percona/percona-toolkit
1
0
Fork 0
mirror of https://github.com/percona/percona-toolkit.git synced 2025-09-01 18:25:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

PT-2168 pt-osc shouldnt fail while unable to monitor a replica node (#676)

Browse Source 
* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Proof of concept
- Fixed regular expression in lib/TableParser.pm mistakenly chaged in the tool's code

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Added basic test case for PT-2168
- Added more details for replica lag information
- Disconnecting replica if lag is not checked. This prevents "Too many
  connections" error

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Implemented option --wait-lost-replicas for pt-osc, added test case

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Added more tests for situations where connection to the replica can
  fail

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Removed extra checks for wait_no_die variable
- Added test cases for SQL queries that pt-osc sends to replicas

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Allow to reload dsns table while waiting for missed replica if
  --recursion-method is dsn
- Fixed logic in replica rediscovery, so it works with replicas on the
  same host but with different ports
- Renamed option wait-lost-replicas to fail-on-stopped-replication, so
  it is in line with pt-table-checksum
- Adjusted tests
- Removed debug code for PT-1760
- Added test case for PT-1760
- Added exception for variable Open_tables_with_triggers in
  lib/bash/collect.sh due to failed test in Percona Server 8.0.34+
- Updated pt-stalk

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Updated modules
- Fixed typo in t/pt-table-sync/bidirectional.t
- Removed trailing whitespaces in lib/MasterSlave.pm

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Help for option --fail-on-stopped-replication

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Added check for availability of the simple_rewrite_plugin in t/pt-online-schema-change/pt-2168.t

* PT-2168 - PT-OSC shouldn't fail while unable to monitor a replica node

- Added link to the simple_rewrite_plugin source code
- Removed tests for code that runs only in the beginning of pt-osc
  action, so should not be affected by the option fail-on-stopped-replication
This commit is contained in:
Sveta Smirnova
2023-09-19 23:31:33 +03:00
committed by GitHub
parent 36a183e2aa
commit f8c43118bd
19 changed files with 1287 additions and 202 deletions
Download Patch File Download Diff File Expand all files Collapse all files

79
bin/pt-archiver
View File

@@ -3692,6 +3692,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -3699,6 +3700,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -3754,6 +3756,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -3763,23 +3779,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -4401,18 +4423,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

79
bin/pt-heartbeat
View File

@@ -187,6 +187,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -194,6 +195,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -249,6 +251,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -258,23 +274,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -896,18 +918,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

79
bin/pt-kill
View File

@@ -3969,6 +3969,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -3976,6 +3977,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -4031,6 +4033,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -4040,23 +4056,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -4678,18 +4700,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

