+$/; # hex data
$val =~ s/(['\\])/\\$1/g;
return "'$val'";
}
sub split_unquote {
my ( $self, $db_tbl, $default_db ) = @_;
$db_tbl =~ s/`//g;
my ( $db, $tbl ) = split(/[.]/, $db_tbl);
if ( !$tbl ) {
$tbl = $db;
$db = $default_db;
}
return ($db, $tbl);
}
sub literal_like {
my ( $self, $like ) = @_;
return unless $like;
$like =~ s/([%_])/\\$1/g;
return "'$like'";
}
sub join_quote {
my ( $self, $default_db, $db_tbl ) = @_;
return unless $db_tbl;
my ($db, $tbl) = split(/[.]/, $db_tbl);
if ( !$tbl ) {
$tbl = $db;
$db = $default_db;
}
$db = "`$db`" if $db && $db !~ m/^`/;
$tbl = "`$tbl`" if $tbl && $tbl !~ m/^`/;
return $db ? "$db.$tbl" : $tbl;
}
sub serialize_list {
my ( $self, @args ) = @_;
return unless @args;
return $args[0] if @args == 1 && !defined $args[0];
die "Cannot serialize multiple values with undef/NULL"
if grep { !defined $_ } @args;
return join ',', map { quotemeta } @args;
}
sub deserialize_list {
my ( $self, $string ) = @_;
return $string unless defined $string;
my @escaped_parts = $string =~ /
\G # Start of string, or end of previous match.
( # Each of these is an element in the original list.
[^\\,]* # Anything not a backslash or a comma
(?: # When we get here, we found one of the above.
\\. # A backslash followed by something so we can continue
[^\\,]* # Same as above.
)* # Repeat zero of more times.
)
, # Comma dividing elements
/sxgc;
push @escaped_parts, pos($string) ? substr( $string, pos($string) ) : $string;
my @unescaped_parts = map {
my $part = $_;
my $char_class = utf8::is_utf8($part) # If it's a UTF-8 string,
? qr/(?=\p{ASCII})\W/ # We only care about non-word
: qr/(?=\p{ASCII})\W|[\x{80}-\x{FF}]/; # Otherwise,
$part =~ s/\\($char_class)/$1/g;
$part;
} @escaped_parts;
return @unescaped_parts;
}
1;
}
# ###########################################################################
# End Quoter package
# ###########################################################################
# ###########################################################################
# TableNibbler package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/TableNibbler.pm
# t/lib/TableNibbler.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package TableNibbler;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(TableParser Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my $self = { %args };
return bless $self, $class;
}
sub generate_asc_stmt {
my ( $self, %args ) = @_;
my @required_args = qw(tbl_struct index);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my ($tbl_struct, $index) = @args{@required_args};
my @cols = $args{cols} ? @{$args{cols}} : @{$tbl_struct->{cols}};
my $q = $self->{Quoter};
die "Index '$index' does not exist in table"
unless exists $tbl_struct->{keys}->{$index};
PTDEBUG && _d('Will ascend index', $index);
my @asc_cols = @{$tbl_struct->{keys}->{$index}->{cols}};
if ( $args{asc_first} ) {
@asc_cols = $asc_cols[0];
PTDEBUG && _d('Ascending only first column');
}
PTDEBUG && _d('Will ascend columns', join(', ', @asc_cols));
my @asc_slice;
my %col_posn = do { my $i = 0; map { $_ => $i++ } @cols };
foreach my $col ( @asc_cols ) {
if ( !exists $col_posn{$col} ) {
push @cols, $col;
$col_posn{$col} = $#cols;
}
push @asc_slice, $col_posn{$col};
}
PTDEBUG && _d('Will ascend, in ordinal position:', join(', ', @asc_slice));
my $asc_stmt = {
cols => \@cols,
index => $index,
where => '',
slice => [],
scols => [],
};
if ( @asc_slice ) {
my $cmp_where;
foreach my $cmp ( qw(< <= >= >) ) {
$cmp_where = $self->generate_cmp_where(
type => $cmp,
slice => \@asc_slice,
cols => \@cols,
quoter => $q,
is_nullable => $tbl_struct->{is_nullable},
);
$asc_stmt->{boundaries}->{$cmp} = $cmp_where->{where};
}
my $cmp = $args{asc_only} ? '>' : '>=';
$asc_stmt->{where} = $asc_stmt->{boundaries}->{$cmp};
$asc_stmt->{slice} = $cmp_where->{slice};
$asc_stmt->{scols} = $cmp_where->{scols};
}
return $asc_stmt;
}
sub generate_cmp_where {
my ( $self, %args ) = @_;
foreach my $arg ( qw(type slice cols is_nullable) ) {
die "I need a $arg arg" unless defined $args{$arg};
}
my @slice = @{$args{slice}};
my @cols = @{$args{cols}};
my $is_nullable = $args{is_nullable};
my $type = $args{type};
my $q = $self->{Quoter};
(my $cmp = $type) =~ s/=//;
my @r_slice; # Resulting slice columns, by ordinal
my @r_scols; # Ditto, by name
my @clauses;
foreach my $i ( 0 .. $#slice ) {
my @clause;
foreach my $j ( 0 .. $i - 1 ) {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
if ( $is_nullable->{$col} ) {
push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @clause, "$quo = ?";
push @r_slice, $ord;
push @r_scols, $col;
}
}
my $ord = $slice[$i];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
push @clause, "(? IS NULL OR $quo $type ?)";
}
elsif ( $type =~ m/>/ ) {
push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
}
else { # If $type =~ m/</ ) {
push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
}
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @r_slice, $ord;
push @r_scols, $col;
push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
}
push @clauses, '(' . join(' AND ', @clause) . ')';
}
my $result = '(' . join(' OR ', @clauses) . ')';
my $where = {
slice => \@r_slice,
scols => \@r_scols,
where => $result,
};
return $where;
}
sub generate_del_stmt {
my ( $self, %args ) = @_;
my $tbl = $args{tbl_struct};
my @cols = $args{cols} ? @{$args{cols}} : ();
my $tp = $self->{TableParser};
my $q = $self->{Quoter};
my @del_cols;
my @del_slice;
my $index = $tp->find_best_index($tbl, $args{index});
die "Cannot find an ascendable index in table" unless $index;
if ( $index ) {
@del_cols = @{$tbl->{keys}->{$index}->{cols}};
}
else {
@del_cols = @{$tbl->{cols}};
}
PTDEBUG && _d('Columns needed for DELETE:', join(', ', @del_cols));
my %col_posn = do { my $i = 0; map { $_ => $i++ } @cols };
foreach my $col ( @del_cols ) {
if ( !exists $col_posn{$col} ) {
push @cols, $col;
$col_posn{$col} = $#cols;
}
push @del_slice, $col_posn{$col};
}
PTDEBUG && _d('Ordinals needed for DELETE:', join(', ', @del_slice));
my $del_stmt = {
cols => \@cols,
index => $index,
where => '',
slice => [],
scols => [],
};
my @clauses;
foreach my $i ( 0 .. $#del_slice ) {
my $ord = $del_slice[$i];
my $col = $cols[$ord];
my $quo = $q->quote($col);
if ( $tbl->{is_nullable}->{$col} ) {
push @clauses, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @{$del_stmt->{slice}}, $ord, $ord;
push @{$del_stmt->{scols}}, $col, $col;
}
else {
push @clauses, "$quo = ?";
push @{$del_stmt->{slice}}, $ord;
push @{$del_stmt->{scols}}, $col;
}
}
$del_stmt->{where} = '(' . join(' AND ', @clauses) . ')';
return $del_stmt;
}
sub generate_ins_stmt {
my ( $self, %args ) = @_;
foreach my $arg ( qw(ins_tbl sel_cols) ) {
die "I need a $arg argument" unless $args{$arg};
}
my $ins_tbl = $args{ins_tbl};
my @sel_cols = @{$args{sel_cols}};
die "You didn't specify any SELECT columns" unless @sel_cols;
my @ins_cols;
my @ins_slice;
for my $i ( 0..$#sel_cols ) {
next unless $ins_tbl->{is_col}->{$sel_cols[$i]};
push @ins_cols, $sel_cols[$i];
push @ins_slice, $i;
}
return {
cols => \@ins_cols,
slice => \@ins_slice,
};
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End TableNibbler package
# ###########################################################################
# ###########################################################################
# TableParser package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/TableParser.pm
# t/lib/TableParser.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package TableParser;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Data::Dumper;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my $self = { %args };
return bless $self, $class;
}
sub get_create_table {
my ( $self, $dbh, $db, $tbl ) = @_;
die "I need a dbh parameter" unless $dbh;
die "I need a db parameter" unless $db;
die "I need a tbl parameter" unless $tbl;
my $q = $self->{Quoter};
my $new_sql_mode
= '/*!40101 SET @OLD_SQL_MODE := @@SQL_MODE, '
. q{@@SQL_MODE := REPLACE(REPLACE(@@SQL_MODE, 'ANSI_QUOTES', ''), ',,', ','), }
. '@OLD_QUOTE := @@SQL_QUOTE_SHOW_CREATE, '
. '@@SQL_QUOTE_SHOW_CREATE := 1 */';
my $old_sql_mode = '/*!40101 SET @@SQL_MODE := @OLD_SQL_MODE, '
. '@@SQL_QUOTE_SHOW_CREATE := @OLD_QUOTE */';
PTDEBUG && _d($new_sql_mode);
eval { $dbh->do($new_sql_mode); };
PTDEBUG && $EVAL_ERROR && _d($EVAL_ERROR);
my $use_sql = 'USE ' . $q->quote($db);
PTDEBUG && _d($dbh, $use_sql);
$dbh->do($use_sql);
my $show_sql = "SHOW CREATE TABLE " . $q->quote($db, $tbl);
PTDEBUG && _d($show_sql);
my $href;
eval { $href = $dbh->selectrow_hashref($show_sql); };
if ( $EVAL_ERROR ) {
PTDEBUG && _d($EVAL_ERROR);
PTDEBUG && _d($old_sql_mode);
$dbh->do($old_sql_mode);
return;
}
PTDEBUG && _d($old_sql_mode);
$dbh->do($old_sql_mode);
my ($key) = grep { m/create (?:table|view)/i } keys %$href;
if ( !$key ) {
die "Error: no 'Create Table' or 'Create View' in result set from "
. "$show_sql: " . Dumper($href);
}
return $href->{$key};
}
sub parse {
my ( $self, $ddl, $opts ) = @_;
return unless $ddl;
if ( $ddl !~ m/CREATE (?:TEMPORARY )?TABLE `/ ) {
die "Cannot parse table definition; is ANSI quoting "
. "enabled or SQL_QUOTE_SHOW_CREATE disabled?";
}
my ($name) = $ddl =~ m/CREATE (?:TEMPORARY )?TABLE\s+(`.+?`)/;
(undef, $name) = $self->{Quoter}->split_unquote($name) if $name;
$ddl =~ s/(`[^`]+`)/\L$1/g;
my $engine = $self->get_engine($ddl);
my @defs = $ddl =~ m/^(\s+`.*?),?$/gm;
my @cols = map { $_ =~ m/`([^`]+)`/ } @defs;
PTDEBUG && _d('Table cols:', join(', ', map { "`$_`" } @cols));
my %def_for;
@def_for{@cols} = @defs;
my (@nums, @null);
my (%type_for, %is_nullable, %is_numeric, %is_autoinc);
foreach my $col ( @cols ) {
my $def = $def_for{$col};
my ( $type ) = $def =~ m/`[^`]+`\s([a-z]+)/;
die "Can't determine column type for $def" unless $type;
$type_for{$col} = $type;
if ( $type =~ m/(?:(?:tiny|big|medium|small)?int|float|double|decimal|year)/ ) {
push @nums, $col;
$is_numeric{$col} = 1;
}
if ( $def !~ m/NOT NULL/ ) {
push @null, $col;
$is_nullable{$col} = 1;
}
$is_autoinc{$col} = $def =~ m/AUTO_INCREMENT/i ? 1 : 0;
}
my ($keys, $clustered_key) = $self->get_keys($ddl, $opts, \%is_nullable);
my ($charset) = $ddl =~ m/DEFAULT CHARSET=(\w+)/;
return {
name => $name,
cols => \@cols,
col_posn => { map { $cols[$_] => $_ } 0..$#cols },
is_col => { map { $_ => 1 } @cols },
null_cols => \@null,
is_nullable => \%is_nullable,
is_autoinc => \%is_autoinc,
clustered_key => $clustered_key,
keys => $keys,
defs => \%def_for,
numeric_cols => \@nums,
is_numeric => \%is_numeric,
engine => $engine,
type_for => \%type_for,
charset => $charset,
};
}
sub sort_indexes {
my ( $self, $tbl ) = @_;
my @indexes
= sort {
(($a ne 'PRIMARY') <=> ($b ne 'PRIMARY'))
|| ( !$tbl->{keys}->{$a}->{is_unique} <=> !$tbl->{keys}->{$b}->{is_unique} )
|| ( $tbl->{keys}->{$a}->{is_nullable} <=> $tbl->{keys}->{$b}->{is_nullable} )
|| ( scalar(@{$tbl->{keys}->{$a}->{cols}}) <=> scalar(@{$tbl->{keys}->{$b}->{cols}}) )
}
grep {
$tbl->{keys}->{$_}->{type} eq 'BTREE'
}
sort keys %{$tbl->{keys}};
PTDEBUG && _d('Indexes sorted best-first:', join(', ', @indexes));
return @indexes;
}
sub find_best_index {
my ( $self, $tbl, $index ) = @_;
my $best;
if ( $index ) {
($best) = grep { uc $_ eq uc $index } keys %{$tbl->{keys}};
}
if ( !$best ) {
if ( $index ) {
die "Index '$index' does not exist in table";
}
else {
($best) = $self->sort_indexes($tbl);
}
}
PTDEBUG && _d('Best index found is', $best);
return $best;
}
sub find_possible_keys {
my ( $self, $dbh, $database, $table, $quoter, $where ) = @_;
return () unless $where;
my $sql = 'EXPLAIN SELECT * FROM ' . $quoter->quote($database, $table)
. ' WHERE ' . $where;
PTDEBUG && _d($sql);
my $expl = $dbh->selectrow_hashref($sql);
$expl = { map { lc($_) => $expl->{$_} } keys %$expl };
if ( $expl->{possible_keys} ) {
PTDEBUG && _d('possible_keys =', $expl->{possible_keys});
my @candidates = split(',', $expl->{possible_keys});
my %possible = map { $_ => 1 } @candidates;
if ( $expl->{key} ) {
PTDEBUG && _d('MySQL chose', $expl->{key});
unshift @candidates, grep { $possible{$_} } split(',', $expl->{key});
PTDEBUG && _d('Before deduping:', join(', ', @candidates));
my %seen;
@candidates = grep { !$seen{$_}++ } @candidates;
}
PTDEBUG && _d('Final list:', join(', ', @candidates));
return @candidates;
}
else {
PTDEBUG && _d('No keys in possible_keys');
return ();
}
}
sub check_table {
my ( $self, %args ) = @_;
my @required_args = qw(dbh db tbl);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($dbh, $db, $tbl) = @args{@required_args};
my $q = $self->{Quoter};
my $db_tbl = $q->quote($db, $tbl);
PTDEBUG && _d('Checking', $db_tbl);
my $sql = "SHOW TABLES FROM " . $q->quote($db)
. ' LIKE ' . $q->literal_like($tbl);
PTDEBUG && _d($sql);
my $row;
eval {
$row = $dbh->selectrow_arrayref($sql);
};
if ( $EVAL_ERROR ) {
PTDEBUG && _d($EVAL_ERROR);
return 0;
}
if ( !$row->[0] || $row->[0] ne $tbl ) {
PTDEBUG && _d('Table does not exist');
return 0;
}
PTDEBUG && _d('Table exists; no privs to check');
return 1 unless $args{all_privs};
$sql = "SHOW FULL COLUMNS FROM $db_tbl";
PTDEBUG && _d($sql);
eval {
$row = $dbh->selectrow_hashref($sql);
};
if ( $EVAL_ERROR ) {
PTDEBUG && _d($EVAL_ERROR);
return 0;
}
if ( !scalar keys %$row ) {
PTDEBUG && _d('Table has no columns:', Dumper($row));
return 0;
}
my $privs = $row->{privileges} || $row->{Privileges};
$sql = "DELETE FROM $db_tbl LIMIT 0";
PTDEBUG && _d($sql);
eval {
$dbh->do($sql);
};
my $can_delete = $EVAL_ERROR ? 0 : 1;
PTDEBUG && _d('User privs on', $db_tbl, ':', $privs,
($can_delete ? 'delete' : ''));
if ( !($privs =~ m/select/ && $privs =~ m/insert/ && $privs =~ m/update/
&& $can_delete) ) {
PTDEBUG && _d('User does not have all privs');
return 0;
}
PTDEBUG && _d('User has all privs');
return 1;
}
sub get_engine {
my ( $self, $ddl, $opts ) = @_;
my ( $engine ) = $ddl =~ m/\).*?(?:ENGINE|TYPE)=(\w+)/;
PTDEBUG && _d('Storage engine:', $engine);
return $engine || undef;
}
sub get_keys {
my ( $self, $ddl, $opts, $is_nullable ) = @_;
my $engine = $self->get_engine($ddl);
my $keys = {};
my $clustered_key = undef;
KEY:
foreach my $key ( $ddl =~ m/^ ((?:[A-Z]+ )?KEY .*)$/gm ) {
next KEY if $key =~ m/FOREIGN/;
my $key_ddl = $key;
PTDEBUG && _d('Parsed key:', $key_ddl);
if ( $engine !~ m/MEMORY|HEAP/ ) {
$key =~ s/USING HASH/USING BTREE/;
}
my ( $type, $cols ) = $key =~ m/(?:USING (\w+))? \((.+)\)/;
my ( $special ) = $key =~ m/(FULLTEXT|SPATIAL)/;
$type = $type || $special || 'BTREE';
if ( $opts->{mysql_version} && $opts->{mysql_version} lt '004001000'
&& $engine =~ m/HEAP|MEMORY/i )
{
$type = 'HASH'; # MySQL pre-4.1 supports only HASH indexes on HEAP
}
my ($name) = $key =~ m/(PRIMARY|`[^`]*`)/;
my $unique = $key =~ m/PRIMARY|UNIQUE/ ? 1 : 0;
my @cols;
my @col_prefixes;
foreach my $col_def ( $cols =~ m/`[^`]+`(?:\(\d+\))?/g ) {
my ($name, $prefix) = $col_def =~ m/`([^`]+)`(?:\((\d+)\))?/;
push @cols, $name;
push @col_prefixes, $prefix;
}
$name =~ s/`//g;
PTDEBUG && _d( $name, 'key cols:', join(', ', map { "`$_`" } @cols));
$keys->{$name} = {
name => $name,
type => $type,
colnames => $cols,
cols => \@cols,
col_prefixes => \@col_prefixes,
is_unique => $unique,
is_nullable => scalar(grep { $is_nullable->{$_} } @cols),
is_col => { map { $_ => 1 } @cols },
ddl => $key_ddl,
};
if ( $engine =~ m/InnoDB/i && !$clustered_key ) {
my $this_key = $keys->{$name};
if ( $this_key->{name} eq 'PRIMARY' ) {
$clustered_key = 'PRIMARY';
}
elsif ( $this_key->{is_unique} && !$this_key->{is_nullable} ) {
$clustered_key = $this_key->{name};
}
PTDEBUG && $clustered_key && _d('This key is the clustered key');
}
}
return $keys, $clustered_key;
}
sub get_fks {
my ( $self, $ddl, $opts ) = @_;
my $q = $self->{Quoter};
my $fks = {};
foreach my $fk (
$ddl =~ m/CONSTRAINT .* FOREIGN KEY .* REFERENCES [^\)]*\)/mg )
{
my ( $name ) = $fk =~ m/CONSTRAINT `(.*?)`/;
my ( $cols ) = $fk =~ m/FOREIGN KEY \(([^\)]+)\)/;
my ( $parent, $parent_cols ) = $fk =~ m/REFERENCES (\S+) \(([^\)]+)\)/;
my ($db, $tbl) = $q->split_unquote($parent, $opts->{database});
my %parent_tbl = (tbl => $tbl);
$parent_tbl{db} = $db if $db;
if ( $parent !~ m/\./ && $opts->{database} ) {
$parent = $q->quote($opts->{database}) . ".$parent";
}
$fks->{$name} = {
name => $name,
colnames => $cols,
cols => [ map { s/[ `]+//g; $_; } split(',', $cols) ],
parent_tbl => \%parent_tbl,
parent_tblname => $parent,
parent_cols => [ map { s/[ `]+//g; $_; } split(',', $parent_cols) ],
parent_colnames=> $parent_cols,
ddl => $fk,
};
}
return $fks;
}
sub remove_auto_increment {
my ( $self, $ddl ) = @_;
$ddl =~ s/(^\).*?) AUTO_INCREMENT=\d+\b/$1/m;
return $ddl;
}
sub get_table_status {
my ( $self, $dbh, $db, $like ) = @_;
my $q = $self->{Quoter};
my $sql = "SHOW TABLE STATUS FROM " . $q->quote($db);
my @params;
if ( $like ) {
$sql .= ' LIKE ?';
push @params, $like;
}
PTDEBUG && _d($sql, @params);
my $sth = $dbh->prepare($sql);
eval { $sth->execute(@params); };
if ($EVAL_ERROR) {
PTDEBUG && _d($EVAL_ERROR);
return;
}
my @tables = @{$sth->fetchall_arrayref({})};
@tables = map {
my %tbl; # Make a copy with lowercased keys
@tbl{ map { lc $_ } keys %$_ } = values %$_;
$tbl{engine} ||= $tbl{type} || $tbl{comment};
delete $tbl{type};
\%tbl;
} @tables;
return @tables;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End TableParser package
# ###########################################################################
# ###########################################################################
# Progress package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Progress.pm
# t/lib/Progress.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Progress;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
foreach my $arg (qw(jobsize)) {