131
bin/pt-online-schema-change
View File

@@ -4255,6 +4255,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -4262,6 +4263,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -4317,6 +4319,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -4326,23 +4342,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -4964,18 +4986,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}
@@ -5048,9 +5091,9 @@ sub wait {
my ($self) = @_;
my ($slaves, $refresher) = ($self->{slaves}, $self->{get_slaves_cb});
return $slaves if ( not defined $refresher );
my $before = join ' ', sort map {$_->name()} @$slaves;
my $before = join ' ', sort map {$_->description()} @$slaves;
$slaves = $refresher->();
my $after = join ' ', sort map {$_->name()} @$slaves;
my $after = join ' ', sort map {$_->description()} @$slaves;
if ($before ne $after) {
$self->{slaves} = $slaves;
printf STDERR "Slave set to watch has changed\n Was: %s\n Now: %s\n",
@@ -5065,9 +5108,10 @@ sub wait {
$pr_callback = sub {
my ($fraction, $elapsed, $remaining, $eta, $completed) = @_;
my $dsn_name = $worst->{cxn}->name();
my $dsn_description = $worst->{cxn}->description();
if ( defined $worst->{lag} ) {
print STDERR "Replica lag is " . ($worst->{lag} || '?')
. " seconds on $dsn_name. Waiting.\n";
. " seconds on $dsn_description. Waiting.\n";
}
else {
if ($self->{fail_on_stopped_replication}) {
@@ -8702,10 +8746,21 @@ sub main {
OptionParser => $o,
set => $set_on_connect,
);
eval { $cxn->connect() }; # connect or die trying
if ( $EVAL_ERROR ) {
_die("Cannot connect to MySQL: $EVAL_ERROR", MYSQL_CONNECTION_ERROR);
}
my $done = 0;
do {
eval { $cxn->connect() }; # connect or die trying
if ( $EVAL_ERROR ) {
if ($args{'wait_no_die'} && ($cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
} else {
die("Cannot connect to MySQL: $EVAL_ERROR", MYSQL_CONNECTION_ERROR);
}
} else {
$done = 1;
}
} until $done;
return $cxn;
};
@@ -8860,6 +8915,7 @@ sub main {
errok => (not $intolerant)
);
},
wait_no_die => !$o->get('fail-on-stopped-replication'),
);
if ($slaves_to_skip) {
@@ -8870,6 +8926,7 @@ sub main {
if ($slave->{dsn}->{h} eq $slave_to_skip->{h} && $slave->{dsn}->{P} eq $slave_to_skip->{P}) {
print "Skipping slave " . $slave->description() . "\n";
$is_skip = 1;
$slave->{dbh}->disconnect();
last;
}
}
@@ -9009,10 +9066,8 @@ sub main {
# to stop lagging. If any replica is stopped, the tool
# waits forever until the replica is started."
# https://bugs.launchpad.net/percona-toolkit/+bug/1402051
#TODO REMOVE DEBUG
PTDEBUG && _d('2> Cannot connect to', $cxn->name(), ':',
$EVAL_ERROR);
die '2> Cannot connect to '. $cxn->name() . ':' . $EVAL_ERROR;
# Make ReplicaLagWaiter::wait() report slave is stopped.
return undef;
}
@@ -9023,7 +9078,6 @@ sub main {
};
if ( $EVAL_ERROR ) {
PTDEBUG && _d('Cannot get lag for', $cxn->name(), ':', $EVAL_ERROR);
die '2> Cannot connect to '. $cxn->name() . ':' . $EVAL_ERROR;
}
return $lag; # undef if error
};
@@ -10011,7 +10065,19 @@ sub main {
# Wait forever for slaves to catch up.
$replica_lag_pr->start() if $replica_lag_pr;
$replica_lag->wait(Progress => $replica_lag_pr);
LAG:
{
eval { $replica_lag->wait(Progress => $replica_lag_pr) };
if ( $EVAL_ERROR ) {
if (!$o->get('fail-on-stopped-replication')) {
PTDEBUG && _d($EVAL_ERROR);
sleep 1;
next LAG;
} else {
_die("Error while waiting for replica lag: " . $EVAL_ERROR);
}
}
}
# Wait forever for system load to abate. wait() will die if
# --critical load is reached.
@@ -13093,6 +13159,13 @@ tool.
For more information, visit L<https://www.percona.com/doc/percona-toolkit/LATEST/version-check.html>.
=item --[no]fail-on-stopped-replication
default: yes
If replication is stopped, fail with an error (exit status 128) instead of waiting
until replication is restarted.
=back
=head1 PLUGIN

79
bin/pt-query-digest
View File

@@ -10560,6 +10560,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -10567,6 +10568,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -10622,6 +10624,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -10631,23 +10647,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -11269,18 +11291,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

79
bin/pt-slave-find
View File

@@ -2323,6 +2323,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -2330,6 +2331,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -2385,6 +2387,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -2394,23 +2410,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -3032,18 +3054,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

79
bin/pt-slave-restart
View File

@@ -2734,6 +2734,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -2741,6 +2742,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -2796,6 +2798,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -2805,23 +2821,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -3443,18 +3465,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

2
bin/pt-stalk
View File

@@ -1103,7 +1103,7 @@ collect_mysql_data_two() {
}
open_tables() {
local open_tables=$($CMD_MYSQLADMIN $EXT_ARGV ext | grep "Open_tables" | awk '{print $4}')
local open_tables=$($CMD_MYSQLADMIN $EXT_ARGV ext | grep "Open_tables" | grep -v "Open_tables_with_triggers" | awk '{print $4}')
if [ -n "$open_tables" -a $open_tables -le 1000 ]; then
$CMD_MYSQL $EXT_ARGV -e 'SHOW OPEN TABLES' &
else