die "I need a $arg argument" unless defined $args{$arg};
}
if ( (!$args{report} || !$args{interval}) ) {
if ( $args{spec} && @{$args{spec}} == 2 ) {
@args{qw(report interval)} = @{$args{spec}};
}
else {
die "I need either report and interval arguments, or a spec";
}
}
my $name = $args{name} || "Progress";
$args{start} ||= time();
my $self;
$self = {
last_reported => $args{start},
fraction => 0, # How complete the job is
callback => sub {
my ($fraction, $elapsed, $remaining, $eta) = @_;
printf STDERR "$name: %3d%% %s remain\n",
$fraction * 100,
Transformers::secs_to_time($remaining),
Transformers::ts($eta);
},
%args,
};
return bless $self, $class;
}
sub validate_spec {
shift @_ if $_[0] eq 'Progress'; # Permit calling as Progress-> or Progress::
my ( $spec ) = @_;
if ( @$spec != 2 ) {
die "spec array requires a two-part argument\n";
}
if ( $spec->[0] !~ m/^(?:percentage|time|iterations)$/ ) {
die "spec array's first element must be one of "
. "percentage,time,iterations\n";
}
if ( $spec->[1] !~ m/^\d+$/ ) {
die "spec array's second element must be an integer\n";
}
}
sub set_callback {
my ( $self, $callback ) = @_;
$self->{callback} = $callback;
}
sub start {
my ( $self, $start ) = @_;
$self->{start} = $self->{last_reported} = $start || time();
$self->{first_report} = 0;
}
sub update {
my ( $self, $callback, %args ) = @_;
my $jobsize = $self->{jobsize};
my $now ||= $args{now} || time;
$self->{iterations}++; # How many updates have happened;
if ( !$self->{first_report} && $args{first_report} ) {
$args{first_report}->();
$self->{first_report} = 1;
}
if ( $self->{report} eq 'time'
&& $self->{interval} > $now - $self->{last_reported}
) {
return;
}
elsif ( $self->{report} eq 'iterations'
&& ($self->{iterations} - 1) % $self->{interval} > 0
) {
return;
}
$self->{last_reported} = $now;
my $completed = $callback->();
$self->{updates}++; # How many times we have run the update callback
return if $completed > $jobsize;
my $fraction = $completed > 0 ? $completed / $jobsize : 0;
if ( $self->{report} eq 'percentage'
&& $self->fraction_modulo($self->{fraction})
>= $self->fraction_modulo($fraction)
) {
$self->{fraction} = $fraction;
return;
}
$self->{fraction} = $fraction;
my $elapsed = $now - $self->{start};
my $remaining = 0;
my $eta = $now;
if ( $completed > 0 && $completed <= $jobsize && $elapsed > 0 ) {
my $rate = $completed / $elapsed;
if ( $rate > 0 ) {
$remaining = ($jobsize - $completed) / $rate;
$eta = $now + int($remaining);
}
}
$self->{callback}->($fraction, $elapsed, $remaining, $eta, $completed);
}
sub fraction_modulo {
my ( $self, $num ) = @_;
$num *= 100; # Convert from fraction to percentage
return sprintf('%d',
sprintf('%d', $num / $self->{interval}) * $self->{interval});
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End Progress package
# ###########################################################################
# ###########################################################################
# Retry package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Retry.pm
# t/lib/Retry.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Retry;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
my $self = {
%args,
};
return bless $self, $class;
}
sub retry {
my ( $self, %args ) = @_;
my @required_args = qw(try fail final_fail);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
};
my ($try, $fail, $final_fail) = @args{@required_args};
my $wait = $args{wait} || sub { sleep 1; };
my $tries = $args{tries} || 3;
my $last_error;
my $tryno = 0;
TRY:
while ( ++$tryno <= $tries ) {
PTDEBUG && _d("Try", $tryno, "of", $tries);
my $result;
eval {
$result = $try->(tryno=>$tryno);
};
if ( $EVAL_ERROR ) {
PTDEBUG && _d("Try code failed:", $EVAL_ERROR);
$last_error = $EVAL_ERROR;
if ( $tryno < $tries ) { # more retries
my $retry = $fail->(tryno=>$tryno, error=>$last_error);
last TRY unless $retry;
PTDEBUG && _d("Calling wait code");
$wait->(tryno=>$tryno);
}
}
else {
PTDEBUG && _d("Try code succeeded");
return $result;
}
}
PTDEBUG && _d('Try code did not succeed');
return $final_fail->(error=>$last_error);
}
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;
}
# ###########################################################################
# End Retry package
# ###########################################################################
# ###########################################################################
# Cxn package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Cxn.pm
# t/lib/Cxn.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Cxn;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use constant PERCONA_TOOLKIT_TEST_USE_DSN_NAMES => $ENV{PERCONA_TOOLKIT_TEST_USE_DSN_NAMES} || 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(DSNParser OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
};
my ($dp, $o) = @args{@required_args};
my $dsn_defaults = $dp->parse_options($o);
my $prev_dsn = $args{prev_dsn};
my $dsn = $args{dsn};
if ( !$dsn ) {
$args{dsn_string} ||= 'h=' . ($dsn_defaults->{h} || 'localhost');
$dsn = $dp->parse(
$args{dsn_string}, $prev_dsn, $dsn_defaults);
}
elsif ( $prev_dsn ) {
$dsn = $dp->copy($prev_dsn, $dsn);
}
my $self = {
dsn => $dsn,
dbh => $args{dbh},
dsn_name => $dp->as_string($dsn, [qw(h P S)]),
hostname => '',
set => $args{set},
dbh_set => 0,
OptionParser => $o,
DSNParser => $dp,
};
return bless $self, $class;
}
sub connect {
my ( $self ) = @_;
my $dsn = $self->{dsn};
my $dp = $self->{DSNParser};
my $o = $self->{OptionParser};
my $dbh = $self->{dbh};
if ( !$dbh || !$dbh->ping() ) {
if ( $o->get('ask-pass') && !$self->{asked_for_pass} ) {
$dsn->{p} = OptionParser::prompt_noecho("Enter MySQL password: ");
$self->{asked_for_pass} = 1;
}
$dbh = $dp->get_dbh($dp->get_cxn_params($dsn), { AutoCommit => 1 });
}
PTDEBUG && _d($dbh, 'Connected dbh to', $self->{name});
return $self->set_dbh($dbh);
}
sub set_dbh {
my ($self, $dbh) = @_;
if ( $self->{dbh} && $self->{dbh} == $dbh && $self->{dbh_set} ) {
PTDEBUG && _d($dbh, 'Already set dbh');
return $dbh;
}
PTDEBUG && _d($dbh, 'Setting dbh');
$dbh->{FetchHashKeyName} = 'NAME_lc';
my $sql = 'SELECT @@hostname, @@server_id';
PTDEBUG && _d($dbh, $sql);
my ($hostname, $server_id) = $dbh->selectrow_array($sql);
PTDEBUG && _d($dbh, 'hostname:', $hostname, $server_id);
if ( $hostname ) {
$self->{hostname} = $hostname;
}
if ( my $set = $self->{set}) {
$set->($dbh);
}
$self->{dbh} = $dbh;
$self->{dbh_set} = 1;
return $dbh;
}
sub dbh {
my ($self) = @_;
return $self->{dbh};
}
sub dsn {
my ($self) = @_;
return $self->{dsn};
}
sub name {
my ($self) = @_;
return $self->{dsn_name} if PERCONA_TOOLKIT_TEST_USE_DSN_NAMES;
return $self->{hostname} || $self->{dsn_name} || 'unknown host';
}
sub DESTROY {
my ($self) = @_;
if ( $self->{dbh} && ref($self->{dbh}) ) {
PTDEBUG && _d('Disconnecting dbh', $self->{dbh}, $self->{name});
$self->{dbh}->disconnect();
}
return;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End Cxn package
# ###########################################################################
# ###########################################################################
# MasterSlave package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/MasterSlave.pm
# t/lib/MasterSlave.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package MasterSlave;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
my $self = {
%args,
replication_thread => {},
};
return bless $self, $class;
}
sub get_slaves {
my ($self, %args) = @_;
my @required_args = qw(make_cxn OptionParser DSNParser Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($make_cxn, $o, $dp) = @args{@required_args};
my $slaves = [];
my $method = $o->get('recursion-method');
PTDEBUG && _d('Slave recursion method:', $method);
if ( !$method || $method =~ m/processlist|hosts/i ) {
my @required_args = qw(dbh dsn);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($dbh, $dsn) = @args{@required_args};
$self->recurse_to_slaves(
{ dbh => $dbh,
dsn => $dsn,
dsn_parser => $dp,
recurse => $o->get('recurse'),
method => $o->get('recursion-method'),
callback => sub {
my ( $dsn, $dbh, $level, $parent ) = @_;
return unless $level;
PTDEBUG && _d('Found slave:', $dp->as_string($dsn));
push @$slaves, $make_cxn->(dsn => $dsn, dbh => $dbh);
return;
},
}
);
}
elsif ( $method =~ m/^dsn=/i ) {
my ($dsn_table_dsn) = $method =~ m/^dsn=(.+)/i;
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
);
}
elsif ( $method =~ m/none/i ) {
PTDEBUG && _d('Not getting to slaves');
}
else {
die "Invalid --recursion-method: $method. Valid values are: "
. "dsn=DSN, hosts, or processlist.\n";
}
return $slaves;
}
sub recurse_to_slaves {
my ( $self, $args, $level ) = @_;
$level ||= 0;
my $dp = $args->{dsn_parser};
my $dsn = $args->{dsn};
if ( lc($args->{method} || '') eq 'none' ) {
PTDEBUG && _d('Not recursing to slaves');
return;
}
my $dbh;
eval {
$dbh = $args->{dbh} || $dp->get_dbh(
$dp->get_cxn_params($dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($dsn), "\n"
or die "Cannot print: $OS_ERROR";
return;
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
|| ( defined $master_thinks_i_am && $master_thinks_i_am != $id )
|| $args->{server_ids_seen}->{$id}++
) {
PTDEBUG && _d('Server ID seen, or not what master said');
if ( $args->{skip_callback} ) {
$args->{skip_callback}->($dsn, $dbh, $level, $args->{parent});
}
return;
}
$args->{callback}->($dsn, $dbh, $level, $args->{parent});
if ( !defined $args->{recurse} || $level < $args->{recurse} ) {
my @slaves =
grep { !$_->{master_id} || $_->{master_id} == $id } # Only my slaves.
$self->find_slave_hosts($dp, $dbh, $dsn, $args->{method});
foreach my $slave ( @slaves ) {
PTDEBUG && _d('Recursing from',
$dp->as_string($dsn), 'to', $dp->as_string($slave));
$self->recurse_to_slaves(
{ %$args, dsn => $slave, dbh => undef, parent => $dsn }, $level + 1 );
}
}
}
sub find_slave_hosts {
my ( $self, $dsn_parser, $dbh, $dsn, $method ) = @_;
my @methods = qw(processlist hosts);
if ( $method ) {
@methods = grep { $_ ne $method } @methods;
unshift @methods, $method;
}
else {
if ( ($dsn->{P} || 3306) != 3306 ) {
PTDEBUG && _d('Port number is non-standard; using only hosts method');
@methods = qw(hosts);
}
}
PTDEBUG && _d('Looking for slaves on', $dsn_parser->as_string($dsn),
'using methods', @methods);
my @slaves;
METHOD:
foreach my $method ( @methods ) {
my $find_slaves = "_find_slaves_by_$method";
PTDEBUG && _d('Finding slaves with', $find_slaves);
@slaves = $self->$find_slaves($dsn_parser, $dbh, $dsn);
last METHOD if @slaves;
}
PTDEBUG && _d('Found', scalar(@slaves), 'slaves');
return @slaves;
}
sub _find_slaves_by_processlist {
my ( $self, $dsn_parser, $dbh, $dsn ) = @_;
my @slaves = map {
my $slave = $dsn_parser->parse("h=$_", $dsn);
$slave->{source} = 'processlist';
$slave;
}
grep { $_ }
map {
my ( $host ) = $_->{host} =~ m/^([^:]+):/;
if ( $host eq 'localhost' ) {
$host = '127.0.0.1'; # Replication never uses sockets.
}
$host;
} $self->get_connected_slaves($dbh);
return @slaves;
}
sub _find_slaves_by_hosts {
my ( $self, $dsn_parser, $dbh, $dsn ) = @_;
my @slaves;
my $sql = 'SHOW SLAVE HOSTS';
PTDEBUG && _d($dbh, $sql);
@slaves = @{$dbh->selectall_arrayref($sql, { Slice => {} })};
if ( @slaves ) {
PTDEBUG && _d('Found some SHOW SLAVE HOSTS info');
@slaves = map {
my %hash;
@hash{ map { lc $_ } keys %$_ } = values %$_;
my $spec = "h=$hash{host},P=$hash{port}"
. ( $hash{user} ? ",u=$hash{user}" : '')
. ( $hash{password} ? ",p=$hash{password}" : '');
my $dsn = $dsn_parser->parse($spec, $dsn);
$dsn->{server_id} = $hash{server_id};
$dsn->{master_id} = $hash{master_id};
$dsn->{source} = 'hosts';
$dsn;
} @slaves;
}
return @slaves;
}
sub get_connected_slaves {
my ( $self, $dbh ) = @_;
my $show = "SHOW GRANTS FOR ";
my $user = 'CURRENT_USER()';
my $vp = $self->{VersionParser};
if ( $vp && !$vp->version_ge($dbh, '4.1.2') ) {
$user = $dbh->selectrow_arrayref('SELECT USER()')->[0];
$user =~ s/([^@]+)@(.+)/'$1'\@'$2'/;
}
my $sql = $show . $user;
PTDEBUG && _d($dbh, $sql);
my $proc;
eval {
$proc = grep {
m/ALL PRIVILEGES.*?\*\.\*|PROCESS/
} @{$dbh->selectcol_arrayref($sql)};
};
if ( $EVAL_ERROR ) {
if ( $EVAL_ERROR =~ m/no such grant defined for user/ ) {
PTDEBUG && _d('Retrying SHOW GRANTS without host; error:',
$EVAL_ERROR);
($user) = split('@', $user);
$sql = $show . $user;
PTDEBUG && _d($sql);
eval {
$proc = grep {
m/ALL PRIVILEGES.*?\*\.\*|PROCESS/
} @{$dbh->selectcol_arrayref($sql)};
};
}
die "Failed to $sql: $EVAL_ERROR" if $EVAL_ERROR;
}
if ( !$proc ) {
die "You do not have the PROCESS privilege";
}
$sql = 'SHOW PROCESSLIST';
PTDEBUG && _d($dbh, $sql);
grep { $_->{command} =~ m/Binlog Dump/i }
map { # Lowercase the column names
my %hash;
@hash{ map { lc $_ } keys %$_ } = values %$_;
\%hash;
}
@{$dbh->selectall_arrayref($sql, { Slice => {} })};
}
sub is_master_of {
my ( $self, $master, $slave ) = @_;
my $master_status = $self->get_master_status($master)
or die "The server specified as a master is not a master";
my $slave_status = $self->get_slave_status($slave)
or die "The server specified as a slave is not a slave";
my @connected = $self->get_connected_slaves($master)
or die "The server specified as a master has no connected slaves";
my (undef, $port) = $master->selectrow_array('SHOW VARIABLES LIKE "port"');
if ( $port != $slave_status->{master_port} ) {
die "The slave is connected to $slave_status->{master_port} "
. "but the master's port is $port";
}
if ( !grep { $slave_status->{master_user} eq $_->{user} } @connected ) {
die "I don't see any slave I/O thread connected with user "
. $slave_status->{master_user};
}
if ( ($slave_status->{slave_io_state} || '')
eq 'Waiting for master to send event' )
{
my ( $master_log_name, $master_log_num )
= $master_status->{file} =~ m/^(.*?)\.0*([1-9][0-9]*)$/;
my ( $slave_log_name, $slave_log_num )
= $slave_status->{master_log_file} =~ m/^(.*?)\.0*([1-9][0-9]*)$/;
if ( $master_log_name ne $slave_log_name
|| abs($master_log_num - $slave_log_num) > 1 )
{
die "The slave thinks it is reading from "
. "$slave_status->{master_log_file}, but the "
. "master is writing to $master_status->{file}";
}
}
return 1;
}
sub get_master_dsn {
my ( $self, $dbh, $dsn, $dsn_parser ) = @_;
my $master = $self->get_slave_status($dbh) or return undef;
my $spec = "h=$master->{master_host},P=$master->{master_port}";
return $dsn_parser->parse($spec, $dsn);
}
sub get_slave_status {
my ( $self, $dbh ) = @_;
if ( !$self->{not_a_slave}->{$dbh} ) {
my $sth = $self->{sths}->{$dbh}->{SLAVE_STATUS}
||= $dbh->prepare('SHOW SLAVE STATUS');
PTDEBUG && _d($dbh, 'SHOW SLAVE STATUS');
$sth->execute();
my ($ss) = @{$sth->fetchall_arrayref({})};
if ( $ss && %$ss ) {
$ss = { map { lc($_) => $ss->{$_} } keys %$ss }; # lowercase the keys
return $ss;
}
PTDEBUG && _d('This server returns nothing for SHOW SLAVE STATUS');
$self->{not_a_slave}->{$dbh}++;
}
}
sub get_master_status {
my ( $self, $dbh ) = @_;
if ( $self->{not_a_master}->{$dbh} ) {
PTDEBUG && _d('Server on dbh', $dbh, 'is not a master');
return;
}
my $sth = $self->{sths}->{$dbh}->{MASTER_STATUS}
||= $dbh->prepare('SHOW MASTER STATUS');
PTDEBUG && _d($dbh, 'SHOW MASTER STATUS');
$sth->execute();
my ($ms) = @{$sth->fetchall_arrayref({})};
PTDEBUG && _d(
$ms ? map { "$_=" . (defined $ms->{$_} ? $ms->{$_} : '') } keys %$ms
: '');
if ( !$ms || scalar keys %$ms < 2 ) {
PTDEBUG && _d('Server on dbh', $dbh, 'does not seem to be a master');
$self->{not_a_master}->{$dbh}++;
}
return { map { lc($_) => $ms->{$_} } keys %$ms }; # lowercase the keys
}
sub wait_for_master {
my ( $self, %args ) = @_;
my @required_args = qw(master_status slave_dbh);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($master_status, $slave_dbh) = @args{@required_args};
my $timeout = $args{timeout} || 60;
my $result;
my $waited;
if ( $master_status ) {
my $sql = "SELECT MASTER_POS_WAIT('$master_status->{file}', "
. "$master_status->{position}, $timeout)";
PTDEBUG && _d($slave_dbh, $sql);
my $start = time;
($result) = $slave_dbh->selectrow_array($sql);
$waited = time - $start;
PTDEBUG && _d('Result of waiting:', $result);
PTDEBUG && _d("Waited", $waited, "seconds");
}
else {
PTDEBUG && _d('Not waiting: this server is not a master');
}
return {
result => $result,
waited => $waited,
};
}
sub stop_slave {
my ( $self, $dbh ) = @_;
my $sth = $self->{sths}->{$dbh}->{STOP_SLAVE}
||= $dbh->prepare('STOP SLAVE');
PTDEBUG && _d($dbh, $sth->{Statement});
$sth->execute();
}
sub start_slave {
my ( $self, $dbh, $pos ) = @_;
if ( $pos ) {
my $sql = "START SLAVE UNTIL MASTER_LOG_FILE='$pos->{file}', "
. "MASTER_LOG_POS=$pos->{position}";
PTDEBUG && _d($dbh, $sql);
$dbh->do($sql);
}
else {
my $sth = $self->{sths}->{$dbh}->{START_SLAVE}
||= $dbh->prepare('START SLAVE');
PTDEBUG && _d($dbh, $sth->{Statement});
$sth->execute();
}
}
sub catchup_to_master {
my ( $self, $slave, $master, $timeout ) = @_;
$self->stop_slave($master);
$self->stop_slave($slave);
my $slave_status = $self->get_slave_status($slave);
my $slave_pos = $self->repl_posn($slave_status);
my $master_status = $self->get_master_status($master);
my $master_pos = $self->repl_posn($master_status);
PTDEBUG && _d('Master position:', $self->pos_to_string($master_pos),
'Slave position:', $self->pos_to_string($slave_pos));
my $result;
if ( $self->pos_cmp($slave_pos, $master_pos) < 0 ) {
PTDEBUG && _d('Waiting for slave to catch up to master');
$self->start_slave($slave, $master_pos);
$result = $self->wait_for_master(
master_status => $master_status,
slave_dbh => $slave,
timeout => $timeout,
master_status => $master_status
);
if ( !defined $result->{result} ) {
$slave_status = $self->get_slave_status($slave);
if ( !$self->slave_is_running($slave_status) ) {
PTDEBUG && _d('Master position:',
$self->pos_to_string($master_pos),
'Slave position:', $self->pos_to_string($slave_pos));
$slave_pos = $self->repl_posn($slave_status);
if ( $self->pos_cmp($slave_pos, $master_pos) != 0 ) {
die "MASTER_POS_WAIT() returned NULL but slave has not "
. "caught up to master";
}
PTDEBUG && _d('Slave is caught up to master and stopped');
}
else {
die "Slave has not caught up to master and it is still running";
}
}
}
else {
PTDEBUG && _d("Slave is already caught up to master");
}
return $result;
}
sub catchup_to_same_pos {
my ( $self, $s1_dbh, $s2_dbh ) = @_;
$self->stop_slave($s1_dbh);
$self->stop_slave($s2_dbh);
my $s1_status = $self->get_slave_status($s1_dbh);
my $s2_status = $self->get_slave_status($s2_dbh);
my $s1_pos = $self->repl_posn($s1_status);
my $s2_pos = $self->repl_posn($s2_status);
if ( $self->pos_cmp($s1_pos, $s2_pos) < 0 ) {
$self->start_slave($s1_dbh, $s2_pos);
}
elsif ( $self->pos_cmp($s2_pos, $s1_pos) < 0 ) {
$self->start_slave($s2_dbh, $s1_pos);
}
$s1_status = $self->get_slave_status($s1_dbh);
$s2_status = $self->get_slave_status($s2_dbh);
$s1_pos = $self->repl_posn($s1_status);
$s2_pos = $self->repl_posn($s2_status);
if ( $self->slave_is_running($s1_status)
|| $self->slave_is_running($s2_status)
|| $self->pos_cmp($s1_pos, $s2_pos) != 0)
{
die "The servers aren't both stopped at the same position";
}
}
sub slave_is_running {
my ( $self, $slave_status ) = @_;
return ($slave_status->{slave_sql_running} || 'No') eq 'Yes';
}
sub has_slave_updates {
my ( $self, $dbh ) = @_;
my $sql = q{SHOW VARIABLES LIKE 'log_slave_updates'};
PTDEBUG && _d($dbh, $sql);
my ($name, $value) = $dbh->selectrow_array($sql);
return $value && $value =~ m/^(1|ON)$/;
}
sub repl_posn {
my ( $self, $status ) = @_;
if ( exists $status->{file} && exists $status->{position} ) {
return {
file => $status->{file},
position => $status->{position},
};
}
else {
return {
file => $status->{relay_master_log_file},
position => $status->{exec_master_log_pos},
};
}
}
sub get_slave_lag {
my ( $self, $dbh ) = @_;
my $stat = $self->get_slave_status($dbh);
return unless $stat; # server is not a slave
return $stat->{seconds_behind_master};
}
sub pos_cmp {
my ( $self, $a, $b ) = @_;
return $self->pos_to_string($a) cmp $self->pos_to_string($b);
}
sub short_host {
my ( $self, $dsn ) = @_;
my ($host, $port);
if ( $dsn->{master_host} ) {
$host = $dsn->{master_host};
$port = $dsn->{master_port};
}
else {
$host = $dsn->{h};
$port = $dsn->{P};
}
return ($host || '[default]') . ( ($port || 3306) == 3306 ? '' : ":$port" );
}
sub is_replication_thread {
my ( $self, $query, %args ) = @_;
return unless $query;
my $type = lc($args{type} || 'all');
die "Invalid type: $type"
unless $type =~ m/^binlog_dump|slave_io|slave_sql|all$/i;
my $match = 0;
if ( $type =~ m/binlog_dump|all/i ) {
$match = 1
if ($query->{Command} || $query->{command} || '') eq "Binlog Dump";
}
if ( !$match ) {
if ( ($query->{User} || $query->{user} || '') eq "system user" ) {
PTDEBUG && _d("Slave replication thread");
if ( $type ne 'all' ) {
my $state = $query->{State} || $query->{state} || '';
if ( $state =~ m/^init|end$/ ) {
PTDEBUG && _d("Special state:", $state);
$match = 1;
}
else {
my ($slave_sql) = $state =~ m/
^(Waiting\sfor\sthe\snext\sevent
|Reading\sevent\sfrom\sthe\srelay\slog
|Has\sread\sall\srelay\slog;\swaiting
|Making\stemp\sfile
|Waiting\sfor\sslave\smutex\son\sexit)/xi;
$match = $type eq 'slave_sql' && $slave_sql ? 1
: $type eq 'slave_io' && !$slave_sql ? 1
: 0;
}
}
else {
$match = 1;
}
}
else {
PTDEBUG && _d('Not system user');
}
if ( !defined $args{check_known_ids} || $args{check_known_ids} ) {
my $id = $query->{Id} || $query->{id};
if ( $match ) {
$self->{replication_thread}->{$id} = 1;
}
else {
if ( $self->{replication_thread}->{$id} ) {
PTDEBUG && _d("Thread ID is a known replication thread ID");
$match = 1;
}
}
}
}
PTDEBUG && _d('Matches', $type, 'replication thread:',
($match ? 'yes' : 'no'), '; match:', $match);
return $match;
}
sub get_replication_filters {
my ( $self, %args ) = @_;
my @required_args = qw(dbh);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($dbh) = @args{@required_args};
my %filters = ();
my $status = $self->get_master_status($dbh);
if ( $status ) {
map { $filters{$_} = $status->{$_} }
grep { defined $status->{$_} && $status->{$_} ne '' }
qw(
binlog_do_db
binlog_ignore_db
);
}
$status = $self->get_slave_status($dbh);
if ( $status ) {
map { $filters{$_} = $status->{$_} }
grep { defined $status->{$_} && $status->{$_} ne '' }
qw(
replicate_do_db
replicate_ignore_db
replicate_do_table
replicate_ignore_table
replicate_wild_do_table
replicate_wild_ignore_table
);
my $sql = "SHOW VARIABLES LIKE 'slave_skip_errors'";
PTDEBUG && _d($dbh, $sql);
my $row = $dbh->selectrow_arrayref($sql);
$filters{slave_skip_errors} = $row->[1] if $row->[1] && $row->[1] ne 'OFF';
}
return \%filters;
}
sub pos_to_string {
my ( $self, $pos ) = @_;
my $fmt = '%s/%020d';
return sprintf($fmt, @{$pos}{qw(file position)});
}
sub reset_known_replication_threads {
my ( $self ) = @_;
$self->{replication_thread} = {};
return;
}
sub get_cxn_from_dsn_table {
my ($self, %args) = @_;
my @required_args = qw(dsn_table_dsn make_cxn DSNParser Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($dsn_table_dsn, $make_cxn, $dp, $q) = @args{@required_args};
PTDEBUG && _d('DSN table DSN:', $dsn_table_dsn);
my $dsn = $dp->parse($dsn_table_dsn);
my $dsn_table;
if ( $dsn->{D} && $dsn->{t} ) {
$dsn_table = $q->quote($dsn->{D}, $dsn->{t});
}
elsif ( $dsn->{t} && $dsn->{t} =~ m/\./ ) {
$dsn_table = $q->quote($q->split_unquote($dsn->{t}));
}
else {
die "DSN table DSN does not specify a database (D) "
. "or a database-qualified table (t)";
}
my $dsn_tbl_cxn = $make_cxn->(dsn => $dsn);
my $dbh = $dsn_tbl_cxn->connect();
my $sql = "SELECT dsn FROM $dsn_table ORDER BY id";
PTDEBUG && _d($sql);
my $dsn_strings = $dbh->selectcol_arrayref($sql);
my @cxn;
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
return \@cxn;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End MasterSlave package
# ###########################################################################
# ###########################################################################
# ReplicaLagWaiter package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/ReplicaLagWaiter.pm
# t/lib/ReplicaLagWaiter.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package ReplicaLagWaiter;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Time::HiRes qw(sleep time);
use Data::Dumper;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(oktorun get_lag sleep max_lag slaves);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my $self = {
%args,
};
return bless $self, $class;
}
sub wait {
my ( $self, %args ) = @_;
my @required_args = qw();
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my $pr = $args{Progress};
my $oktorun = $self->{oktorun};
my $get_lag = $self->{get_lag};
my $sleep = $self->{sleep};
my $slaves = $self->{slaves};
my $max_lag = $self->{max_lag};
my $worst; # most lagging slave
my $pr_callback;
my $pr_first_report;
if ( $pr ) {
$pr_callback = sub {
my ($fraction, $elapsed, $remaining, $eta, $completed) = @_;
my $dsn_name = $worst->{cxn}->name();
if ( defined $worst->{lag} ) {
print STDERR "Replica lag is " . ($worst->{lag} || '?')
. " seconds on $dsn_name. Waiting.\n";
}
else {
print STDERR "Replica $dsn_name is stopped. Waiting.\n";
}
return;
};
$pr->set_callback($pr_callback);
$pr_first_report = sub {
my $dsn_name = $worst->{cxn}->name();
if ( !defined $worst->{lag} ) {
print STDERR "Replica $dsn_name is stopped. Waiting.\n";
}
return;
};
}
my @lagged_slaves = map { {cxn=>$_, lag=>undef} } @$slaves;
while ( $oktorun->() && @lagged_slaves ) {
PTDEBUG && _d('Checking slave lag');
for my $i ( 0..$#lagged_slaves ) {
my $lag = $get_lag->($lagged_slaves[$i]->{cxn});
PTDEBUG && _d($lagged_slaves[$i]->{cxn}->name(),
'slave lag:', $lag);
if ( !defined $lag || $lag > $max_lag ) {
$lagged_slaves[$i]->{lag} = $lag;
}
else {
delete $lagged_slaves[$i];
}
}
@lagged_slaves = grep { defined $_ } @lagged_slaves;
if ( @lagged_slaves ) {
@lagged_slaves = reverse sort {
defined $a->{lag} && defined $b->{lag} ? $a->{lag} <=> $b->{lag}
: defined $a->{lag} ? -1
: 1;
} @lagged_slaves;
$worst = $lagged_slaves[0];
PTDEBUG && _d(scalar @lagged_slaves, 'slaves are lagging, worst:',
$worst->{lag}, 'on', Dumper($worst->{cxn}->dsn()));
if ( $pr ) {
$pr->update(
sub { return 0; },
first_report => $pr_first_report,
);
}
PTDEBUG && _d('Calling sleep callback');
$sleep->($worst->{cxn}, $worst->{lag});
}
}
PTDEBUG && _d('All slaves caught up');
return;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End ReplicaLagWaiter package
# ###########################################################################
# ###########################################################################
# MySQLStatusWaiter package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/MySQLStatusWaiter.pm
# t/lib/MySQLStatusWaiter.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package MySQLStatusWaiter;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(max_spec get_status sleep oktorun);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
PTDEBUG && _d('Parsing spec for max thresholds');
my $max_val_for = _parse_spec($args{max_spec});
if ( $max_val_for ) {
_check_and_set_vals(
vars => $max_val_for,
get_status => $args{get_status},
threshold_factor => 0.2, # +20%
);
}
PTDEBUG && _d('Parsing spec for critical thresholds');
my $critical_val_for = _parse_spec($args{critical_spec} || []);
if ( $critical_val_for ) {
_check_and_set_vals(
vars => $critical_val_for,
get_status => $args{get_status},
threshold_factor => 1.0, # double (x2; +100%)
);
}
my $self = {
get_status => $args{get_status},
sleep => $args{sleep},
oktorun => $args{oktorun},
max_val_for => $max_val_for,
critical_val_for => $critical_val_for,
};
return bless $self, $class;
}
sub _parse_spec {
my ($spec) = @_;
return unless $spec && scalar @$spec;
my %max_val_for;
foreach my $var_val ( @$spec ) {
die "Empty or undefined spec\n" unless $var_val;
$var_val =~ s/^\s+//;
$var_val =~ s/\s+$//g;
my ($var, $val) = split /[:=]/, $var_val;
die "$var_val does not contain a variable\n" unless $var;
die "$var is not a variable name\n" unless $var =~ m/^[a-zA-Z_]+$/;
if ( !$val ) {
PTDEBUG && _d('Will get intial value for', $var, 'later');
$max_val_for{$var} = undef;
}
else {
die "The value for $var must be a number\n"
unless $val =~ m/^[\d\.]+$/;
$max_val_for{$var} = $val;
}
}
return \%max_val_for;
}
sub max_values {
my ($self) = @_;
return $self->{max_val_for};
}
sub critical_values {
my ($self) = @_;
return $self->{critical_val_for};
}
sub wait {
my ( $self, %args ) = @_;
return unless $self->{max_val_for};
my $pr = $args{Progress}; # optional
my $oktorun = $self->{oktorun};
my $get_status = $self->{get_status};
my $sleep = $self->{sleep};
my %vals_too_high = %{$self->{max_val_for}};
my $pr_callback;
if ( $pr ) {
$pr_callback = sub {
print STDERR "Pausing because "
. join(', ',
map {
"$_="
. (defined $vals_too_high{$_} ? $vals_too_high{$_}
: 'unknown')
} sort keys %vals_too_high
)
. ".\n";
return;
};
$pr->set_callback($pr_callback);
}
while ( $oktorun->() ) {
PTDEBUG && _d('Checking status variables');
foreach my $var ( sort keys %vals_too_high ) {
my $val = $get_status->($var);
PTDEBUG && _d($var, '=', $val);
if ( $val
&& exists $self->{critical_val_for}->{$var}
&& $val >= $self->{critical_val_for}->{$var} ) {
die "$var=$val exceeds its critical threshold "
. "$self->{critical_val_for}->{$var}\n";
}
if ( !$val || $val >= $self->{max_val_for}->{$var} ) {
$vals_too_high{$var} = $val;
}
else {
delete $vals_too_high{$var};
}
}
last unless scalar keys %vals_too_high;
PTDEBUG && _d(scalar keys %vals_too_high, 'values are too high:',
%vals_too_high);
if ( $pr ) {
$pr->update(sub { return 0; });
}
PTDEBUG && _d('Calling sleep callback');
$sleep->();
%vals_too_high = %{$self->{max_val_for}}; # recheck all vars
}
PTDEBUG && _d('All var vals are low enough');
return;
}
sub _check_and_set_vals {
my (%args) = @_;
my @required_args = qw(vars get_status threshold_factor);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my ($vars, $get_status, $threshold_factor) = @args{@required_args};
PTDEBUG && _d('Checking and setting values');
return unless $vars && scalar %$vars;
foreach my $var ( keys %$vars ) {
my $init_val = $get_status->($var);
die "Variable $var does not exist or its value is undefined\n"
unless defined $init_val;
my $val;
if ( defined $vars->{$var} ) {
$val = $vars->{$var};
}
else {
PTDEBUG && _d('Initial', $var, 'value:', $init_val);
$val = int(($init_val * $threshold_factor) + $init_val);
$vars->{$var} = $val;
}
PTDEBUG && _d('Wait if', $var, '>=', $val);
}
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End MySQLStatusWaiter package
# ###########################################################################
# ###########################################################################
# WeightedAvgRate package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/WeightedAvgRate.pm
# t/lib/WeightedAvgRate.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package WeightedAvgRate;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(target_t);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my $self = {
%args,
avg_n => 0,
avg_t => 0,
weight => $args{weight} || 0.75,
};
return bless $self, $class;
}
sub update {
my ($self, $n, $t) = @_;
PTDEBUG && _d('Master op time:', $n, 'n /', $t, 's');
if ( $self->{avg_n} && $self->{avg_t} ) {
$self->{avg_n} = ($self->{avg_n} * $self->{weight}) + $n;
$self->{avg_t} = ($self->{avg_t} * $self->{weight}) + $t;
$self->{avg_rate} = $self->{avg_n} / $self->{avg_t};
PTDEBUG && _d('Weighted avg rate:', $self->{avg_rate}, 'n/s');
}
else {
$self->{avg_n} = $n;
$self->{avg_t} = $t;
$self->{avg_rate} = $self->{avg_n} / $self->{avg_t};
PTDEBUG && _d('Initial avg rate:', $self->{avg_rate}, 'n/s');
}
my $new_n = int($self->{avg_rate} * $self->{target_t});
PTDEBUG && _d('Adjust n to', $new_n);
return $new_n;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End WeightedAvgRate package
# ###########################################################################
# ###########################################################################
# NibbleIterator package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/NibbleIterator.pm
# t/lib/NibbleIterator.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package NibbleIterator;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Data::Dumper;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(Cxn tbl chunk_size OptionParser Quoter TableNibbler TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $chunk_size, $o, $q) = @args{@required_args};
my $nibble_params = can_nibble(%args);
my %comments = (
bite => "bite table",
nibble => "nibble table",
);
if ( $args{comments} ) {
map { $comments{$_} = $args{comments}->{$_} }
grep { defined $args{comments}->{$_} }
keys %{$args{comments}};
}
my $where = $o->has('where') ? $o->get('where') : '';
my $tbl_struct = $tbl->{tbl_struct};
my $ignore_col = $o->has('ignore-columns')
? ($o->get('ignore-columns') || {})
: {};
my $all_cols = $o->has('columns')
? ($o->get('columns') || $tbl_struct->{cols})
: $tbl_struct->{cols};
my @cols = grep { !$ignore_col->{$_} } @$all_cols;
my $self;
if ( $nibble_params->{one_nibble} ) {
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. " /*$comments{bite}*/";
PTDEBUG && _d('One nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. " /*explain $comments{bite}*/";
PTDEBUG && _d('Explain one nibble statement:', $explain_nibble_sql);
$self = {
%args,
one_nibble => 1,
limit => 0,
nibble_sql => $nibble_sql,
explain_nibble_sql => $explain_nibble_sql,
};
}
else {
my $index = $nibble_params->{index}; # brevity
my $index_cols = $tbl->{tbl_struct}->{keys}->{$index}->{cols};
my $asc = $args{TableNibbler}->generate_asc_stmt(
%args,
tbl_struct => $tbl->{tbl_struct},
index => $index,
cols => \@cols,
asc_only => 1,
);
PTDEBUG && _d('Ascend params:', Dumper($asc));
my $from = "$tbl->{name} FORCE INDEX(`$index`)";
my $order_by = join(', ', map {$q->quote($_)} @{$index_cols});
my $first_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY $order_by"
. " LIMIT 1"
. " /*first lower boundary*/";
PTDEBUG && _d('First lower boundary statement:', $first_lb_sql);
my $resume_lb_sql;
if ( $args{resume} ) {
$resume_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>'}
. ($where ? " AND ($where)" : '')
. " ORDER BY $order_by"
. " LIMIT 1"
. " /*resume lower boundary*/";
PTDEBUG && _d('Resume lower boundary statement:', $resume_lb_sql);
}
my $last_ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY "
. join(' DESC, ', map {$q->quote($_)} @{$index_cols}) . ' DESC'
. " LIMIT 1"
. " /*last upper boundary*/";
PTDEBUG && _d('Last upper boundary statement:', $last_ub_sql);
my $ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='}
. ($where ? " AND ($where)" : '')
. " ORDER BY $order_by"
. " LIMIT ?, 2"
. " /*next chunk boundary*/";
PTDEBUG && _d('Upper boundary statement:', $ub_sql);
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. " /*$comments{nibble}*/";
PTDEBUG && _d('Nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. " /*explain $comments{nibble}*/";
PTDEBUG && _d('Explain nibble statement:', $explain_nibble_sql);
my $limit = $chunk_size - 1;
PTDEBUG && _d('Initial chunk size (LIMIT):', $limit);
$self = {
%args,
index => $index,