86
bin/pt-table-checksum
View File

@@ -5218,6 +5218,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -5225,6 +5226,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -5280,6 +5282,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -5289,23 +5305,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -5927,18 +5949,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}
@@ -8719,9 +8762,9 @@ sub wait {
my ($self) = @_;
my ($slaves, $refresher) = ($self->{slaves}, $self->{get_slaves_cb});
return $slaves if ( not defined $refresher );
my $before = join ' ', sort map {$_->name()} @$slaves;
my $before = join ' ', sort map {$_->description()} @$slaves;
$slaves = $refresher->();
my $after = join ' ', sort map {$_->name()} @$slaves;
my $after = join ' ', sort map {$_->description()} @$slaves;
if ($before ne $after) {
$self->{slaves} = $slaves;
printf STDERR "Slave set to watch has changed\n Was: %s\n Now: %s\n",
@@ -8736,9 +8779,10 @@ sub wait {
$pr_callback = sub {
my ($fraction, $elapsed, $remaining, $eta, $completed) = @_;
my $dsn_name = $worst->{cxn}->name();
my $dsn_description = $worst->{cxn}->description();
if ( defined $worst->{lag} ) {
print STDERR "Replica lag is " . ($worst->{lag} || '?')
. " seconds on $dsn_name. Waiting.\n";
. " seconds on $dsn_description. Waiting.\n";
}
else {
if ($self->{fail_on_stopped_replication}) {

79
bin/pt-table-sync
View File

@@ -6734,6 +6734,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -6741,6 +6742,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -6796,6 +6798,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -6805,23 +6821,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -7443,18 +7465,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

79
lib/MasterSlave.pm
View File

@@ -104,6 +104,7 @@ sub get_slaves {
push @$slaves, $make_cxn->(dsn => $slave_dsn, dbh => $dbh, parent => $parent);
return;
},
wait_no_die => $args{'wait_no_die'},
}
);
} elsif ( $methods->[0] =~ m/^dsn=/i ) {
@@ -111,6 +112,7 @@ sub get_slaves {
$slaves = $self->get_cxn_from_dsn_table(
%args,
dsn_table_dsn => $dsn_table_dsn,
wait_no_die => $args{'wait_no_die'},
);
}
elsif ( $methods->[0] =~ m/none/i ) {
@@ -188,6 +190,20 @@ sub recurse_to_slaves {
my $dbh = $args->{dbh};
my $get_dbh = sub {
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 }
);
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
};
DBH: {
if ( !defined $dbh ) {
foreach my $known_slave ( @{$args->{slaves}} ) {
@@ -197,23 +213,29 @@ sub recurse_to_slaves {
last DBH;
}
}
eval {
$dbh = $dp->get_dbh(
$dp->get_cxn_params($slave_dsn), { AutoCommit => 1 });
PTDEBUG && _d('Connected to', $dp->as_string($slave_dsn));
};
if ( $EVAL_ERROR ) {
print STDERR "Cannot connect to ", $dp->as_string($slave_dsn), ": ", $EVAL_ERROR, "\n"
or die "Cannot print: $OS_ERROR";
return;
}
$get_dbh->();
}
}
my $sql = 'SELECT @@SERVER_ID';
PTDEBUG && _d($sql);
my ($id) = $dbh->selectrow_array($sql);
my $id = undef;
do {
eval {
($id) = $dbh->selectrow_array($sql);
};
if ( $EVAL_ERROR ) {
if ( $args->{wait_no_die} ) {
print STDERR "Error getting server id: ", $EVAL_ERROR,
"\nRetrying query for server ", $slave_dsn->{h}, ":", $slave_dsn->{P}, "\n";
sleep 1;
$dbh->disconnect();
$get_dbh->();
} else {
die $EVAL_ERROR;
}
}
} until ($id);
PTDEBUG && _d('Working on server ID', $id);
my $master_thinks_i_am = $dsn->{server_id};
if ( !defined $id
@@ -994,18 +1016,39 @@ sub get_cxn_from_dsn_table {
. "or a database-qualified table (t)";
}
my $done = 0;
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);
use Data::Dumper;
DSN:
do {
@cxn = ();
my $dsn_strings = $dbh->selectcol_arrayref($sql);
if ( $dsn_strings ) {
foreach my $dsn_string ( @$dsn_strings ) {
PTDEBUG && _d('DSN from DSN table:', $dsn_string);
if ($args{wait_no_die}) {
my $lcxn;
eval {
$lcxn = $make_cxn->(dsn_string => $dsn_string);
};
if ( $EVAL_ERROR && ($dsn_tbl_cxn->lost_connection($EVAL_ERROR)
|| $EVAL_ERROR =~ m/Can't connect to MySQL server/)) {
PTDEBUG && _d("Server is not accessible, waiting when it is online again");
sleep(1);
goto DSN;
}
push @cxn, $lcxn;
} else {
push @cxn, $make_cxn->(dsn_string => $dsn_string);
}
}
}
}
$done = 1;
} until $done;
return \@cxn;
}