limit => $limit,
first_lb_sql => $first_lb_sql,
last_ub_sql => $last_ub_sql,
ub_sql => $ub_sql,
nibble_sql => $nibble_sql,
explain_ub_sql => "EXPLAIN $ub_sql",
explain_nibble_sql => $explain_nibble_sql,
resume_lb_sql => $resume_lb_sql,
sql => {
columns => $asc->{scols},
from => $from,
where => $where,
boundaries => $asc->{boundaries},
order_by => $order_by,
},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
return bless $self, $class;
}
sub next {
my ($self) = @_;
if ( !$self->{oktonibble} ) {
PTDEBUG && _d('Not ok to nibble');
return;
}
my %callback_args = (
Cxn => $self->{Cxn},
tbl => $self->{tbl},
NibbleIterator => $self,
);
if ($self->{nibbleno} == 0) {
$self->_prepare_sths();
$self->_get_bounds();
if ( my $callback = $self->{callbacks}->{init} ) {
$self->{oktonibble} = $callback->(%callback_args);
PTDEBUG && _d('init callback returned', $self->{oktonibble});
if ( !$self->{oktonibble} ) {
$self->{no_more_boundaries} = 1;
return;
}
}
}
NIBBLE:
while ( $self->{have_rows} || $self->_next_boundaries() ) {
if ( !$self->{have_rows} ) {
$self->{nibbleno}++;
PTDEBUG && _d('Nibble:', $self->{nibble_sth}->{Statement}, 'params:',
join(', ', (@{$self->{lower} || []}, @{$self->{upper} || []})));
if ( my $callback = $self->{callbacks}->{exec_nibble} ) {
$self->{have_rows} = $callback->(%callback_args);
}
else {
$self->{nibble_sth}->execute(@{$self->{lower}}, @{$self->{upper}});
$self->{have_rows} = $self->{nibble_sth}->rows();
}
PTDEBUG && _d($self->{have_rows}, 'rows in nibble', $self->{nibbleno});
}
if ( $self->{have_rows} ) {
my $row = $self->{nibble_sth}->fetchrow_arrayref();
if ( $row ) {
$self->{rowno}++;
PTDEBUG && _d('Row', $self->{rowno}, 'in nibble',$self->{nibbleno});
return [ @$row ];
}
}
PTDEBUG && _d('No rows in nibble or nibble skipped');
if ( my $callback = $self->{callbacks}->{after_nibble} ) {
$callback->(%callback_args);
}
$self->{rowno} = 0;
$self->{have_rows} = 0;
}
PTDEBUG && _d('Done nibbling');
if ( my $callback = $self->{callbacks}->{done} ) {
$callback->(%callback_args);
}
return;
}
sub nibble_number {
my ($self) = @_;
return $self->{nibbleno};
}
sub set_nibble_number {
my ($self, $n) = @_;
die "I need a number" unless $n;
$self->{nibbleno} = $n;
PTDEBUG && _d('Set new nibble number:', $n);
return;
}
sub nibble_index {
my ($self) = @_;
return $self->{index};
}
sub statements {
my ($self) = @_;
return {
nibble => $self->{nibble_sth},
explain_nibble => $self->{explain_nibble_sth},
upper_boundary => $self->{ub_sth},
explain_upper_boundary => $self->{explain_ub_sth},
}
}
sub boundaries {
my ($self) = @_;
return {
first_lower => $self->{first_lower},
lower => $self->{lower},
upper => $self->{upper},
next_lower => $self->{next_lower},
last_upper => $self->{last_upper},
};
}
sub set_boundary {
my ($self, $boundary, $values) = @_;
die "I need a boundary parameter"
unless $boundary;
die "Invalid boundary: $boundary"
unless $boundary =~ m/^(?:lower|upper|next_lower|last_upper)$/;
die "I need a values arrayref parameter"
unless $values && ref $values eq 'ARRAY';
$self->{$boundary} = $values;
PTDEBUG && _d('Set new', $boundary, 'boundary:', Dumper($values));
return;
}
sub one_nibble {
my ($self) = @_;
return $self->{one_nibble};
}
sub chunk_size {
my ($self) = @_;
return $self->{limit} + 1;
}
sub set_chunk_size {
my ($self, $limit) = @_;
return if $self->{one_nibble};
die "Chunk size must be > 0" unless $limit;
$self->{limit} = $limit - 1;
PTDEBUG && _d('Set new chunk size (LIMIT):', $limit);
return;
}
sub sql {
my ($self) = @_;
return $self->{sql};
}
sub more_boundaries {
my ($self) = @_;
return !$self->{no_more_boundaries};
}
sub row_estimate {
my ($self) = @_;
return $self->{row_est};
}
sub can_nibble {
my (%args) = @_;
my @required_args = qw(Cxn tbl chunk_size OptionParser TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $chunk_size, $o) = @args{@required_args};
my $where = $o->has('where') ? $o->get('where') : '';
my ($row_est, $mysql_index) = get_row_estimate(
Cxn => $cxn,
tbl => $tbl,
where => $where,
);
if ( !$where ) {
$mysql_index = undef;
}
my $chunk_size_limit = $o->get('chunk-size-limit') || 1;
my $one_nibble = !defined $args{one_nibble} || $args{one_nibble}
? $row_est <= $chunk_size * $chunk_size_limit
: 0;
PTDEBUG && _d('One nibble:', $one_nibble ? 'yes' : 'no');
if ( $args{resume}
&& !defined $args{resume}->{lower_boundary}
&& !defined $args{resume}->{upper_boundary} ) {
PTDEBUG && _d('Resuming from one nibble table');
$one_nibble = 1;
}
my $index = _find_best_index(%args, mysql_index => $mysql_index);
if ( !$index && !$one_nibble ) {
die "There is no good index and the table is oversized.";
}
return {
row_est => $row_est, # nibble about this many rows
index => $index, # using this index
one_nibble => $one_nibble, # if the table fits in one nibble/chunk
};
}
sub _find_best_index {
my (%args) = @_;
my @required_args = qw(Cxn tbl TableParser);
my ($cxn, $tbl, $tp) = @args{@required_args};
my $tbl_struct = $tbl->{tbl_struct};
my $indexes = $tbl_struct->{keys};
my $want_index = $args{chunk_index};
if ( $want_index ) {
PTDEBUG && _d('User wants to use index', $want_index);
if ( !exists $indexes->{$want_index} ) {
PTDEBUG && _d('Cannot use user index because it does not exist');
$want_index = undef;
}
}
if ( !$want_index && $args{mysql_index} ) {
PTDEBUG && _d('MySQL wants to use index', $args{mysql_index});
$want_index = $args{mysql_index};
}
my $best_index;
my @possible_indexes;
if ( $want_index ) {
if ( $indexes->{$want_index}->{is_unique} ) {
PTDEBUG && _d('Will use wanted index');
$best_index = $want_index;
}
else {
PTDEBUG && _d('Wanted index is a possible index');
push @possible_indexes, $want_index;
}
}
else {
PTDEBUG && _d('Auto-selecting best index');
foreach my $index ( $tp->sort_indexes($tbl_struct) ) {
if ( $index eq 'PRIMARY' || $indexes->{$index}->{is_unique} ) {
$best_index = $index;
last;
}
else {
push @possible_indexes, $index;
}
}
}
if ( !$best_index && @possible_indexes ) {
PTDEBUG && _d('No PRIMARY or unique indexes;',
'will use index with highest cardinality');
foreach my $index ( @possible_indexes ) {
$indexes->{$index}->{cardinality} = _get_index_cardinality(
%args,
index => $index,
);
}
@possible_indexes = sort {
my $cmp
= $indexes->{$b}->{cardinality} <=> $indexes->{$b}->{cardinality};
if ( $cmp == 0 ) {
$cmp = scalar @{$indexes->{$b}->{cols}}
<=> scalar @{$indexes->{$a}->{cols}};
}
$cmp;
} @possible_indexes;
$best_index = $possible_indexes[0];
}
PTDEBUG && _d('Best index:', $best_index);
return $best_index;
}
sub _get_index_cardinality {
my (%args) = @_;
my @required_args = qw(Cxn tbl index);
my ($cxn, $tbl, $index) = @args{@required_args};
my $sql = "SHOW INDEXES FROM $tbl->{name} "
. "WHERE Key_name = '$index'";
PTDEBUG && _d($sql);
my $cardinality = 1;
my $dbh = $cxn->dbh();
my $key_name = $dbh && ($dbh->{FetchHashKeyName} || '') eq 'NAME_lc'
? 'key_name'
: 'Key_name';
my $rows = $dbh->selectall_hashref($sql, $key_name);
foreach my $row ( values %$rows ) {
$cardinality *= $row->{cardinality} if $row->{cardinality};
}
PTDEBUG && _d('Index', $index, 'cardinality:', $cardinality);
return $cardinality;
}
sub get_row_estimate {
my (%args) = @_;
my @required_args = qw(Cxn tbl);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl) = @args{@required_args};
my $sql = "EXPLAIN SELECT * FROM $tbl->{name} "
. "WHERE " . ($args{where} || '1=1');
PTDEBUG && _d($sql);
my $expl = $cxn->dbh()->selectrow_hashref($sql);
PTDEBUG && _d(Dumper($expl));
my $mysql_index = $expl->{key} || '';
if ( $mysql_index ne 'PRIMARY' ) {
$mysql_index = lc($mysql_index);
}
return ($expl->{rows} || 0), $mysql_index;
}
sub _prepare_sths {
my ($self) = @_;
PTDEBUG && _d('Preparing statement handles');
my $dbh = $self->{Cxn}->dbh();
$self->{nibble_sth} = $dbh->prepare($self->{nibble_sql});
$self->{explain_nibble_sth} = $dbh->prepare($self->{explain_nibble_sql});
if ( !$self->{one_nibble} ) {
$self->{ub_sth} = $dbh->prepare($self->{ub_sql});
$self->{explain_ub_sth} = $dbh->prepare($self->{explain_ub_sql});
}
return;
}
sub _get_bounds {
my ($self) = @_;
if ( $self->{one_nibble} ) {
if ( $self->{resume} ) {
$self->{no_more_boundaries} = 1;
}
return;
}
my $dbh = $self->{Cxn}->dbh();
$self->{first_lower} = $dbh->selectrow_arrayref($self->{first_lb_sql});
PTDEBUG && _d('First lower boundary:', Dumper($self->{first_lower}));
if ( my $nibble = $self->{resume} ) {
if ( defined $nibble->{lower_boundary}
&& defined $nibble->{upper_boundary} ) {
my $sth = $dbh->prepare($self->{resume_lb_sql});
my @ub = split ',', $nibble->{upper_boundary};
PTDEBUG && _d($sth->{Statement}, 'params:', @ub);
$sth->execute(@ub);
$self->{next_lower} = $sth->fetchrow_arrayref();
$sth->finish();
}
}
else {
$self->{next_lower} = $self->{first_lower};
}
PTDEBUG && _d('Next lower boundary:', Dumper($self->{next_lower}));
if ( !$self->{next_lower} ) {
PTDEBUG && _d('At end of table, or no more boundaries to resume');
$self->{no_more_boundaries} = 1;
$self->{last_upper} = $dbh->selectrow_arrayref($self->{last_ub_sql});
PTDEBUG && _d('Last upper boundary:', Dumper($self->{last_upper}));
}
return;
}
sub _next_boundaries {
my ($self) = @_;
if ( $self->{no_more_boundaries} ) {
PTDEBUG && _d('No more boundaries');
return; # stop nibbling
}
if ( $self->{one_nibble} ) {
$self->{lower} = $self->{upper} = [];
$self->{no_more_boundaries} = 1; # for next call
return 1; # continue nibbling
}
if ( $self->identical_boundaries($self->{lower}, $self->{next_lower}) ) {
PTDEBUG && _d('Infinite loop detected');
my $tbl = $self->{tbl};
my $index = $tbl->{tbl_struct}->{keys}->{$self->{index}};
my $n_cols = scalar @{$index->{cols}};
my $chunkno = $self->{nibbleno};
die "Possible infinite loop detected! "
. "The lower boundary for chunk $chunkno is "
. "<" . join(', ', @{$self->{lower}}) . "> and the lower "
. "boundary for chunk " . ($chunkno + 1) . " is also "
. "<" . join(', ', @{$self->{next_lower}}) . ">. "
. "This usually happens when using a non-unique single "
. "column index. The current chunk index for table "
. "$tbl->{db}.$tbl->{tbl} is $self->{index} which is"
. ($index->{is_unique} ? '' : ' not') . " unique and covers "
. ($n_cols > 1 ? "$n_cols columns" : "1 column") . ".\n";
}
$self->{lower} = $self->{next_lower};
if ( my $callback = $self->{callbacks}->{next_boundaries} ) {
my $oktonibble = $callback->(
Cxn => $self->{Cxn},
tbl => $self->{tbl},
NibbleIterator => $self,
);
PTDEBUG && _d('next_boundaries callback returned', $oktonibble);
if ( !$oktonibble ) {
$self->{no_more_boundaries} = 1;
return; # stop nibbling
}
}
PTDEBUG && _d($self->{ub_sth}->{Statement}, 'params:',
join(', ', @{$self->{lower}}), $self->{limit});
$self->{ub_sth}->execute(@{$self->{lower}}, $self->{limit});
my $boundary = $self->{ub_sth}->fetchall_arrayref();
PTDEBUG && _d('Next boundary:', Dumper($boundary));
if ( $boundary && @$boundary ) {
$self->{upper} = $boundary->[0];
if ( $boundary->[1] ) {
$self->{next_lower} = $boundary->[1];
}
else {
PTDEBUG && _d('End of table boundary:', Dumper($boundary->[0]));
$self->{no_more_boundaries} = 1; # for next call
$self->{last_upper} = $boundary->[0];
}
}
else {
my $dbh = $self->{Cxn}->dbh();
$self->{upper} = $dbh->selectrow_arrayref($self->{last_ub_sql});
PTDEBUG && _d('Last upper boundary:', Dumper($self->{upper}));
$self->{no_more_boundaries} = 1; # for next call
$self->{last_upper} = $self->{upper};
}
$self->{ub_sth}->finish();
return 1; # continue nibbling
}
sub identical_boundaries {
my ($self, $b1, $b2) = @_;
return 0 if ($b1 && !$b2) || (!$b1 && $b2);
return 1 if !$b1 && !$b2;
die "Boundaries have different numbers of values"
if scalar @$b1 != scalar @$b2; # shouldn't happen
my $n_vals = scalar @$b1;
for my $i ( 0..($n_vals-1) ) {
return 0 if $b1->[$i] ne $b2->[$i]; # diff
}
return 1;
}
sub DESTROY {
my ( $self ) = @_;
foreach my $key ( keys %$self ) {
if ( $key =~ m/_sth$/ ) {
PTDEBUG && _d('Finish', $key);
$self->{$key}->finish();
}
}
return;
}
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;
}
# ###########################################################################
# End NibbleIterator package
# ###########################################################################
# ###########################################################################
# Transformers package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/Transformers.pm
# t/lib/Transformers.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package Transformers;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Time::Local qw(timegm timelocal);
use Digest::MD5 qw(md5_hex);
require Exporter;
our @ISA = qw(Exporter);
our %EXPORT_TAGS = ();
our @EXPORT = ();
our @EXPORT_OK = qw(
micro_t
percentage_of
secs_to_time
time_to_secs
shorten
ts
parse_timestamp
unix_timestamp
any_unix_timestamp
make_checksum
crc32
);
our $mysql_ts = qr/(\d\d)(\d\d)(\d\d) +(\d+):(\d+):(\d+)(\.\d+)?/;
our $proper_ts = qr/(\d\d\d\d)-(\d\d)-(\d\d)[T ](\d\d):(\d\d):(\d\d)(\.\d+)?/;
our $n_ts = qr/(\d{1,5})([shmd]?)/; # Limit \d{1,5} because \d{6} looks
sub micro_t {
my ( $t, %args ) = @_;
my $p_ms = defined $args{p_ms} ? $args{p_ms} : 0; # precision for ms vals
my $p_s = defined $args{p_s} ? $args{p_s} : 0; # precision for s vals
my $f;
$t = 0 if $t < 0;
$t = sprintf('%.17f', $t) if $t =~ /e/;
$t =~ s/\.(\d{1,6})\d*/\.$1/;
if ($t > 0 && $t <= 0.000999) {
$f = ($t * 1000000) . 'us';
}
elsif ($t >= 0.001000 && $t <= 0.999999) {
$f = sprintf("%.${p_ms}f", $t * 1000);
$f = ($f * 1) . 'ms'; # * 1 to remove insignificant zeros
}
elsif ($t >= 1) {
$f = sprintf("%.${p_s}f", $t);
$f = ($f * 1) . 's'; # * 1 to remove insignificant zeros
}
else {
$f = 0; # $t should = 0 at this point
}
return $f;
}
sub percentage_of {
my ( $is, $of, %args ) = @_;
my $p = $args{p} || 0; # float precision
my $fmt = $p ? "%.${p}f" : "%d";
return sprintf $fmt, ($is * 100) / ($of ||= 1);
}
sub secs_to_time {
my ( $secs, $fmt ) = @_;
$secs ||= 0;
return '00:00' unless $secs;
$fmt ||= $secs >= 86_400 ? 'd'
: $secs >= 3_600 ? 'h'
: 'm';
return
$fmt eq 'd' ? sprintf(
"%d+%02d:%02d:%02d",
int($secs / 86_400),
int(($secs % 86_400) / 3_600),
int(($secs % 3_600) / 60),
$secs % 60)
: $fmt eq 'h' ? sprintf(
"%02d:%02d:%02d",
int(($secs % 86_400) / 3_600),
int(($secs % 3_600) / 60),
$secs % 60)
: sprintf(
"%02d:%02d",
int(($secs % 3_600) / 60),
$secs % 60);
}
sub time_to_secs {
my ( $val, $default_suffix ) = @_;
die "I need a val argument" unless defined $val;
my $t = 0;
my ( $prefix, $num, $suffix ) = $val =~ m/([+-]?)(\d+)([a-z])?$/;
$suffix = $suffix || $default_suffix || 's';
if ( $suffix =~ m/[smhd]/ ) {
$t = $suffix eq 's' ? $num * 1 # Seconds
: $suffix eq 'm' ? $num * 60 # Minutes
: $suffix eq 'h' ? $num * 3600 # Hours
: $num * 86400; # Days
$t *= -1 if $prefix && $prefix eq '-';
}
else {
die "Invalid suffix for $val: $suffix";
}
return $t;
}
sub shorten {
my ( $num, %args ) = @_;
my $p = defined $args{p} ? $args{p} : 2; # float precision
my $d = defined $args{d} ? $args{d} : 1_024; # divisor
my $n = 0;
my @units = ('', qw(k M G T P E Z Y));
while ( $num >= $d && $n < @units - 1 ) {
$num /= $d;
++$n;
}
return sprintf(
$num =~ m/\./ || $n
? "%.${p}f%s"
: '%d',
$num, $units[$n]);
}
sub ts {
my ( $time, $gmt ) = @_;
my ( $sec, $min, $hour, $mday, $mon, $year )
= $gmt ? gmtime($time) : localtime($time);
$mon += 1;
$year += 1900;
my $val = sprintf("%d-%02d-%02dT%02d:%02d:%02d",
$year, $mon, $mday, $hour, $min, $sec);
if ( my ($us) = $time =~ m/(\.\d+)$/ ) {
$us = sprintf("%.6f", $us);
$us =~ s/^0\././;
$val .= $us;
}
return $val;
}
sub parse_timestamp {
my ( $val ) = @_;
if ( my($y, $m, $d, $h, $i, $s, $f)
= $val =~ m/^$mysql_ts$/ )
{
return sprintf "%d-%02d-%02d %02d:%02d:"
. (defined $f ? '%09.6f' : '%02d'),
$y + 2000, $m, $d, $h, $i, (defined $f ? $s + $f : $s);
}
return $val;
}
sub unix_timestamp {
my ( $val, $gmt ) = @_;
if ( my($y, $m, $d, $h, $i, $s, $us) = $val =~ m/^$proper_ts$/ ) {
$val = $gmt
? timegm($s, $i, $h, $d, $m - 1, $y)
: timelocal($s, $i, $h, $d, $m - 1, $y);
if ( defined $us ) {
$us = sprintf('%.6f', $us);
$us =~ s/^0\././;
$val .= $us;
}
}
return $val;
}
sub any_unix_timestamp {
my ( $val, $callback ) = @_;
if ( my ($n, $suffix) = $val =~ m/^$n_ts$/ ) {
$n = $suffix eq 's' ? $n # Seconds
: $suffix eq 'm' ? $n * 60 # Minutes
: $suffix eq 'h' ? $n * 3600 # Hours
: $suffix eq 'd' ? $n * 86400 # Days
: $n; # default: Seconds
PTDEBUG && _d('ts is now - N[shmd]:', $n);
return time - $n;
}
elsif ( $val =~ m/^\d{9,}/ ) {
PTDEBUG && _d('ts is already a unix timestamp');
return $val;
}
elsif ( my ($ymd, $hms) = $val =~ m/^(\d{6})(?:\s+(\d+:\d+:\d+))?/ ) {
PTDEBUG && _d('ts is MySQL slow log timestamp');
$val .= ' 00:00:00' unless $hms;
return unix_timestamp(parse_timestamp($val));
}
elsif ( ($ymd, $hms) = $val =~ m/^(\d{4}-\d\d-\d\d)(?:[T ](\d+:\d+:\d+))?/) {
PTDEBUG && _d('ts is properly formatted timestamp');
$val .= ' 00:00:00' unless $hms;
return unix_timestamp($val);
}
else {
PTDEBUG && _d('ts is MySQL expression');
return $callback->($val) if $callback && ref $callback eq 'CODE';
}
PTDEBUG && _d('Unknown ts type:', $val);
return;
}
sub make_checksum {
my ( $val ) = @_;
my $checksum = uc substr(md5_hex($val), -16);
PTDEBUG && _d($checksum, 'checksum for', $val);
return $checksum;
}
sub crc32 {
my ( $string ) = @_;
return unless $string;
my $poly = 0xEDB88320;
my $crc = 0xFFFFFFFF;
foreach my $char ( split(//, $string) ) {
my $comp = ($crc ^ ord($char)) & 0xFF;
for ( 1 .. 8 ) {
$comp = $comp & 1 ? $poly ^ ($comp >> 1) : $comp >> 1;
}
$crc = (($crc >> 8) & 0x00FFFFFF) ^ $comp;
}
return $crc ^ 0xFFFFFFFF;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End Transformers package
# ###########################################################################
# ###########################################################################
# CleanupTask package
# This package is a copy without comments from the original. The original
# with comments and its test file can be found in the Bazaar repository at,
# lib/CleanupTask.pm
# t/lib/CleanupTask.t
# See https://launchpad.net/percona-toolkit for more information.
# ###########################################################################
{
package CleanupTask;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
sub new {
my ( $class, $task ) = @_;
die "I need a task parameter" unless $task;
die "The task parameter must be a coderef" unless ref $task eq 'CODE';
my $self = {
task => $task,
};
PTDEBUG && _d('Created cleanup task', $task);
return bless $self, $class;
}
sub DESTROY {
my ($self) = @_;
my $task = $self->{task};
if ( ref $task ) {
PTDEBUG && _d('Calling cleanup task', $task);
$task->();
}
else {
warn "Lost cleanup task";
}
return;
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End CleanupTask package
# ###########################################################################
# ###########################################################################
# This is a combination of modules and programs in one -- a runnable module.
# http://www.perl.com/pub/a/2006/07/13/lightning-articles.html?page=last
# Or, look it up in the Camel book on pages 642 and 643 in the 3rd edition.
#
# Check at the end of this package for the call to main() which actually runs
# the program.
# ###########################################################################
package pt_online_schema_change;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Time::HiRes qw(time sleep);
use Data::Dumper;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
use sigtrap 'handler', \&sig_int, 'normal-signals';
my $exit_status = 0;
my $oktorun = 1;
my @drop_trigger_sqls;
$OUTPUT_AUTOFLUSH = 1;
sub main {
# Reset global vars else tests will fail.
@ARGV = @_;
$oktorun = 1;
@drop_trigger_sqls = ();
$exit_status = 0;
# ########################################################################
# Get configuration information.
# ########################################################################
my $q = new Quoter();
my $o = new OptionParser();
$o->get_specs();
$o->get_opts();
my $dp = $o->DSNParser();
$dp->prop('set-vars', $o->get('set-vars'));
# The original table, i.e. the one being altered, must be specified
# on the command line via the DSN.
my ($db, $tbl);
my $dsn = shift @ARGV;
if ( !$dsn ) {
$o->save_error('A DSN must be specified');
}
else {
# Parse DSN string and convert it to a DSN data struct.
$dsn = $dp->parse($dsn, $dp->parse_options($o));
$db = $dsn->{D};
$tbl = $dsn->{t};
}
my $alter_fk_method = $o->get('alter-foreign-keys-method') || '';
if ( $alter_fk_method eq 'drop_swap' ) {
$o->set('swap-tables', 0);
$o->set('drop-old-table', 0);
}
# Explicit --chunk-size disable auto chunk sizing.
$o->set('chunk-time', 0) if $o->got('chunk-size');
foreach my $opt ( qw(max-load critical-load) ) {
next unless $o->has($opt);
my $spec = $o->get($opt);
eval {
MySQLStatusWaiter::_parse_spec($o->get($opt));
};
if ( $EVAL_ERROR ) {
chomp $EVAL_ERROR;
$o->save_error("Invalid --$opt: $EVAL_ERROR");
}
}
if ( !$o->get('help') ) {
if ( @ARGV ) {
$o->save_error('Specify only one DSN on the command line');
}
if ( !$db || !$tbl ) {
$o->save_error("The DSN must specify a database (D) and a table (t)");
}
if ( $o->get('progress') ) {
eval { Progress->validate_spec($o->get('progress')) };
if ( $EVAL_ERROR ) {
chomp $EVAL_ERROR;
$o->save_error("--progress $EVAL_ERROR");
}
}
# See the "pod-based-option-value-validation" spec for how this may
# be automagically validated.
if ( $alter_fk_method
&& $alter_fk_method ne 'auto'
&& $alter_fk_method ne 'rebuild_constraints'
&& $alter_fk_method ne 'drop_swap'
&& $alter_fk_method ne 'none' )
{
$o->save_error("Invalid --alter-foreign-keys-method value: $alter_fk_method");
}
if ( $alter_fk_method eq 'drop_swap' && !$o->get('drop-new-table') ) {
$o->save_error("--alter-foreign-keys-method=drop_swap does not work with --no-drop-new-table.");
}
}
$o->usage_or_errors();
if ( $o->get('quiet') ) {
# BARON: this will fail on Windows, where there is no /dev/null. I feel
# it's a hack, like ignoring a problem instead of fixing it somehow. We
# should take a look at the things that get printed in a "normal"
# non-quiet run, and "if !quiet" them, and then do some kind of Logger.pm
# or Messager.pm module for a future release.
close STDOUT;
open STDOUT, '>', '/dev/null'
or warn "Cannot reopen STDOUT to /dev/null: $OS_ERROR";
}
# ########################################################################
# Connect to MySQL.
# ########################################################################
my $set_on_connect = sub {
my ($dbh) = @_;
# See the same code in pt-table-checksum.
my $lock_wait_timeout = $o->get('lock-wait-timeout');
my $set_lwt = "SET SESSION innodb_lock_wait_timeout=$lock_wait_timeout";
PTDEBUG && _d($set_lwt);
eval {
$dbh->do($set_lwt);
};
if ( $EVAL_ERROR ) {
PTDEBUG && _d($EVAL_ERROR);
# Get the server's current value.
my $sql = "SHOW SESSION VARIABLES LIKE 'innodb_lock_wait_timeout'";
PTDEBUG && _d($dbh, $sql);
my (undef, $curr_lwt) = $dbh->selectrow_array($sql);
PTDEBUG && _d('innodb_lock_wait_timeout on server:', $curr_lwt);
if ( !defined $curr_lwt ) {
PTDEBUG && _d('innodb_lock_wait_timeout does not exist;',
'InnoDB is probably disabled');
}
elsif ( $curr_lwt > $lock_wait_timeout ) {
warn "Failed to $set_lwt: $EVAL_ERROR\n"
. "The current innodb_lock_wait_timeout value "
. "$curr_lwt is greater than the --lock-wait-timeout "
. "value $lock_wait_timeout and the variable cannot be "
. "changed. innodb_lock_wait_timeout is only dynamic when "
. "using the InnoDB plugin. To prevent this warning, either "
. "specify --lock-wait-time=$curr_lwt, or manually set "
. "innodb_lock_wait_timeout to a value less than or equal "
. "to $lock_wait_timeout and restart MySQL.\n";
}
}
};
# Do not call "new Cxn(" directly; use this sub so that set_on_connect
# is applied to every cxn.
# BARON: why not make this a subroutine instead of a subroutine variable? I
# think that can be less confusing. Also, the $set_on_connect variable can be
# inlined into this subroutine. Many of our tools have a get_dbh() subroutine
# and it might be good to just make a convention of it.
my $make_cxn = sub {
my (%args) = @_;
my $cxn = new Cxn(
%args,
DSNParser => $dp,
OptionParser => $o,
set => $set_on_connect,
);
eval { $cxn->connect() }; # connect or die trying
if ( $EVAL_ERROR ) {
die "Cannot connect to MySQL: $EVAL_ERROR\n";
}
return $cxn;
};
my $cxn = $make_cxn->(dsn => $dsn);
# ########################################################################
# Check if MySQL is new enough to have the triggers we need.
# Although triggers were introduced in 5.0.2, "Prior to MySQL 5.0.10,
# triggers cannot contain direct references to tables by name."
# ########################################################################
my $vp = new VersionParser();
if ( !$vp->version_ge($cxn->dbh(), '5.0.10') ) {
die "This tool requires MySQL 5.0.10 or newer.\n";
}
# ########################################################################
# Setup lag and load monitors.
# ########################################################################
my $slaves; # all slaves that are found or specified
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
my $sys_load_pr; # Progress for MySQLStatusWaiter object
if ( $o->get('execute') ) {
# #####################################################################
# Find and connect to slaves.
# #####################################################################
my $ms = new MasterSlave();
$slaves = $ms->get_slaves(
dbh => $cxn->dbh(),
dsn => $cxn->dsn(),
OptionParser => $o,
DSNParser => $dp,
Quoter => $q,
make_cxn => sub {
return $make_cxn->(@_, prev_dsn => $cxn->dsn());
},
);
PTDEBUG && _d(scalar @$slaves, 'slaves found');
if ( $o->get('check-slave-lag') ) {
PTDEBUG && _d('Will use --check-slave-lag to check for slave lag');
my $cxn = $make_cxn->(
dsn_string => $o->get('check-slave-lag'),
prev_dsn => $cxn->dsn(),
);
$slave_lag_cxns = [ $cxn ];
}
else {
PTDEBUG && _d('Will check slave lag on all slaves');
$slave_lag_cxns = $slaves;
}
# #####################################################################
# Check for replication filters.
# #####################################################################
if ( $o->get('check-replication-filters') ) {
PTDEBUG && _d("Checking slave replication filters");
my @all_repl_filters;
foreach my $slave ( @$slaves ) {
my $repl_filters = $ms->get_replication_filters(
dbh => $slave->dbh(),
);
if ( keys %$repl_filters ) {
push @all_repl_filters,
{ name => $slave->name(),
filters => $repl_filters,
};
}
}
if ( @all_repl_filters ) {
my $msg = "Replication filters are set on these hosts:\n";
foreach my $host ( @all_repl_filters ) {
my $filters = $host->{filters};
$msg .= " $host->{name}\n"
. join("\n", map { " $_ = $host->{filters}->{$_}" }
keys %{$host->{filters}})
. "\n";
}
$msg .= "Please read the --check-replication-filters documentation "
. "to learn how to solve this problem.";
die $msg;
}
}
# #####################################################################
# Make a ReplicaLagWaiter to help wait for slaves after each chunk.
# #####################################################################
my $sleep = sub {
# Don't let the master dbh die while waiting for slaves because we
# may wait a very long time for slaves.
my $dbh = $cxn->dbh();
if ( !$dbh || !$dbh->ping() ) {
eval { $dbh = $cxn->connect() }; # connect or die trying
if ( $EVAL_ERROR ) {
$oktorun = 0; # flag for cleanup tasks
chomp $EVAL_ERROR;
die "Lost connection to " . $cxn->name() . " while waiting for "
. "replica lag ($EVAL_ERROR)\n";
}
}
$dbh->do("SELECT 'pt-online-schema-change keepalive'");
sleep $o->get('check-interval');
return;
};
my $get_lag = sub {
my ($cxn) = @_;
my $dbh = $cxn->dbh();
if ( !$dbh || !$dbh->ping() ) {
eval { $dbh = $cxn->connect() }; # connect or die trying
if ( $EVAL_ERROR ) {
$oktorun = 0; # flag for cleanup tasks
chomp $EVAL_ERROR;
die "Lost connection to replica " . $cxn->name()
. " while attempting to get its lag ($EVAL_ERROR)\n";
}
}
return $ms->get_slave_lag($dbh);
};
$replica_lag = new ReplicaLagWaiter(
slaves => $slave_lag_cxns,
max_lag => $o->get('max-lag'),
oktorun => sub { return $oktorun },
get_lag => $get_lag,
sleep => $sleep,
);
my $get_status;
{
my $sql = "SHOW GLOBAL STATUS LIKE ?";
my $sth = $cxn->dbh()->prepare($sql);
$get_status = sub {
my ($var) = @_;
PTDEBUG && _d($sth->{Statement}, $var);
$sth->execute($var);
my (undef, $val) = $sth->fetchrow_array();
return $val;
};
}
eval {
$sys_load = new MySQLStatusWaiter(
max_spec => $o->get('max-load'),
critical_spec => $o->get('critical-load'),
get_status => $get_status,
oktorun => sub { return $oktorun },
sleep => $sleep,
);
};
if ( $EVAL_ERROR ) {
chomp $EVAL_ERROR;
die "Error checking --max-load or --critial-load: $EVAL_ERROR. "
. "Check that the variables specified for --max-load and "
. "--critical-load are spelled correctly and exist in "
. "SHOW GLOBAL STATUS. Current values for these options are:\n"
. " --max-load " . (join(',', @{$o->get('max-load')})) . "\n"
. " --critial-load " . (join(',', @{$o->get('critical-load')}))
. "\n";
}
if ( $o->get('progress') ) {
$replica_lag_pr = new Progress(
jobsize => scalar @$slaves,
spec => $o->get('progress'),
name => "Waiting for replicas to catch up", # not used
);
$sys_load_pr = new Progress(
jobsize => scalar @{$o->get('max-load')},
spec => $o->get('progress'),
name => "Waiting for --max-load", # not used
);
}
}
# ########################################################################
# Setup and check the original table.
# ########################################################################
my $tp = new TableParser(Quoter => $q);
# Common table data struct (that modules like NibbleIterator expect).
my $orig_tbl = {
db => $db,
tbl => $tbl,
name => $q->quote($db, $tbl),
};
check_orig_table(
orig_tbl => $orig_tbl,
Cxn => $cxn,
OptionParser => $o,
TableParser => $tp,
Quoter => $q,
);
# ########################################################################
# Get child tables of the original table, if necessary.
# ########################################################################
my $child_tables;
if ( ($alter_fk_method || '') eq 'none' ) {
print "Not updating foreign keys because "
. "--alter-foreign-keys-method=none. Foreign keys "
. "that reference the table will no longer work.\n";
}
else {
$child_tables = find_child_tables(
tbl => $orig_tbl,
Cxn => $cxn,
Quoter => $q,
);
if ( !$child_tables ) {
if ( $alter_fk_method ) {
warn "No foreign keys reference $orig_tbl->{name}; ignoring "
. "--alter-foreign-keys-method.\n";
if ( $alter_fk_method eq 'drop_swap' ) {
# These opts are disabled at the start if the user specifies
# the drop_swap method, but now that we know there are no
# child tables, we must re-enable these to make the alter work.
$o->set('swap-tables', 1);
$o->set('drop-old-table', 1);
}
$alter_fk_method = '';
}
# No child tables and --alter-fk-method wasn't specified,
# so nothing to do.
}
else {
print "Child tables:\n";
foreach my $child_table ( @$child_tables ) {
printf " %s (approx. %s rows)\n",
$child_table->{name},
$child_table->{row_est} || '?';
}
if ( $alter_fk_method ) {
# Let the user know how we're going to update the child table
# fk refs.
my $choice
= $alter_fk_method eq 'none' ? "not"
: $alter_fk_method eq 'auto' ? "automatically choose the method to"
: "use the $alter_fk_method method to";
print "Will $choice update foreign keys.\n";
}
else {
print "You did not specify --alter-foreign-keys-method, but there "
. "are foreign keys that reference the table. "
. "Please read the tool's documentation carefully.\n";
return 1;
}
}
}
# ########################################################################
# XXX
# Ready to begin the alter! Nothing has been changed on the server at
# this point; we've just checked and looked for things. Past this point,
# the code is live if --execute, else it's doing a --dry-run. Or, if
# the user didn't read the docs, we may bail out here.
# XXX
# ########################################################################
if ( $o->get('dry-run') ) {
print "Starting a dry run. $orig_tbl->{name} will not be altered. "
. "Specify --execute instead of --dry-run to alter the table.\n";
}
elsif ( $o->get('execute') ) {
print "Altering $orig_tbl->{name}...\n";
}
else {
print "Exiting without altering $orig_tbl->{name} because neither "
. "--dry-run nor --execute was specified. Please read the tool's "
. "documentation carefully before using this tool.\n";
return 1;
}
# ########################################################################
# Create a cleanup task object to undo changes (i.e. clean up) if the
# code dies, or we may call this explicitly at the end if all goes well.
# ########################################################################
my @cleanup_tasks;
my $cleanup = new CleanupTask(
sub {
# XXX We shouldn't copy $EVAL_ERROR here, but I found that
# errors are not re-thrown in tests. If you comment out this
# line and the die below, an error fails:
# not ok 5 - Doesn't try forever to find a new table name
# Failed test 'Doesn't try forever to find a new table name'
# at /Users/daniel/p/pt-osc-2.1.1/lib/PerconaTest.pm line 559.
# ''
# doesn't match '(?-xism:Failed to find a unique new table name)'
my $original_error = $EVAL_ERROR;
foreach my $task ( reverse @cleanup_tasks ) {
eval {
$task->();
};
if ( $EVAL_ERROR ) {
warn "Error cleaning up: $EVAL_ERROR\n";
}
}
die $original_error if $original_error; # rethrow original error
return;
}
);
# The last cleanup task is to report whether or not the orig table
# was altered.
push @cleanup_tasks, sub {
PTDEBUG && _d('Clean up done, report if orig table was altered');
if ( $o->get('dry-run') ) {
print "Dry run complete. $orig_tbl->{name} was not altered.\n";
}
else {
if ( $orig_tbl->{swapped} ) {
if ( $orig_tbl->{success} ) {
print "Successfully altered $orig_tbl->{name}.\n";
}
else {
print "Altered $orig_tbl->{name} but there were errors "
. "or warnings.\n";
}
}
else {
print "$orig_tbl->{name} was not altered.\n";
}
}
return;
};
# ########################################################################
# Check and create PID file if user specified --pid.
# ########################################################################
my $daemon;
if ( $o->get('execute') && $o->get('pid') ) {
# We're not daemoninzing, it just handles PID stuff.
$daemon = new Daemon(o=>$o);
$daemon->make_PID_file();
}
# #####################################################################
# Step 1: Create the new table.
# #####################################################################
my $new_tbl;
eval {
$new_tbl = create_new_table(
orig_tbl => $orig_tbl,
suffix => '_new',
Cxn => $cxn,
Quoter => $q,
OptionParser => $o,
TableParser => $tp,
);
};
if ( $EVAL_ERROR ) {
die "Error creating new table: $EVAL_ERROR\n";
}
# If the new table still exists, drop it unless the tool was interrupted.
push @cleanup_tasks, sub {
PTDEBUG && _d('Clean up new table');
my $new_tbl_exists = $tp->check_table(
dbh => $cxn->dbh(),
db => $new_tbl->{db},
tbl => $new_tbl->{tbl},
);
PTDEBUG && _d('New table exists:', $new_tbl_exists ? 'yes' : 'no');
return unless $new_tbl_exists;
my $sql = "DROP TABLE IF EXISTS $new_tbl->{name};";
if ( !$oktorun ) {
# The tool was interrupted, so do not drop the new table
# in case the user wants to resume (once resume capability
# is implemented).
print "Not dropping the new table $new_tbl->{name} because "
. "the tool was interrupted. To drop the new table, "
. "execute:\n$sql\n";
}
elsif ( $orig_tbl->{copied} && !$orig_tbl->{swapped} ) {
print "Not dropping the new table $new_tbl->{name} because "
. "--swap-tables failed. To drop the new table, "
. "execute:\n$sql\n";
}
elsif ( !$o->get('drop-new-table') ) {
# https://bugs.launchpad.net/percona-toolkit/+bug/998831
print "Not dropping the new table $new_tbl->{name} because "
. "--no-drop-new-table was specified. To drop the new table, "
. "execute:\n$sql\n";
}
else {
print "Dropping new table...\n";
print $sql, "\n" if $o->get('print');
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
warn "Error dropping new table $new_tbl->{name}: $EVAL_ERROR\n"
. "To try dropping the new table again, execute:\n$sql\n";
}
print "Dropped new table OK.\n";
}
};
# #####################################################################
# Step 2: Alter the new, empty table. This should be very quick,
# or die if the user specified a bad alter statement.
# #####################################################################
if ( my $alter = $o->get('alter') ) {
print "Altering new table...\n";
my $sql = "ALTER TABLE $new_tbl->{name} $alter";
print $sql, "\n" if $o->get('print');
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
die "Error altering new table $new_tbl->{name}: $EVAL_ERROR\n"
}
print "Altered $new_tbl->{name} OK.\n"
}
# Get the new table struct. This shouldn't die because
# we just created the table successfully so we know it's
# there. But the ghost of Ryan is everywhere.
my $ddl = $tp->get_create_table(
$cxn->dbh(),
$new_tbl->{db},
$new_tbl->{tbl},
);
$new_tbl->{tbl_struct} = $tp->parse($ddl);
# Determine what columns the original and new table share.
# If the user drops a col, that's easy: just don't copy it. If they
# add a column, it must have a default value. Other alterations
# may or may not affect the copy process--we'll know when we try!
# Note: we don't want to examine the --alter statement to see if the
# cols have changed because that's messy and prone to parsing errors.
# Col posn (position) is just for looks because user's like
# to see columns listed in their original order, not Perl's
# random hash key sorting.
my $col_posn = $orig_tbl->{tbl_struct}->{col_posn};
my $orig_cols = $orig_tbl->{tbl_struct}->{is_col};
my $new_cols = $new_tbl->{tbl_struct}->{is_col};
my @common_cols = sort { $col_posn->{$a} <=> $col_posn->{$b} }
grep { $new_cols->{$_} }
keys %$orig_cols;
PTDEBUG && _d('Common columns', @common_cols);