7
lib/ReplicaLagWaiter.pm
View File

@@ -87,9 +87,9 @@ sub wait {
my ($self) = @_;
my ($slaves, $refresher) = ($self->{slaves}, $self->{get_slaves_cb});
return $slaves if ( not defined $refresher );
my $before = join ' ', sort map {$_->name()} @$slaves;
my $before = join ' ', sort map {$_->description()} @$slaves;
$slaves = $refresher->();
my $after = join ' ', sort map {$_->name()} @$slaves;
my $after = join ' ', sort map {$_->description()} @$slaves;
if ($before ne $after) {
$self->{slaves} = $slaves;
printf STDERR "Slave set to watch has changed\n Was: %s\n Now: %s\n",
@@ -106,9 +106,10 @@ sub wait {
$pr_callback = sub {
my ($fraction, $elapsed, $remaining, $eta, $completed) = @_;
my $dsn_name = $worst->{cxn}->name();
my $dsn_description = $worst->{cxn}->description();
if ( defined $worst->{lag} ) {
print STDERR "Replica lag is " . ($worst->{lag} || '?')
. " seconds on $dsn_name. Waiting.\n";
. " seconds on $dsn_description. Waiting.\n";
}
else {
if ($self->{fail_on_stopped_replication}) {

2
lib/bash/collect.sh
View File

@@ -392,7 +392,7 @@ collect_mysql_data_two() {
}
open_tables() {
local open_tables=$($CMD_MYSQLADMIN $EXT_ARGV ext | grep "Open_tables" | awk '{print $4}')
local open_tables=$($CMD_MYSQLADMIN $EXT_ARGV ext | grep "Open_tables" | grep -v "Open_tables_with_triggers" | awk '{print $4}')
if [ -n "$open_tables" -a $open_tables -le 1000 ]; then
$CMD_MYSQL $EXT_ARGV -e 'SHOW OPEN TABLES' &
else

2
t/lib/bash/collect.sh
View File

@@ -30,6 +30,8 @@ collect "$PT_TMPDIR/collect" "2011_12_05" > $p-output 2>&1
wait_for_files "$p-hostname" "$p-opentables2" "$p-variables" "$p-df" "$p-innodbstatus2"
cat "$p-opentables2" > /tmp/collect.test
# Even if this system doesn't have all the cmds, collect should still
# have created some files for cmds that (hopefully) all systems have.
ls -1 $PT_TMPDIR/collect | sort > $PT_TMPDIR/collect-files