# ########################################################################
# Step 3: Create the triggers to capture changes on the original table and
# apply them to the new table.
# ########################################################################
# Drop the triggers. We can save this cleanup task before
# adding the triggers because if adding them fails, this will be
# called which will drop whichever triggers were created.
push @cleanup_tasks, sub {
PTDEBUG && _d('Clean up triggers');
if ( $oktorun ) {
drop_triggers(
tbl => $orig_tbl,
Cxn => $cxn,
Quoter => $q,
OptionParser => $o,
);
}
else {
print "Not dropping triggers because the tool was interrupted. "
. "To drop the triggers, execute:\n"
. join("\n", @drop_trigger_sqls) . "\n";
}
};
eval {
create_triggers(
orig_tbl => $orig_tbl,
new_tbl => $new_tbl,
columns => \@common_cols,
Cxn => $cxn,
Quoter => $q,
OptionParser => $o,
);
};
if ( $EVAL_ERROR ) {
die "Error creating triggers: $EVAL_ERROR\n";
};
# #####################################################################
# Step 4: Copy rows.
# #####################################################################
# The hashref of callbacks below is what NibbleIterator calls internally
# to do all the copy work. The callbacks do not need to eval their work
# because the higher call to $nibble_iter->next() is eval'ed which will
# catch any errors in the callbacks.
my $total_rows = 0;
my $total_time = 0;
my $avg_rate = 0; # rows/second
my $retry = new Retry(); # for retrying to exec the copy statement
my $limit = $o->get('chunk-size-limit'); # brevity
my $chunk_time = $o->get('chunk-time'); # brevity
my $callbacks = {
init => sub {
my (%args) = @_;
my $tbl = $args{tbl};
my $nibble_iter = $args{NibbleIterator};
if ( $o->get('dry-run') ) {
print "Not copying rows because this is a dry run.\n";
}
else {
print "Copying approximately ", $nibble_iter->row_estimate(),
" rows...\n";
}
if ( $o->get('print') ) {
# Print the checksum and next boundary statements.
my $statements = $nibble_iter->statements();
foreach my $sth ( sort keys %$statements ) {
next if $sth =~ m/^explain/;
if ( $statements->{$sth} ) {
print $statements->{$sth}->{Statement}, "\n";
}
}
}
return unless $o->get('execute');
# If table is a single chunk on the master, make sure it's also
# a single chunk on all slaves. E.g. if a slave is out of sync
# and has a lot more rows than the master, single chunking on the
# master could cause the slave to choke.
if ( $nibble_iter->one_nibble() ) {
PTDEBUG && _d('Getting table row estimate on replicas');
my @too_large;
foreach my $slave ( @$slaves ) {
my ($n_rows) = NibbleIterator::get_row_estimate(
Cxn => $slave,
tbl => $tbl,
);
PTDEBUG && _d('Table on',$slave->name(),'has', $n_rows, 'rows');
if ( $n_rows && $n_rows > ($tbl->{chunk_size} * $limit) ) {
PTDEBUG && _d('Table too large on', $slave->name());
push @too_large, [$slave->name(), $n_rows || 0];
}
}
if ( @too_large ) {
my $msg
= "Cannot copy table $tbl->{name} because"
. " on the master it would be checksummed in one chunk"
. " but on these replicas it has too many rows:\n";
foreach my $info ( @too_large ) {
$msg .= " $info->[1] rows on $info->[0]\n";
}
$msg .= "The current chunk size limit is "
. ($tbl->{chunk_size} * $limit)
. " rows (chunk size=$tbl->{chunk_size}"
. " * chunk size limit=$limit).\n";
die $msg;
}
}
return 1; # continue nibbling table
},
next_boundaries => sub {
my (%args) = @_;
my $tbl = $args{tbl};
my $nibble_iter = $args{NibbleIterator};
my $sth = $nibble_iter->statements();
my $boundary = $nibble_iter->boundaries();
return 0 if $o->get('dry-run');
return 1 if $nibble_iter->one_nibble();
# Check that MySQL will use the nibble index for the next upper
# boundary sql. This check applies to the next nibble. So if
# the current nibble number is 5, then nibble 5 is already done
# and we're checking nibble number 6.
my $expl = explain_statement(
tbl => $tbl,
sth => $sth->{explain_upper_boundary},
vals => [ @{$boundary->{lower}}, $nibble_iter->chunk_size() ],
);
if (lc($expl->{key} || '') ne lc($nibble_iter->nibble_index() || '')) {
my $msg
= "Aborting copying table $tbl->{name} at chunk "
. ($nibble_iter->nibble_number() + 1)
. " because it is not safe to ascend. Chunking should "
. "use the "
. ($nibble_iter->nibble_index() || '?')
. " index, but MySQL EXPLAIN reports that "
. ($expl->{key} ? "the $expl->{key}" : "no")
. " index will be used for "
. $sth->{upper_boundary}->{Statement}
. " with values "
. join(", ", map { defined $_ ? $_ : "NULL" }
(@{$boundary->{lower}}, $nibble_iter->chunk_size()))
. "\n";
die $msg;
}
# Once nibbling begins for a table, control does not return to this
# tool until nibbling is done because, as noted above, all work is
# done in these callbacks. This callback is the only place where we
# can prematurely stop nibbling by returning false. This allows
# Ctrl-C to stop the tool between nibbles instead of between tables.
return $oktorun; # continue nibbling table?
},
exec_nibble => sub {
my (%args) = @_;
my $tbl = $args{tbl};
my $nibble_iter = $args{NibbleIterator};
my $sth = $nibble_iter->statements();
my $boundary = $nibble_iter->boundaries();
return if $o->get('dry-run');
# Count every chunk, even if it's ultimately skipped, etc.
$tbl->{results}->{n_chunks}++;
# If the table is being chunk (i.e., it's not small enough to be
# consumed by one nibble), then check index usage and chunk size.
if ( !$nibble_iter->one_nibble() ) {
my $expl = explain_statement(
tbl => $tbl,
sth => $sth->{explain_nibble},
vals => [ @{$boundary->{lower}}, @{$boundary->{upper}} ],
);
# Ensure that MySQL is using the chunk index.
if ( lc($expl->{key} || '')
ne lc($nibble_iter->nibble_index() || '') ) {
my $msg
= "Aborting copying table $tbl->{name} at chunk "
. $nibble_iter->nibble_number()
. " because it is not safe to chunk. Chunking should "
. "use the "
. ($nibble_iter->nibble_index() || '?')
. " index, but MySQL EXPLAIN reports that "
. ($expl->{key} ? "the $expl->{key}" : "no")
. " index will be used for "
. $sth->{explain_nibble}->{Statement}
. " with values "
. join(", ", map { defined $_ ? $_ : "NULL" }
(@{$boundary->{lower}}, @{$boundary->{upper}}))
. "\n";
die $msg;
}
# Check chunk size limit if the upper boundary and next lower
# boundary are identical.
if ( $limit ) {
my $boundary = $nibble_iter->boundaries();
my $oversize_chunk
= $limit ? ($expl->{rows} || 0) >= $tbl->{chunk_size} * $limit
: 0;
if ( $oversize_chunk
&& $nibble_iter->identical_boundaries(
$boundary->{upper}, $boundary->{next_lower}) )
{
my $msg
= "Aborting copying table $tbl->{name} at chunk "
. $nibble_iter->nibble_number()
. " because the chunk is too large: MySQL estimates "
. ($expl->{rows} || 0) . "rows. The current chunk "
. "size limit is " . ($tbl->{chunk_size} * $limit)
. " rows (chunk size=$tbl->{chunk_size}"
. " * chunk size limit=$limit).\n";
die $msg;
}
}
}
# Exec and time the chunk checksum query.
$tbl->{nibble_time} = exec_nibble(
%args,
Retry => $retry,
Quoter => $q,
OptionParser => $o,
);
PTDEBUG && _d('Nibble time:', $tbl->{nibble_time});
# We're executing REPLACE queries which don't return rows.
# Returning 0 from this callback causes the nibble iter to
# get the next boundaries/nibble.
return 0;
},
after_nibble => sub {
my (%args) = @_;
my $tbl = $args{tbl};
my $nibble_iter = $args{NibbleIterator};
return unless $o->get('execute');
# Update rate, chunk size, and progress if the nibble actually
# selected some rows.
my $cnt = $tbl->{row_cnt};
if ( ($cnt || 0) > 0 ) {
# Update the rate of rows per second for the entire server.
# This is used for the initial chunk size of the next table.
$total_rows += $cnt;
$total_time += $tbl->{nibble_time};
$avg_rate = int($total_rows / $total_time);
PTDEBUG && _d('Average copy rate (rows/s):', $avg_rate);
# Adjust chunk size. This affects the next chunk.
if ( $chunk_time ) {
# Calcuate a new chunk-size based on the rate of rows/s.
$tbl->{chunk_size} = $tbl->{rate}->update(
$cnt, # processed this many rows
$tbl->{nibble_time}, # is this amount of time
);
if ( $tbl->{chunk_size} < 1 ) {
# This shouldn't happen. WeightedAvgRate::update() may
# return a value < 1, but minimum chunk size is 1.
$tbl->{chunk_size} = 1;
# This warning is printed once per table.
if ( !$tbl->{warned_slow} ) {
warn "Rows are copying very slowly. "
. "--chunk-size has been automatically reduced to 1. "
. "Check that the server is not being overloaded, "
. "or increase --chunk-time. The last chunk "
. "selected $cnt rows and took "
. sprintf('%.3f', $tbl->{nibble_time})
. " seconds to execute.\n";
$tbl->{warned_slow} = 1;
}
}
# Update chunk-size based on the rate of rows/s.
$nibble_iter->set_chunk_size($tbl->{chunk_size});
}
# Every table should have a Progress obj; update it.
if ( my $tbl_pr = $tbl->{progress} ) {
$tbl_pr->update( sub { return $total_rows } );
}
}
# Wait forever for slaves to catch up.
$replica_lag_pr->start() if $replica_lag_pr;
$replica_lag->wait(Progress => $replica_lag_pr);
# Wait forever for system load to abate. wait() will die if
# --critical load is reached.
$sys_load_pr->start() if $sys_load_pr;
$sys_load->wait(Progress => $sys_load_pr);
return;
},
done => sub {
if ( $o->get('execute') ) {
print "Copied rows OK.\n";
}
},
};
# NibbleIterator combines these two statements and adds
# "FROM $orig_table->{name} WHERE <nibble stuff>".
my $dml = "INSERT LOW_PRIORITY IGNORE INTO $new_tbl->{name} "
. "(" . join(', ', map { $q->quote($_) } @common_cols) . ") "
. "SELECT";
my $select = join(', ', map { $q->quote($_) } @common_cols);
# The chunk size is auto-adjusted, so use --chunk-size as
# the initial value, but then save and update the adjusted
# chunk size in the table data struct.
$orig_tbl->{chunk_size} = $o->get('chunk-size');
# This won't (shouldn't) fail because we already verified in
# check_orig_table() table we can NibbleIterator::can_nibble().
my $nibble_iter = new NibbleIterator(
Cxn => $cxn,
tbl => $orig_tbl,
chunk_size => $orig_tbl->{chunk_size},
chunk_index => $o->get('chunk-index'),
dml => $dml,
select => $select,
callbacks => $callbacks,
OptionParser => $o,
Quoter => $q,
TableParser => $tp,
TableNibbler => new TableNibbler(TableParser => $tp, Quoter => $q),
comments => {
bite => "pt-online-schema-change $PID copy table",
nibble => "pt-online-schema-change $PID copy nibble",
},
);
# Init a new weighted avg rate calculator for the table.
$orig_tbl->{rate} = new WeightedAvgRate(target_t => $chunk_time);
# Make a Progress obj for this table. It may not be used;
# depends on how many rows, chunk size, how fast the server
# is, etc. But just in case, all tables have a Progress obj.
if ( $o->get('progress')
&& !$nibble_iter->one_nibble()
&& $nibble_iter->row_estimate() )
{
$orig_tbl->{progress} = new Progress(
jobsize => $nibble_iter->row_estimate(),
spec => $o->get('progress'),
name => "Copying $orig_tbl->{name}",
);
}
# Start copying rows. This may take awhile, but --progress is on
# by default so there will be progress updates to stderr.
eval {
1 while $nibble_iter->next();
};
if ( $EVAL_ERROR ) {
die "Error copying rows from $orig_tbl->{name} to "
. "$new_tbl->{name}: $EVAL_ERROR\n";
}
$orig_tbl->{copied} = 1; # flag for cleanup tasks
# XXX Auto-choose the alter fk method BEFORE swapping/renaming tables
# else everything will break because if drop_swap is chosen, then we
# most NOT rename tables or drop the old table.
if ( $alter_fk_method eq 'auto' ) {
# If chunk time is set, then use the average rate of rows/s
# from copying the orig table to determine the max size of
# a child table that can be altered within one chunk time.
# The limit is a fudge factor. Chunk time won't be set if
# the user specified --chunk-size=N on the cmd line, in which
# case the max child table size is their specified chunk size
# times the fudge factor.
my $max_rows
= $o->get('dry-run') ? $o->get('chunk-size') * $limit
: $chunk_time ? $avg_rate * $chunk_time * $limit
: $o->get('chunk-size') * $limit;
PTDEBUG && _d('Max allowed child table size:', $max_rows);
$alter_fk_method = determine_alter_fk_method(
child_tables => $child_tables,
max_rows => $max_rows,
Cxn => $cxn,
OptionParser => $o,
);
if ( $alter_fk_method eq 'drop_swap' ) {
$o->set('swap-tables', 0);
$o->set('drop-old-table', 0);
}
}
# #####################################################################
# XXX
# Step 5: Rename tables: orig -> old, new -> orig
# Past this step, the original table has been altered. This shouldn't
# fail, but if it does, the failure could be serious depending on what
# state the tables are left in.
# XXX
# #####################################################################
my $old_tbl;
if ( $o->get('swap-tables') ) {
eval {
$old_tbl = swap_tables(
orig_tbl => $orig_tbl,
new_tbl => $new_tbl,
suffix => '_old',
Cxn => $cxn,
Quoter => $q,
OptionParser => $o,
);
};
if ( $EVAL_ERROR ) {
die "Error swapping the tables: $EVAL_ERROR\n"
. "Verify that the original table $orig_tbl->{name} has not "
. "been modified or renamed to the old table $old_tbl->{name}. "
. "Then drop the new table $new_tbl->{name} if it exists.\n";
}
}
$orig_tbl->{swapped} = 1; # flag for cleanup tasks
PTDEBUG && _d('Old table:', Dumper($old_tbl));
# #####################################################################
# Step 6: Update foreign key constraints if there are child tables.
# #####################################################################
if ( $child_tables ) {
eval {
if ( $alter_fk_method eq 'none' ) {
# This shouldn't happen, but in case it does we should know.
warn "The tool detected child tables but "
. "--alter-foreign-keys-method=none";
}
elsif ( $alter_fk_method eq 'rebuild_constraints' ) {
rebuild_constraints(
orig_tbl => $orig_tbl,
old_tbl => $old_tbl,
child_tables => $child_tables,
OptionParser => $o,
Quoter => $q,
Cxn => $cxn,
TableParser => $tp,
);
}
elsif ( $alter_fk_method eq 'drop_swap' ) {
drop_swap(
orig_tbl => $orig_tbl,
new_tbl => $new_tbl,
Cxn => $cxn,
OptionParser => $o,
);
}
elsif ( !$alter_fk_method
&& $o->has('alter-foreign-keys-method')
&& ($o->get('alter-foreign-keys-method') || '') eq 'auto' ) {
# If --alter-foreign-keys-method is 'auto' and we are on a dry run,
# $alter_fk_method is left as an empty string.
print "Not updating foreign key constraints because this is a dry run.\n";
}
else {
# This should "never" happen because we check this var earlier.
die "Invalid --alter-foreign-keys-method: $alter_fk_method\n";
}
};
if ( $EVAL_ERROR ) {
# TODO: improve error message and handling.
die "Error updating foreign key constraints: $EVAL_ERROR\n";
}
}
# ########################################################################
# Step 7: Drop the old table.
# ########################################################################
if ( $o->get('drop-old-table') ) {
if ( $o->get('dry-run') ) {
print "Not dropping old table because this is a dry run.\n";
}
elsif ( !$old_tbl ) {
print "Not dropping old table because --no-swap-tables was specified.\n";
}
else {
print "Dropping old table...\n";
if ( $alter_fk_method eq 'none' ) {
# Child tables still reference the old table, but the user
# has chosen to break fks, so we need to disable fk checks
# in order to drop the old table.
my $sql = "SET foreign_key_checks=0";
PTDEBUG && _d($sql);
print $sql, "\n" if $o->get('print');
$cxn->dbh()->do($sql);
}
my $sql = "DROP TABLE IF EXISTS $old_tbl->{name}";
print $sql, "\n" if $o->get('print');
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
die "Error dropping the old table: $EVAL_ERROR\n";
}
print "Dropped old table $old_tbl->{name} OK.\n";
}
}
# ########################################################################
# Done.
# ########################################################################
$orig_tbl->{success} = 1; # flag for cleanup tasks
$cleanup = undef; # exec cleanup tasks
return $exit_status;
}
# ############################################################################
# Subroutines.
# ############################################################################
sub create_new_table{
my (%args) = @_;
my @required_args = qw(orig_tbl Cxn Quoter OptionParser TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $cxn, $q, $o, $tp) = @args{@required_args};
# Get the original table struct.
my $ddl = $tp->get_create_table(
$cxn->dbh(),
$orig_tbl->{db},
$orig_tbl->{tbl},
);
my $tries = $args{tries} || 10; # don't try forever
my $prefix = $args{prefix} || '_';
my $suffix = $args{suffix} || '_new';
my $table_name = $orig_tbl->{tbl} . $suffix;
print "Creating new table...\n";
my $tryno = 1;
my @old_tables;
while ( $tryno++ < $tries ) {
$table_name = $prefix . $table_name;
my $quoted = $q->quote($orig_tbl->{db}, $table_name);
# Generate SQL to create the new table. We do not use CREATE TABLE LIKE
# because it doesn't preserve foreign key constraints. Here we need to
# rename the FK constraints, too. This is because FK constraints are
# internally stored as <database>.<constraint> and there cannot be
# duplicates. If we don't rename the constraints, then InnoDB will throw
# error 121 (duplicate key violation) when we try to execute the CREATE
# TABLE. TODO: this code isn't perfect. If we rename a constraint from
# foo to _foo and there is already a constraint with that name in this
# or another table, we can still have a collision. But if there are
# multiple FKs on this table, it's hard to know which one is causing the
# trouble. Should we generate random/UUID FK names or something instead?
my $sql = $ddl;
$sql =~ s/\ACREATE TABLE .*?\($/CREATE TABLE $quoted (/m;
$sql =~ s/^ CONSTRAINT `/ CONSTRAINT `_/gm;
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
# Ignore this error because if multiple instances of the tool
# are running, or previous runs failed and weren't cleaned up,
# then there will be other similarly named tables with fewer
# leading prefix chars. Or, in rarer cases, the db just happens
# to have a similarly named table created by the user for other
# purposes.
if ( $EVAL_ERROR =~ m/table.+?already exists/i ) {
push @old_tables, $q->quote($orig_tbl->{db}, $table_name);
next;
}
# Some other error happened. Let the caller catch it.
die $EVAL_ERROR;
}
print $sql, "\n" if $o->get('print'); # the sql that work
print "Created new table $orig_tbl->{db}.$table_name OK.\n";
return { # success
db => $orig_tbl->{db},
tbl => $table_name,
name => $q->quote($orig_tbl->{db}, $table_name),
};
}
die "Failed to find a unique new table name after $tries attemps. "
. "The following tables exist which may be left over from previous "
. "failed runs of the tool:\n"
. join("\n", map { " $_" } @old_tables)
. "\nExamine these tables and drop some or all of them if they are "
. "no longer need, then re-run the tool.\n";
}
sub swap_tables {
my (%args) = @_;
my @required_args = qw(orig_tbl new_tbl Cxn Quoter OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $new_tbl, $cxn, $q, $o) = @args{@required_args};
my $prefix = '_';
my $table_name = $orig_tbl->{tbl} . ($args{suffix} || '');
my $tries = 10; # don't try forever
# This sub only works for --execute. Since the options are
# mutually exclusive and we return in the if case, the elsif
# is just a paranoid check because swapping the tables is one
# of the most sensitive/dangerous operations.
if ( $o->get('dry-run') ) {
print "Not swapping tables because this is a dry run.\n";
# A return value really isn't needed, but this trick allows
# rebuild_constraints() to parse and show the sql statements
# it would used. Otherwise, this has no effect.
return $orig_tbl;
}
elsif ( $o->get('execute') ) {
print "Swapping tables...\n";
while ( $tries-- ) {
$table_name = $prefix . $table_name;
my $sql = "RENAME TABLE $orig_tbl->{name} "
. "TO " . $q->quote($orig_tbl->{db}, $table_name)
. ", $new_tbl->{name} TO $orig_tbl->{name}";
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
# Ignore this error because if multiple instances of the tool
# are running, or previous runs failed and weren't cleaned up,
# then there will be other similarly named tables with fewer
# leading prefix chars. Or, in rarer cases, the db just happens
# to have a similarly named table created by the user for other
# purposes.
next if $EVAL_ERROR =~ m/table.+?already exists/i;
# Some other error happened. Let caller catch it.
die $EVAL_ERROR;
}
print $sql, "\n" if $o->get('print');
print "Swapped original and new tables OK.\n";
return { # success
db => $orig_tbl->{db},
tbl => $table_name,
name => $q->quote($orig_tbl->{db}, $table_name),
};