83
t/pt-online-schema-change/pt-1760.t Normal file
View File

@@ -0,0 +1,83 @@
#!/usr/bin/env perl
BEGIN {
die "The PERCONA_TOOLKIT_BRANCH environment variable is not set.\n"
unless $ENV{PERCONA_TOOLKIT_BRANCH} && -d $ENV{PERCONA_TOOLKIT_BRANCH};
unshift @INC, "$ENV{PERCONA_TOOLKIT_BRANCH}/lib";
};
use strict;
use warnings FATAL => 'all';
use threads;
use English qw(-no_match_vars);
use Test::More;
use Data::Dumper;
use PerconaTest;
use Sandbox;
use SqlModes;
use File::Temp qw/ tempdir tempfile /;
our $delay = 10;
my $tmp_file = File::Temp->new();
my $tmp_file_name = $tmp_file->filename;
unlink $tmp_file_name;
require "$trunk/bin/pt-online-schema-change";
my $dp = new DSNParser(opts=>$dsn_opts);
my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
if ($sb->is_cluster_mode) {
plan skip_all => 'Not for PXC';
}
my $master_dbh = $sb->get_dbh_for('master');
my $slave_dbh1 = $sb->get_dbh_for('slave1');
my $slave_dbh2 = $sb->get_dbh_for('slave2');
my $master_dsn = 'h=127.0.0.1,P=12345,u=msandbox,p=msandbox';
my $slave_dsn1 = 'h=127.0.0.1,P=12346,u=msandbox,p=msandbox';
my $slave_dsn2 = 'h=127.0.0.1,P=12347,u=msandbox,p=msandbox';
my $sample = "t/pt-online-schema-change/samples";
$slave_dbh1->do("stop slave");
$slave_dbh1->do("reset slave all");
$slave_dbh1->do("CHANGE MASTER TO MASTER_HOST='127.0.0.1', MASTER_PORT=12345, MASTER_USER = 'msandbox', MASTER_PASSWORD='msandbox' FOR CHANNEL 'channel1';");
$slave_dbh1->do("start slave");
diag('Loading test data');
$sb->load_file('master', "t/pt-online-schema-change/samples/slave_lag.sql");
my $num_rows = 5000;
diag("Loading $num_rows into the table. This might take some time.");
diag(`util/mysql_random_data_load --host=127.0.0.1 --port=12345 --user=msandbox --password=msandbox test pt178 $num_rows`);
my $output = output(
sub { pt_online_schema_change::main("$master_dsn,D=test,t=pt178",
'--execute',
'--alter', "force",
'--recurse', '1',
'--max-lag', '2',
'--channel', 'channel1')
},
);
like(
$output,
qr/Successfully altered `test`.`pt178`/s,
'pt-osc completes successfully when replication channel used',
);
# #############################################################################
# Done.
# #############################################################################
$slave_dbh1->do('STOP SLAVE');
$master_dbh->do("RESET MASTER");
$slave_dbh1->do('RESET SLAVE ALL');
$slave_dbh1->do("CHANGE MASTER TO MASTER_HOST='127.0.0.1', MASTER_PORT=12345, MASTER_USER = 'msandbox', MASTER_PASSWORD='msandbox';");
$slave_dbh1->do('START SLAVE');
$sb->wipe_clean($master_dbh);
ok($sb->ok(), "Sandbox servers") or BAIL_OUT(__FILE__ . " broke the sandbox");
done_testing;