}
# This shouldn't happen.
# Here and in the attempt to find a new table name we probably ought to
# use --retries (and maybe a Retry object?)
die "Failed to find a unique old table name after serveral attempts.\n";
}
}
sub check_orig_table {
my ( %args ) = @_;
my @required_args = qw(orig_tbl Cxn TableParser OptionParser Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $cxn, $tp, $o, $q) = @args{@required_args};
my $dbh = $cxn->dbh();
# The original table must exist, of course.
if (!$tp->check_table(dbh=>$dbh,db=>$orig_tbl->{db},tbl=>$orig_tbl->{tbl})) {
die "The original table $orig_tbl->{name} does not exist.\n";
}
# There cannot be any triggers on the original table.
my $sql = 'SHOW TRIGGERS FROM ' . $q->quote($orig_tbl->{db})
. ' LIKE ' . $q->literal_like($orig_tbl->{tbl});
PTDEBUG && _d($sql);
my $triggers = $dbh->selectall_arrayref($sql);
if ( $triggers && @$triggers ) {
die "The table $orig_tbl->{name} has triggers. This tool "
. "needs to create its own triggers, so the table cannot "
. "already have triggers.\n";
}
# Get the table struct. NibbleIterator needs this, and so do we.
my $ddl = $tp->get_create_table(
$cxn->dbh(),
$orig_tbl->{db},
$orig_tbl->{tbl},
);
$orig_tbl->{tbl_struct} = $tp->parse($ddl);
# Must be able to nibble the original table (to copy rows to the new table).
eval {
NibbleIterator::can_nibble(
Cxn => $cxn,
tbl => $orig_tbl,
chunk_size => $o->get('chunk-size'),
chunk_indx => $o->get('chunk-index'),
OptionParser => $o,
TableParser => $tp,
);
};
if ( $EVAL_ERROR ) {
die "Cannot chunk the original table $orig_tbl->{name}: $EVAL_ERROR\n";
}
# Find a pk or unique index to use for the delete trigger. can_nibble()
# above returns an index, but NibbleIterator will use non-unique indexes,
# so we have to do this again here.
my $indexes = $orig_tbl->{tbl_struct}->{keys}; # brevity
foreach my $index ( $tp->sort_indexes($orig_tbl->{tbl_struct}) ) {
if ( $index eq 'PRIMARY' || $indexes->{$index}->{is_unique} ) {
PTDEBUG && _d('Delete trigger index:', Dumper($index));
$orig_tbl->{del_index} = $index;
last;
}
}
if ( !$orig_tbl->{del_index} ) {
die "The original table $orig_tbl->{name} does not have a PRIMARY KEY "
. "or a unique index which is required for the DELETE trigger.\n";
}
return; # success
}
sub find_child_tables {
my ( %args ) = @_;
my @required_args = qw(tbl Cxn Quoter);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($tbl, $cxn, $q) = @args{@required_args};
PTDEBUG && _d('Finding child tables');
my $sql = "SELECT table_schema, table_name "
. "FROM information_schema.key_column_usage "
. "WHERE constraint_schema='$tbl->{db}' "
. "AND referenced_table_name='$tbl->{tbl}'";
PTDEBUG && _d($sql);
my $rows = $cxn->dbh()->selectall_arrayref($sql);
if ( !$rows || !@$rows ) {
PTDEBUG && _d('No child tables found');
return;
}
my @child_tables;
foreach my $row ( @$rows ) {
my $tbl = {
db => $row->[0],
tbl => $row->[1],
name => $q->quote(@$row),
};
# Get row estimates for each child table so we can give the user
# some input on choosing an --alter-foreign-keys-method if they
# don't use "auto".
my ($n_rows) = NibbleIterator::get_row_estimate(
Cxn => $cxn,
tbl => $tbl,
);
$tbl->{row_est} = $n_rows;
push @child_tables, $tbl;
}
PTDEBUG && _d('Child tables:', Dumper(\@child_tables));
return \@child_tables;
}
sub determine_alter_fk_method {
my ( %args ) = @_;
my @required_args = qw(child_tables max_rows Cxn OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($child_tables, $max_rows, $cxn, $o) = @args{@required_args};
if ( $o->get('dry-run') ) {
print "Not determining the method to update foreign keys "
. "because this is a dry run.\n";
return ''; # $alter_fk_method can't be undef
}
# The rebuild_constraints method is the default becuase it's safer
# and doesn't cause the orig table to go missing for a moment.
my $method = 'rebuild_constraints';
print "Max rows for the rebuild_constraints method: $max_rows\n",
"Determining the method to update foreign keys...\n";
foreach my $child_tbl ( @$child_tables ) {
print " $child_tbl->{name}: ";
my ($n_rows) = NibbleIterator::get_row_estimate(
Cxn => $cxn,
tbl => $child_tbl,
);
if ( $n_rows > $max_rows ) {
print "too many rows: $n_rows; must use drop_swap\n";
$method = 'drop_swap';
last;
}
else {
print "$n_rows rows; can use rebuild_constraints\n";
}
}
return $method || ''; # $alter_fk_method can't be undef
}
sub rebuild_constraints {
my ( %args ) = @_;
my @required_args = qw(orig_tbl old_tbl child_tables
Cxn Quoter OptionParser TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $old_tbl, $child_tables, $cxn, $q, $o, $tp)
= @args{@required_args};
# MySQL has a "feature" where if the parent tbl is in the same db,
# then the child tbl ref is simply `parent_tbl`, but if the parent tbl
# is in another db, then the child tbl ref is `other_db`.`parent_tbl`.
# When we recreate the ref below, we use the db-qualified form, and
# MySQL will automatically trim the db if the tables are in the same db.
my $quoted_old_table = $q->quote($old_tbl->{tbl});
my $constraint = qr/
^\s+
(
CONSTRAINT.+?
REFERENCES\s(?:$quoted_old_table|$old_tbl->{name})
.+
)$
/xm;
PTDEBUG && _d('Rebuilding fk constraint matching', $constraint);
if ( $o->get('dry-run') ) {
print "Not rebuilding foreign key constraints because this is a dry run.\n";
}
else {
print "Rebuilding foreign key constraints...\n";
}
CHILD_TABLE:
foreach my $child_tbl ( @$child_tables ) {
my $table_def = $tp->get_create_table(
$cxn->dbh(),
$child_tbl->{db},
$child_tbl->{tbl},
);
my @constraints = $table_def =~ m/$constraint/g;
if ( !@constraints ) {
warn "$child_tbl->{name} has no foreign key "
. "constraints referencing $old_tbl->{name}.\n";
next CHILD_TABLE;
}
my @rebuilt_constraints;
foreach my $constraint ( @constraints ) {
PTDEBUG && _d('Rebuilding fk constraint:', $constraint);
# Remove trailing commas in case there are multiple constraints on the
# table.
$constraint =~ s/,$//;
# Find the constraint name. It will be quoted already.
my ($fk) = $constraint =~ m/CONSTRAINT\s+`([^`]+)`/;
# Drop the reference to the old table/renamed orig table, and add a new
# reference to the new table. InnoDB will throw an error if the new
# constraint has the same name as the old one, so we must rename it.
# Example: after renaming sakila.actor to sakila.actor_old (for
# example), the foreign key on film_actor looks like this:
# CONSTRAINT `fk_film_actor_actor` FOREIGN KEY (`actor_id`) REFERENCES
# `actor_old` (`actor_id`) ON UPDATE CASCADE
# We need it to look like this instead:
# CONSTRAINT `_fk_film_actor_actor` FOREIGN KEY (`actor_id`) REFERENCES
# `actor` (`actor_id`) ON UPDATE CASCADE
# Reference the correct table name...
$constraint =~ s/REFERENCES[^\(]+/REFERENCES $orig_tbl->{name} /;
# And rename the constraint to avoid conflict
$constraint =~ s/CONSTRAINT `$fk`/CONSTRAINT `_$fk`/;
my $sql = "DROP FOREIGN KEY `$fk`, "
. "ADD $constraint";
push @rebuilt_constraints, $sql;
}
my $sql = "ALTER TABLE $child_tbl->{name} "
. join(', ', @rebuilt_constraints);
print $sql, "\n" if $o->get('print');
if ( $o->get('execute') ) {
PTDEBUG && _d($sql);
$cxn->dbh()->do($sql);
}
}
if ( $o->get('execute') ) {
print "Rebuilt foreign key constraints OK.\n";
}
return;
}
sub drop_swap {
my ( %args ) = @_;
my @required_args = qw(orig_tbl new_tbl Cxn OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $new_tbl, $cxn, $o) = @args{@required_args};
if ( $o->get('dry-run') ) {
print "Not drop-swapping tables because this is a dry run.\n";
}
else {
print "Drop-swapping tables...\n";
}
my @sqls = (
"SET foreign_key_checks=0",
"DROP TABLE IF EXISTS $orig_tbl->{name}",
"RENAME TABLE $new_tbl->{name} TO $orig_tbl->{name}",
);
foreach my $sql ( @sqls ) {
PTDEBUG && _d($sql);
print $sql, "\n" if $o->get('print');
$cxn->dbh()->do($sql) if $o->get('execute');
}
if ( $o->get('execute') ) {
print "Dropped and swapped tables OK.\n";
}
return;
}
sub create_triggers {
my ( %args ) = @_;
my @required_args = qw(orig_tbl new_tbl columns Cxn Quoter OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($orig_tbl, $new_tbl, $cols, $cxn, $q, $o) = @args{@required_args};
# This sub works for --dry-run and --execute. With --dry-run it's
# only interesting if --print is specified, too; then the user can
# see the create triggers statements for --execute.
if ( $o->get('dry-run') ) {
print "Not creating triggers because this is a dry run.\n";
}
else {
print "Creating triggers...\n";
}
# Create a unique trigger name prefix based on the orig table name
# so multiple instances of the tool can run on different tables.
my $prefix = 'pt_osc_' . $orig_tbl->{db} . '_' . $orig_tbl->{tbl};
$prefix =~ s/\W/_/g;
# To be safe, the delete trigger must specify all the columns of the
# primary key/unique index. We use null-safe equals, because unique
# unique indexes can be nullable.
my $tbl_struct = $orig_tbl->{tbl_struct};
my $del_index = $orig_tbl->{del_index};
my $del_index_cols = join(" AND ",
map {
my $col = $q->quote($_);
"$new_tbl->{name}.$col <=> OLD.$col"
} @{$tbl_struct->{keys}->{$del_index}->{cols}} );
my $delete_trigger
= "CREATE TRIGGER `${prefix}_del` AFTER DELETE ON $orig_tbl->{name} "
. "FOR EACH ROW "
. "DELETE IGNORE FROM $new_tbl->{name} "
. "WHERE $del_index_cols";
# The insert and update triggers should only use values for columns
# that exist in both tables.
my $qcols = join(', ', map { $q->quote($_) } @$cols);
my $new_vals = join(', ', map { "NEW.".$q->quote($_) } @$cols);
my $insert_trigger
= "CREATE TRIGGER `${prefix}_ins` AFTER INSERT ON $orig_tbl->{name} "
. "FOR EACH ROW "
. "REPLACE INTO $new_tbl->{name} ($qcols) VALUES ($new_vals)";
my $update_trigger
= "CREATE TRIGGER `${prefix}_upd` AFTER UPDATE ON $orig_tbl->{name} "
. "FOR EACH ROW "
. "REPLACE INTO $new_tbl->{name} ($qcols) VALUES ($new_vals)";
my @triggers = (
['del', $delete_trigger],
['upd', $update_trigger],
['ins', $insert_trigger],
);
@drop_trigger_sqls = ();
foreach my $trg ( @triggers ) {
my ($name, $sql) = @$trg;
print $sql, "\n" if $o->get('print');
if ( $o->get('execute') ) {
# Let caller catch errors.
PTDEBUG && _d($sql);
$cxn->dbh()->do($sql);
}
# Only save the trigger once it has been created
# (or faked to be created) so if the 2nd trigger
# fails to create, we know to only drop the 1st.
push @drop_trigger_sqls,
"DROP TRIGGER IF EXISTS "
. $q->quote($orig_tbl->{db}, "${prefix}_$name") . ";";
}
if ( $o->get('execute') ) {
print "Created triggers OK.\n";
}
return;
}
sub drop_triggers {
my ( %args ) = @_;
my @required_args = qw(tbl Cxn Quoter OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($tbl, $cxn, $q, $o) = @args{@required_args};
# This sub works for --dry-run and --execute, although --dry-run is
# only interesting with --print so the user can see the drop trigger
# statements for --execute.
if ( $o->get('dry-run') ) {
print "Not dropping triggers because this is a dry run.\n";
}
else {
print "Dropping triggers...\n";
}
my @not_dropped;
foreach my $sql ( @drop_trigger_sqls ) {
print $sql, "\n" if $o->get('print');
if ( $o->get('execute') ) {
PTDEBUG && _d($sql);
eval {
$cxn->dbh()->do($sql);
};
if ( $EVAL_ERROR ) {
warn "Error dropping trigger: $EVAL_ERROR\n";
push @not_dropped, $sql;
$exit_status = 1;
}
}
}
if ( $o->get('execute') ) {
if ( !@not_dropped ) {
print "Dropped triggers OK.\n";
}
else {
warn "To try dropping the triggers again, execute:\n"
. join("\n", @not_dropped) . "\n";
}
}
return;
}
sub exec_nibble {
my (%args) = @_;
my @required_args = qw(Cxn tbl NibbleIterator Retry Quoter OptionParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $nibble_iter, $retry, $q, $o)= @args{@required_args};
my $dbh = $cxn->dbh();
my $sth = $nibble_iter->statements();
my $boundary = $nibble_iter->boundaries();
my $lb_quoted = $q->serialize_list(@{$boundary->{lower}});
my $ub_quoted = $q->serialize_list(@{$boundary->{upper}});
my $chunk = $nibble_iter->nibble_number();
my $chunk_index = $nibble_iter->nibble_index();
# Completely ignore these error codes.
my %ignore_code = (
# Error: 1592 SQLSTATE: HY000 (ER_BINLOG_UNSAFE_STATEMENT)
# Message: Statement may not be safe to log in statement format.
# Ignore this warning because we have purposely set statement-based
# replication.
1592 => 1,
);
# Warn once per-table for these error codes if the error message
# matches the pattern.
my %warn_code = (
# Error: 1265 SQLSTATE: 01000 (WARN_DATA_TRUNCATED)
# Message: Data truncated for column '%s' at row %ld
1265 => {
# any pattern
# use MySQL's message for this warning
},
);
return $retry->retry(
tries => $o->get('retries'),
wait => sub { sleep 0.25; return; },
try => sub {
# ###################################################################
# Start timing the query.
# ###################################################################
my $t_start = time;
# Execute the INSERT..SELECT query.
PTDEBUG && _d($sth->{nibble}->{Statement},
'lower boundary:', @{$boundary->{lower}},
'upper boundary:', @{$boundary->{upper}});
$sth->{nibble}->execute(
# WHERE
@{$boundary->{lower}}, # upper boundary values
@{$boundary->{upper}}, # lower boundary values
);
my $t_end = time;
# ###################################################################
# End timing the query.
# ###################################################################
# How many rows were inserted this time. Used for auto chunk sizing.
$tbl->{row_cnt} = $sth->{nibble}->rows();
# Check if query caused any warnings.
my $sql_warn = 'SHOW WARNINGS';
PTDEBUG && _d($sql_warn);
my $warnings = $dbh->selectall_arrayref($sql_warn, { Slice => {} } );
foreach my $warning ( @$warnings ) {
my $code = ($warning->{code} || 0);
my $message = $warning->{message};
if ( $ignore_code{$code} ) {
PTDEBUG && _d('Ignoring warning:', $code, $message);
next;
}
elsif ( $warn_code{$code}
&& (!$warn_code{$code}->{pattern}
|| $message =~ m/$warn_code{$code}->{pattern}/) )
{
if ( !$tbl->{"warned_code_$code"} ) { # warn once per table
warn "Copying rows caused a MySQL error $code: "
. ($warn_code{$code}->{message}
? $warn_code{$code}->{message}
: $message)
. "\nThis MySQL error is being ignored and further "
. "occurrences of it will not be reported.\n";
$tbl->{"warned_code_$code"} = 1;
}
}
else {
# This die will propagate to fail which will return 0
# and propagate it to final_fail which will die with
# this error message.
die "Copying rows caused a MySQL error $code:\n"
. " Level: " . ($warning->{level} || '') . "\n"
. " Code: " . ($warning->{code} || '') . "\n"
. " Message: " . ($warning->{message} || '') . "\n"
. " Query: " . $sth->{nibble}->{Statement} . "\n";
}
}
# Success: no warnings, no errors. Return nibble time.
return $t_end - $t_start;
},
fail => sub {
my (%args) = @_;
my $error = $args{error};
# The query failed/caused an error. If the error is one of these,
# then we can possibly retry.
if ( $error =~ m/Lock wait timeout exceeded/
|| $error =~ m/Deadlock found/
|| $error =~ m/Query execution was interrupted/
) {
# These errors/warnings can be retried, so don't print
# a warning yet; do that in final_fail.
return 1; # try again
}
elsif ( $error =~ m/MySQL server has gone away/
|| $error =~ m/Lost connection to MySQL server/
) {
# The 2nd pattern means that MySQL itself died or was stopped.
# The 3rd pattern means that our cxn was killed (KILL <id>).
$dbh = $cxn->connect(); # connect or die trying
return 1; # reconnected, try again
}
# At this point, either the error/warning cannot be retried,
# or we failed to reconnect. Don't retry; call final_fail.
return 0;
},
final_fail => sub {
my (%args) = @_;
# This die should be caught by the caller. Copying rows and
# the tool will stop, which is probably good because by this
# point the error or warning indicates that something is wrong.
die $args{error};
}
);
}
# Sub: explain_statement
# EXPLAIN a statement.
#
# Required Arguments:
# * tbl - Standard tbl hashref
# * sth - Sth with EXLAIN <statement>
# * vals - Values for sth, if any
#
# Returns:
# Hashref with EXPLAIN plan
sub explain_statement {
my ( %args ) = @_;
my @required_args = qw(tbl sth vals);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my ($tbl, $sth, $vals) = @args{@required_args};
my $expl;
eval {
PTDEBUG && _d($sth->{Statement}, 'params:', @$vals);
$sth->execute(@$vals);
$expl = $sth->fetchrow_hashref();
$sth->finish();
};
if ( $EVAL_ERROR ) {
# This shouldn't happen.
die "Error executing " . $sth->{Statement} . ": $EVAL_ERROR\n";
}
PTDEBUG && _d('EXPLAIN plan:', Dumper($expl));
return $expl;
}
# Catches signals so we can exit gracefully.
sub sig_int {
my ( $signal ) = @_;
$oktorun = 0; # flag for cleanup tasks
print STDERR "# Exiting on SIG$signal.\n";
exit 1;
}
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";
}
# ############################################################################
# Run the program.
# ############################################################################
if ( !caller ) { exit main(@ARGV); }
1; # Because this is a module as well as a script.
# ############################################################################
# Documentation
# ############################################################################
=pod
=head1 NAME
pt-online-schema-change - ALTER tables without locking them.
=head1 SYNOPSIS
Usage: pt-online-schema-change [OPTIONS] DSN
pt-online-schema-change alters a table's structure without blocking reads or
writes. Specify the database and table in the DSN. Do not use this tool before
reading its documentation and checking your backups carefully.
Add a column to sakila.actor:
pt-online-schema-change --alter "ADD COLUMN c1 INT" D=sakila,t=actor
Change sakila.actor to InnoDB, effectively performing OPTIMIZE TABLE in a
non-blocking fashion because it is already an InnoDB table:
pt-online-schema-change --alter "ENGINE=InnoDB" D=sakila,t=actor
=head1 RISKS
The following section is included to inform users about the potential risks,
whether known or unknown, of using this tool. The two main categories of risks
are those created by the nature of the tool (e.g. read-only tools vs. read-write
tools) and those created by bugs.
pt-online-schema-change modifies data and structures. You should be careful with
it, and test it before using it in production. You should also ensure that you
have recoverable backups before using this tool.
At the time of this release, we know of no bugs that could cause harm to users.
The authoritative source for updated information is always the online issue
tracking system. Issues that affect this tool will be marked as such. You can
see a list of such issues at the following URL:
L<http://www.percona.com/bugs/pt-online-schema-change>.
See also L<"BUGS"> for more information on filing bugs and getting help.
=head1 DESCRIPTION
pt-online-schema-change emulates the way that MySQL alters tables internally,
but it works on a copy of the table you wish to alter. This means that the
original table is not locked, and clients may continue to read and change data
in it.
pt-online-schema-change works by creating an empty copy of the table to alter,
modifying it as desired, and then copying rows from the original table into the
new table. When the copy is complete, it moves away the original table and
replaces it with the new one. By default, it also drops the original table.