188
t/pt-online-schema-change/pt-2168-cancel.t Normal file
View File

@@ -0,0 +1,188 @@
#!/usr/bin/env perl
BEGIN {
die "The PERCONA_TOOLKIT_BRANCH environment variable is not set.\n"
unless $ENV{PERCONA_TOOLKIT_BRANCH} && -d $ENV{PERCONA_TOOLKIT_BRANCH};
unshift @INC, "$ENV{PERCONA_TOOLKIT_BRANCH}/lib";
};
use strict;
use warnings FATAL => 'all';
use threads;
use English qw(-no_match_vars);
use Test::More;
use Data::Dumper;
use PerconaTest;
use Sandbox;
use SqlModes;
use File::Temp qw/ tempdir tempfile /;
our $delay = 10;
my $tmp_file = File::Temp->new();
my $tmp_file_name = $tmp_file->filename;
unlink $tmp_file_name;
require "$trunk/bin/pt-online-schema-change";
my $dp = new DSNParser(opts=>$dsn_opts);
my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
if ($sb->is_cluster_mode) {
plan skip_all => 'Not for PXC';
}
# We need third slave to redirect pt-osc in case of one or standard disconnects
diag(`$trunk/sandbox/start-sandbox slave 12348 12345`);
my $master_dbh = $sb->get_dbh_for('master');
my $slave_dbh1 = $sb->get_dbh_for('slave1');
my $slave_dbh2 = $sb->get_dbh_for('slave2');
my $slave_dbh3 = $sb->get_dbh_for('master1');
my $master_dsn = 'h=127.0.0.1,P=12345,u=msandbox,p=msandbox';
my $slave_dsn1 = 'h=127.0.0.1,P=12346,u=msandbox,p=msandbox';
my $slave_dsn2 = 'h=127.0.0.1,P=12347,u=msandbox,p=msandbox';
my $sample = "t/pt-online-schema-change/samples";
# We need sync_relay_log=1 to have
my $cnf = '/tmp/12347/my.sandbox.cnf';
diag(`cp $cnf $cnf.bak`);
diag(`echo "[mysqld]" > /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "sync_relay_log=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "sync_relay_log_info=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "relay_log_recovery=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "!include /tmp/12347/my.sandbox.2.cnf" >> $cnf`);
diag(`/tmp/12347/stop >/dev/null`);
sleep 1;
diag(`/tmp/12347/start >/dev/null`);
# DSN table for further tests
$sb->load_file('master', "$sample/create_dsns.sql");
#diag(`mysql -h127.0.0.1 -P12345 -umsandbox -pmsandbox -e "show databases"`);
#diag(`mysql -h127.0.0.1 -P12347 -umsandbox -pmsandbox -e "show databases"`);
#diag(`mysql -h127.0.0.1 -P12348 -umsandbox -pmsandbox -e "show databases"`);
#diag(`mysql -h127.0.0.1 -P12348 -umsandbox -pmsandbox -E -e "show slave status"`);
sub reset_query_cache {
my @dbhs = @_;
return if ($sandbox_version >= '8.0');
foreach my $dbh (@dbhs) {
$dbh->do('RESET QUERY CACHE');
}
}
# 1) Set the slave delay to 0 just in case we are re-running the tests without restarting the sandbox.
# 2) Load sample data
# 3) Set the slave delay to 30 seconds to be able to see the 'waiting' message.
diag("Setting slave delay to 0 seconds");
$sb->wait_for_slaves(slave => 'master1');
$slave_dbh1->do('STOP SLAVE');
$slave_dbh3->do('STOP SLAVE');
$master_dbh->do("RESET MASTER");
$slave_dbh1->do('RESET SLAVE');
$slave_dbh1->do('START SLAVE');
$slave_dbh3->do('RESET SLAVE');
$slave_dbh3->do('START SLAVE');
diag('Loading test data');
$sb->load_file('master', "t/pt-online-schema-change/samples/slave_lag.sql");
# Should be greater than chunk-size and big enough, so pt-osc will wait for delay
my $num_rows = 5000;
diag("Loading $num_rows into the table. This might take some time.");
diag(`util/mysql_random_data_load --host=127.0.0.1 --port=12345 --user=msandbox --password=msandbox test pt178 $num_rows`);
$sb->wait_for_slaves();
$sb->wait_for_slaves(slave => 'master1');
# Plan for tests
# 1. Basic test: start tool on some huge table, stop slave, wait few seconds, start slave. Check if tool restarted with option and failed with error without.
# 2. Delayed slaves
# 3. Places to test:
# - get_dbh
# - SELECT @@SERVER_ID
# 4. Slave never returns
# - die after timeout
# - inject new slave
# - ignore after timeout
diag("Setting slave delay to $delay seconds");
$slave_dbh1->do('STOP SLAVE');
$slave_dbh1->do("CHANGE MASTER TO MASTER_DELAY=$delay");
$slave_dbh1->do('START SLAVE');
# Run a full table scan query to ensure the slave is behind the master
# There is no query cache in MySQL 8.0+
reset_query_cache($master_dbh, $master_dbh);
# Update one row so slave is delayed
$master_dbh->do('UPDATE `test`.`pt178` SET f2 = f2 + 1 LIMIT 1');
$master_dbh->do('UPDATE `test`.`pt178` SET f2 = f2 + 1 WHERE f1 = ""');
diag("Starting tests...");
my $max_lag = $delay / 2;
# We need to sleep, otherwise pt-osc can finish before slave is delayed
sleep($max_lag);
my $args = "$master_dsn,D=test,t=pt178 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5 --nofail-on-stopped-replication";
my ($fh, $filename) = tempfile();
my $pid = fork();
if (!$pid) {
open(STDERR, '>', $filename);
open(STDOUT, '>', $filename);
exec("$trunk/bin/pt-online-schema-change $args");
}
sleep($max_lag + $max_lag/2);
# restart slave 12347
diag(`/tmp/12347/stop >/dev/null`);
sleep 1;
$master_dbh->do("UPDATE test_recursion_method.dsns SET dsn='D=test_recursion_method,t=dsns,P=12348,h=127.0.0.1,u=root,p=msandbox' WHERE id=2");
waitpid($pid, 0);
my $output = do {
local $/ = undef;
<$fh>;
};
like(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc completes successfully when one of replicas is stopped, option --nofail-on-stopped-replication is specified, and another replica was specified in the dsns table as a replacement",
);
diag(`/tmp/12347/start >/dev/null`);
# #############################################################################
# Done.
# #############################################################################
diag("Cleaning");
diag(`$trunk/sandbox/stop-sandbox 12348`);
$slave_dbh2 = $sb->get_dbh_for('slave2');
diag("Setting slave delay to 0 seconds");
$slave_dbh1->do('STOP SLAVE');
$slave_dbh2->do('STOP SLAVE');
$master_dbh->do("RESET MASTER");
$slave_dbh1->do('RESET SLAVE');
$slave_dbh2->do('RESET SLAVE');
$slave_dbh1->do('START SLAVE');
$slave_dbh2->do('START SLAVE');
#$slave_dbh2->do("uninstall plugin simple_rewrite_plugin");
diag(`mv $cnf.bak $cnf`);
diag(`/tmp/12347/stop >/dev/null`);
diag(`/tmp/12347/start >/dev/null`);
diag("Dropping test database");
$master_dbh->do("DROP DATABASE IF EXISTS test");
$sb->wait_for_slaves();
$sb->wipe_clean($master_dbh);
ok($sb->ok(), "Sandbox servers") or BAIL_OUT(__FILE__ . " broke the sandbox");
done_testing;