The data copy process is performed in small chunks of data, which are varied to
attempt to make them execute in a specific amount of time (see
L<"--chunk-time">). This process is very similar to how other tools, such as
pt-table-checksum, work. Any modifications to data in the original tables
during the copy will be reflected in the new table, because the tool creates
triggers on the original table to update the corresponding rows in the new
table. The use of triggers means that the tool will not work if any triggers
are already defined on the table.
When the tool finishes copying data into the new table, it uses an atomic
C<RENAME TABLE> operation to simultaneously rename the original and new tables.
After this is complete, the tool drops the original table.
Foreign keys complicate the tool's operation and introduce additional risk. The
technique of atomically renaming the original and new tables does not work when
foreign keys refer to the table. The tool must update foreign keys to refer to
the new table after the schema change is complete. The tool supports two methods
for accomplishing this. You can read more about this in the documentation for
L<"--alter-foreign-keys-method">.
Foreign keys also cause some side effects. The final table will have the same
foreign keys and indexes as the original table (unless you specify differently
in your ALTER statement), but the names of the objects may be changed slightly
to avoid object name collisions in MySQL and InnoDB.
For safety, the tool does not modify the table unless you specify the
L<"--execute"> option, which is not enabled by default. The tool supports a
variety of other measures to prevent unwanted load or other problems, including
automatically detecting replicas, connecting to them, and using the following
safety checks:
=over
=item *
The tool refuses to operate if it detects replication filters. See
L<"--[no]check-replication-filters"> for details.
=item *
The tool pauses the data copy operation if it observes any replicas that are
delayed in replication. See L<"--max-lag"> for details.
=item *
The tool pauses or aborts its operation if it detects too much load on the
server. See L<"--max-load"> and L<"--critical-load"> for details.
=item *
The tool sets its lock wait timeout to 1 second so that it is more likely to be
the victim of any lock contention, and less likely to disrupt other
transactions. See L<"--lock-wait-timeout"> for details.
=item *
The tool refuses to alter the table if foreign key constraints reference it,
unless you specify L<"--alter-foreign-keys-method">.
=back
=head1 OUTPUT
The tool prints information about its activities to STDOUT so that you can see
what it is doing. During the data copy phase, it prints progress reports to
STDERR. You can get additional information with the L<"--print"> option.
=head1 OPTIONS
L<"--dry-run"> and L<"--execute"> are mutually exclusive.
This tool accepts additional command-line arguments. Refer to the
L<"SYNOPSIS"> and usage information for details.
=over
=item --alter
type: string
The schema modification, without the ALTER TABLE keywords. You can perform
multiple modifications to the table by specifying them with commas. Please refer
to the MySQL manual for the syntax of ALTER TABLE.
The following limitations apply which, if attempted, will cause the tool
to fail in unpredictable ways:
=over
=item *
The C<RENAME> clause cannot be used to rename the table.
=item *
Columns cannot be renamed by dropping and re-adding with the new name.
The tool will not copy the original column's data to the new column.
=back
=item --alter-foreign-keys-method
type: string
How to modify foreign keys so they reference the new table. Foreign keys that
reference the table to be altered must be treated specially to ensure that they
continue to reference the correct table. When the tool renames the original
table to let the new one take its place, the foreign keys "follow" the renamed
table, and must be changed to reference the new table instead.
The tool supports two techniques to achieve this. It automatically finds "child
tables" that reference the table to be altered.
=over
=item auto
Automatically determine which method is best. The tool uses
C<rebuild_constraints> if possible (see the description of that method for
details), and if not, then it uses C<drop_swap>.
=item rebuild_constraints
This method uses C<ALTER TABLE> to drop and re-add foreign key constraints that
reference the new table. This is the preferred technique, unless one or more of
the "child" tables is so large that the C<ALTER> would take too long. The tool
determines that by comparing the number of rows in the child table to the rate
at which the tool is able to copy rows from the old table to the new table. If
the tool estimates that the child table can be altered in less time than the
L<"--chunk-time">, then it will use this technique. For purposes of estimating
the time required to alter the child table, the tool multiplies the row-copying
rate by L<"--chunk-size-limit">, because MySQL's C<ALTER TABLE> is typically
much faster than the external process of copying rows.
Due to a limitation in MySQL, foreign keys will not have the same names after
the ALTER that they did prior to it. The tool has to rename the foreign key
when it redefines it, which adds a leading underscore to the name. In some
cases, MySQL also automatically renames indexes required for the foreign key.
=item drop_swap
Disable foreign key checks (FOREIGN_KEY_CHECKS=0), then drop the original table
before renaming the new table into its place. This is different from the normal
method of swapping the old and new table, which uses an atomic C<RENAME> that is
undetectable to client applications.
This method is faster and does not block, but it is riskier for two reasons.
First, for a short time between dropping the original table and renaming the
temporary table, the table to be altered simply does not exist, and queries
against it will result in an error. Secondly, if there is an error and the new
table cannot be renamed into the place of the old one, then it is too late to
abort, because the old table is gone permanently.
This method forces C<--no-swap-tables> and C<--no-drop-old-table>.
=item none
This method is like C<drop_swap> without the "swap". Any foreign keys that
referenced the original table will now reference a nonexistent table. This will
typically cause foreign key violations that are visible in C<SHOW ENGINE INNODB
STATUS>, similar to the following:
Trying to add to index `idx_fk_staff_id` tuple:
DATA TUPLE: 2 fields;
0: len 1; hex 05; asc ;;
1: len 4; hex 80000001; asc ;;
But the parent table `sakila`.`staff_old`
or its .ibd file does not currently exist!
This is because the original table (in this case, sakila.staff) was renamed to
sakila.staff_old and then dropped. This method of handling foreign key
constraints is provided so that the database administrator can disable the
tool's built-in functionality if desired.
=back
=item --ask-pass
Prompt for a password when connecting to MySQL.
=item --charset
short form: -A; type: string
Default character set. If the value is utf8, sets Perl's binmode on
STDOUT to utf8, passes the mysql_enable_utf8 option to DBD::mysql, and runs SET
NAMES UTF8 after connecting to MySQL. Any other value sets binmode on STDOUT
without the utf8 layer, and runs SET NAMES after connecting to MySQL.
=item --check-interval
type: time; default: 1
Sleep time between checks for L<"--max-lag">.
=item --[no]check-replication-filters
default: yes
Abort if any replication filter is set on any server. The tool looks for
server options that filter replication, such as binlog_ignore_db and
replicate_do_db. If it finds any such filters, it aborts with an error.
If the replicas are configured with any filtering options, you should be careful
not to modify any databases or tables that exist on the master and not the
replicas, because it could cause replication to fail. For more information on
replication rules, see L<http://dev.mysql.com/doc/en/replication-rules.html>.
=item --check-slave-lag
type: string
Pause the data copy until this replica's lag is less than L<"--max-lag">. The
value is a DSN that inherits properties from the the connection options
(L<"--port">, L<"--user">, etc.). This option overrides the normal behavior of
finding and continually monitoring replication lag on ALL connected replicas.
If you don't want to monitor ALL replicas, but you want more than just one
replica to be monitored, then use the DSN option to the L<"--recursion-method">
option instead of this option.
=item --chunk-index
type: string
Prefer this index for chunking tables. By default, the tool chooses the most
appropriate index for chunking. This option lets you specify the index that you
prefer. If the index doesn't exist, then the tool will fall back to its default
behavior of choosing an index. The tool adds the index to the SQL statements in
a C<FORCE INDEX> clause. Be careful when using this option; a poor choice of
index could cause bad performance.
=item --chunk-size
type: size; default: 1000
Number of rows to select for each chunk copied. Allowable suffixes are
k, M, G.
This option can override the default behavior, which is to adjust chunk size
dynamically to try to make chunks run in exactly L<"--chunk-time"> seconds.
When this option isn't set explicitly, its default value is used as a starting
point, but after that, the tool ignores this option's value. If you set this
option explicitly, however, then it disables the dynamic adjustment behavior and
tries to make all chunks exactly the specified number of rows.
There is a subtlety: if the chunk index is not unique, then it's possible that
chunks will be larger than desired. For example, if a table is chunked by an
index that contains 10,000 of a given value, there is no way to write a WHERE
clause that matches only 1,000 of the values, and that chunk will be at least
10,000 rows large. Such a chunk will probably be skipped because of
L<"--chunk-size-limit">.
=item --chunk-size-limit
type: float; default: 4.0
Do not copy chunks this much larger than the desired chunk size.
When a table has no unique indexes, chunk sizes can be inaccurate. This option
specifies a maximum tolerable limit to the inaccuracy. The tool uses <EXPLAIN>
to estimate how many rows are in the chunk. If that estimate exceeds the
desired chunk size times the limit, then the tool skips the chunk.
The minimum value for this option is 1, which means that no chunk can be larger
than L<"--chunk-size">. You probably don't want to specify 1, because rows
reported by EXPLAIN are estimates, which can be different from the real number
of rows in the chunk. You can disable oversized chunk checking by specifying a
value of 0.
The tool also uses this option to determine how to handle foreign keys that
reference the table to be altered. See L<"--alter-foreign-keys-method"> for
details.
=item --chunk-time
type: float; default: 0.5
Adjust the chunk size dynamically so each data-copy query takes this long to
execute. The tool tracks the copy rate (rows per second) and adjusts the chunk
size after each data-copy query, so that the next query takes this amount of
time (in seconds) to execute. It keeps an exponentially decaying moving average
of queries per second, so that if the server's performance changes due to
changes in server load, the tool adapts quickly.
If this option is set to zero, the chunk size doesn't auto-adjust, so query
times will vary, but query chunk sizes will not. Another way to do the same
thing is to specify a value for L<"--chunk-size"> explicitly, instead of leaving
it at the default.
=item --config
type: Array
Read this comma-separated list of config files; if specified, this must be the
first option on the command line.
=item --critical-load
type: Array; default: Threads_running=50
Examine SHOW GLOBAL STATUS after every chunk, and abort if the load is too high.
The option accepts a comma-separated list of MySQL status variables and
thresholds. An optional C<=MAX_VALUE> (or C<:MAX_VALUE>) can follow each
variable. If not given, the tool determines a threshold by examining the
current value at startup and doubling it.
See L<"--max-load"> for further details. These options work similarly, except
that this option will abort the tool's operation instead of pausing it, and the
default value is computed differently if you specify no threshold. The reason
for this option is as a safety check in case the triggers on the original table
add so much load to the server that it causes downtime. There is probably no
single value of Threads_running that is wrong for every server, but a default of
50 seems likely to be unacceptably high for most servers, indicating that the
operation should be canceled immediately.
=item --defaults-file
short form: -F; type: string
Only read mysql options from the given file. You must give an absolute
pathname.
=item --[no]drop-new-table
default: yes
Drop the new table if copying the original table fails.
Specifying C<--no-drop-new-table> and C<--no-swap-tables> leaves the new,
altered copy of the table without modifying the original table. The new
table name is like C<_TBL_new> where C<TBL> is the table name.
L<--no-drop-new-table> does not work with
C<alter-foreign-keys-method drop_swap>.
=item --[no]drop-old-table
default: yes
Drop the original table after renaming it. After the original table has been
successfully renamed to let the new table take its place, and if there are no
errors, the tool drops the original table by default. If there are any errors,
the tool leaves the original table in place.
If C<--no-swap-tables> is specified, then there is no old table to drop.
=item --dry-run
Create and alter the new table, but do not create triggers, copy data, or
replace the original table.
=item --execute
Indicate that you have read the documentation and want to alter the table. You
must specify this option to alter the table. If you do not, then the tool will
only perform some safety checks and exit. This helps ensure that you have read the
documentation and understand how to use this tool. If you have not read the
documentation, then do not specify this option.
=item --help
Show help and exit.
=item --host
short form: -h; type: string
Connect to host.
=item --lock-wait-timeout
type: int; default: 1
Set the session value of C<innodb_lock_wait_timeout>. This option helps guard
against long lock waits if the data-copy queries become slow for some reason.
Setting this option dynamically requires the InnoDB plugin, so this works only
on newer InnoDB and MySQL versions. If the setting's current value is greater
than the specified value, and the tool cannot set the value as desired, then it
prints a warning. If the tool cannot set the value but the current value is less
than or equal to the desired value, there is no error.
=item --max-lag
type: time; default: 1s
Pause the data copy until all replicas' lag is less than this value. After each
data-copy query (each chunk), the tool looks at the replication lag of
all replicas to which it connects, using Seconds_Behind_Master. If any replica
is lagging more than the value of this option, then the tool will sleep
for L<"--check-interval"> seconds, then check all replicas again. If you
specify L<"--check-slave-lag">, then the tool only examines that server for
lag, not all servers. If you want to control exactly which servers the tool
monitors, use the DSN value to L<"--recursion-method">.
The tool waits forever for replicas to stop lagging. If any replica is
stopped, the tool waits forever until the replica is started. The data copy
continues when all replicas are running and not lagging too much.
The tool prints progress reports while waiting. If a replica is stopped, it
prints a progress report immediately, then again at every progress report
interval.
=item --max-load
type: Array; default: Threads_running=25
Examine SHOW GLOBAL STATUS after every chunk, and pause if any status variables
are higher than their thresholds. The option accepts a comma-separated list of
MySQL status variables. An optional C<=MAX_VALUE> (or C<:MAX_VALUE>) can follow
each variable. If not given, the tool determines a threshold by examining the
current value and increasing it by 20%.
For example, if you want the tool to pause when Threads_connected gets too high,
you can specify "Threads_connected", and the tool will check the current value
when it starts working and add 20% to that value. If the current value is 100,
then the tool will pause when Threads_connected exceeds 120, and resume working
when it is below 120 again. If you want to specify an explicit threshold, such
as 110, you can use either "Threads_connected:110" or "Threads_connected=110".
The purpose of this option is to prevent the tool from adding too much load to
the server. If the data-copy queries are intrusive, or if they cause lock waits,
then other queries on the server will tend to block and queue. This will
typically cause Threads_running to increase, and the tool can detect that by
running SHOW GLOBAL STATUS immediately after each query finishes. If you
specify a threshold for this variable, then you can instruct the tool to wait
until queries are running normally again. This will not prevent queueing,
however; it will only give the server a chance to recover from the queueing. If
you notice queueing, it is best to decrease the chunk time.
=item --password
short form: -p; type: string
Password to use when connecting.
=item --pid
type: string
Create the given PID file. The file contains the process ID of the tool's
instance. The PID file is removed when the tool exits. The tool checks for
the existence of the PID file when starting; if it exists and the process with
the matching PID exists, the tool exits.
=item --port
short form: -P; type: int
Port number to use for connection.
=item --print
Print SQL statements to STDOUT. Specifying this option allows you to see most
of the statements that the tool executes. You can use this option with
L<"--dry-run">, for example.
=item --progress
type: array; default: time,30
Print progress reports to STDERR while copying rows. The value is a
comma-separated list with two parts. The first part can be percentage, time, or
iterations; the second part specifies how often an update should be printed, in
percentage, seconds, or number of iterations.
=item --quiet
short form: -q
Do not print messages to STDOUT. Errors and warnings are still printed to
STDERR.
=item --recurse
type: int
Number of levels to recurse in the hierarchy when discovering replicas.
Default is infinite. See also L<"--recursion-method">.
=item --recursion-method
type: string
Preferred recursion method for discovering replicas. Possible methods are:
METHOD USES
=========== ==================
processlist SHOW PROCESSLIST
hosts SHOW SLAVE HOSTS
dsn=DSN DSNs from a table
none Do not find slaves
The processlist method is the default, because SHOW SLAVE HOSTS is not
reliable. However, the hosts method can work better if the server uses a
non-standard port (not 3306). The tool usually does the right thing and
finds all replicas, but you may give a preferred method and it will be used
first.
The hosts method requires replicas to be configured with report_host,
report_port, etc.
The dsn method is special: it specifies a table from which other DSN strings
are read. The specified DSN must specify a D and t, or a database-qualified
t. The DSN table should have the following structure:
CREATE TABLE `dsns` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`parent_id` int(11) DEFAULT NULL,
`dsn` varchar(255) NOT NULL,
PRIMARY KEY (`id`)
);
To make the tool monitor only the hosts 10.10.1.16 and 10.10.1.17 for
replication lag, insert the values C<h=10.10.1.16> and C<h=10.10.1.17> into the
table. Currently, the DSNs are ordered by id, but id and parent_id are otherwise
ignored.
=item --retries
type: int; default: 3
Retry a chunk this many times when there is a nonfatal error. Nonfatal errors
are problems such as a lock wait timeout or the query being killed. This option
applies to the data copy operation.
=item --set-vars
type: string; default: wait_timeout=10000
Set these MySQL variables. Immediately after connecting to MySQL, this string
will be appended to SET and executed.
=item --socket
short form: -S; type: string
Socket file to use for connection.
=item --[no]swap-tables
default: yes
Swap the original table and the new, altered table. This step completes the
online schema change process by making the table with the new schema take the
place of the original table. The original table becomes the "old table," and
the tool drops it unless you disable L<"--[no]drop-old-table">.
=item --user
short form: -u; type: string
User for login if not current user.
=item --version
Show version and exit.
=back
=head1 DSN OPTIONS
These DSN options are used to create a DSN. Each option is given like
C<option=value>. The options are case-sensitive, so P and p are not the
same option. There cannot be whitespace before or after the C<=> and
if the value contains whitespace it must be quoted. DSN options are
comma-separated. See the L<percona-toolkit> manpage for full details.
=over
=item * A
dsn: charset; copy: yes
Default character set.
=item * D
dsn: database; copy: yes
Database for the old and new table.
=item * F
dsn: mysql_read_default_file; copy: yes
Only read default options from the given file
=item * h
dsn: host; copy: yes
Connect to host.
=item * p
dsn: password; copy: yes
Password to use when connecting.
=item * P
dsn: port; copy: yes
Port number to use for connection.
=item * S
dsn: mysql_socket; copy: yes
Socket file to use for connection.
=item * t
dsn: table; copy: no
Table to alter.
=item * u
dsn: user; copy: yes
User for login if not current user.
=back
=head1 ENVIRONMENT
The environment variable C<PTDEBUG> enables verbose debugging output to STDERR.
To enable debugging and capture all output to a file, run the tool like:
PTDEBUG=1 pt-online-schema-change ... > FILE 2>&1
Be careful: debugging output is voluminous and can generate several megabytes
of output.
=head1 SYSTEM REQUIREMENTS
You need Perl, DBI, DBD::mysql, and some core packages that ought to be
installed in any reasonably new version of Perl.
This tool works only on MySQL 5.0.2 and newer versions, because earlier versions
do not support triggers.
=head1 BUGS
For a list of known bugs, see L<http://www.percona.com/bugs/pt-online-schema-change>.
Please report bugs at L<https://bugs.launchpad.net/percona-toolkit>.
Include the following information in your bug report:
=over
=item * Complete command-line used to run the tool
=item * Tool L<"--version">
=item * MySQL version of all servers involved
=item * Output from the tool including STDERR
=item * Input files (log/dump/config files, etc.)
=back
If possible, include debugging output by running the tool with C<PTDEBUG>;
see L<"ENVIRONMENT">.
=head1 DOWNLOADING
Visit L<http://www.percona.com/software/percona-toolkit/> to download the
latest release of Percona Toolkit. Or, get the latest release from the
command line:
wget percona.com/get/percona-toolkit.tar.gz
wget percona.com/get/percona-toolkit.rpm
wget percona.com/get/percona-toolkit.deb
You can also get individual tools from the latest release:
wget percona.com/get/TOOL
Replace C<TOOL> with the name of any tool.
=head1 AUTHORS
Daniel Nichter and Baron Schwartz
=head1 ACKNOWLEDGMENTS
The "online schema change" concept was first implemented by Shlomi Noach
in his tool C<oak-online-alter-table>, part of
L<http://code.google.com/p/openarkkit/>. Engineers at Facebook then built
another version called C<OnlineSchemaChange.php> as explained by their blog
post: L<http://tinyurl.com/32zeb86>. This tool is a hybrid of both approaches,
with additional features and functionality not present in either.
=head1 ABOUT PERCONA TOOLKIT
This tool is part of Percona Toolkit, a collection of advanced command-line
tools developed by Percona for MySQL support and consulting. Percona Toolkit
was forked from two projects in June, 2011: Maatkit and Aspersa. Those
projects were created by Baron Schwartz and developed primarily by him and
Daniel Nichter, both of whom are employed by Percona. Visit
L<http://www.percona.com/software/> for more software developed by Percona.
=head1 COPYRIGHT, LICENSE, AND WARRANTY
This program is copyright 2011-2012 Percona Inc.
Feedback and improvements are welcome.
THIS PROGRAM IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED
WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
This program 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, version 2; OR the Perl Artistic License. On UNIX and similar
systems, you can issue `man perlgpl' or `man perlartistic' to read these
licenses.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc., 59 Temple
Place, Suite 330, Boston, MA 02111-1307 USA.
=head1 VERSION
pt-online-schema-change 2.1.1
=cut