350
t/pt-online-schema-change/pt-2168.t Normal file
View File

@@ -0,0 +1,350 @@
#!/usr/bin/env perl
BEGIN {
die "The PERCONA_TOOLKIT_BRANCH environment variable is not set.\n"
unless $ENV{PERCONA_TOOLKIT_BRANCH} && -d $ENV{PERCONA_TOOLKIT_BRANCH};
unshift @INC, "$ENV{PERCONA_TOOLKIT_BRANCH}/lib";
};
use strict;
use warnings FATAL => 'all';
use threads;
use English qw(-no_match_vars);
use Test::More;
use Data::Dumper;
use PerconaTest;
use Sandbox;
use SqlModes;
use File::Temp qw/ tempdir tempfile /;
our $delay = 10;
my $tmp_file = File::Temp->new();
my $tmp_file_name = $tmp_file->filename;
unlink $tmp_file_name;
require "$trunk/bin/pt-online-schema-change";
my $dp = new DSNParser(opts=>$dsn_opts);
my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
if ($sb->is_cluster_mode) {
plan skip_all => 'Not for PXC';
}
my $master_dbh = $sb->get_dbh_for('master');
my $slave_dbh1 = $sb->get_dbh_for('slave1');
my $slave_dbh2 = $sb->get_dbh_for('slave2');
my $master_dsn = 'h=127.0.0.1,P=12345,u=msandbox,p=msandbox';
my $slave_dsn1 = 'h=127.0.0.1,P=12346,u=msandbox,p=msandbox';
my $slave_dsn2 = 'h=127.0.0.1,P=12347,u=msandbox,p=msandbox';
my $sample = "t/pt-online-schema-change/samples";
# We need sync_relay_log=1 to have
my $cnf = '/tmp/12347/my.sandbox.cnf';
diag(`cp $cnf $cnf.bak`);
diag(`echo "[mysqld]" > /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "sync_relay_log=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "sync_relay_log_info=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "relay_log_recovery=1" >> /tmp/12347/my.sandbox.2.cnf`);
diag(`echo "!include /tmp/12347/my.sandbox.2.cnf" >> $cnf`);
diag(`/tmp/12347/stop >/dev/null`);
sleep 1;
diag(`/tmp/12347/start >/dev/null`);
sub reset_query_cache {
my @dbhs = @_;
return if ($sandbox_version >= '8.0');
foreach my $dbh (@dbhs) {
$dbh->do('RESET QUERY CACHE');
}
}
# 1) Set the slave delay to 0 just in case we are re-running the tests without restarting the sandbox.
# 2) Load sample data
# 3) Set the slave delay to 30 seconds to be able to see the 'waiting' message.
diag("Setting slave delay to 0 seconds");
$slave_dbh1->do('STOP SLAVE');
$master_dbh->do("RESET MASTER");
$slave_dbh1->do('RESET SLAVE');
$slave_dbh1->do('START SLAVE');
diag('Loading test data');
$sb->load_file('master', "t/pt-online-schema-change/samples/slave_lag.sql");
# Should be greater than chunk-size and big enough, so pt-osc will wait for delay
my $num_rows = 5000;
diag("Loading $num_rows into the table. This might take some time.");
diag(`util/mysql_random_data_load --host=127.0.0.1 --port=12345 --user=msandbox --password=msandbox test pt178 $num_rows`);
# DSN table for further tests
$sb->load_file('master', "$sample/create_dsns.sql");
$sb->wait_for_slaves();
# Plan for tests
# 1. Basic test: start tool on some huge table, stop slave, wait few seconds, start slave. Check if tool restarted with option and failed with error without.
# 2. Delayed slaves
# 3. Places to test:
# - get_dbh
# - SELECT @@SERVER_ID
# 4. Slave never returns
# - die after timeout
# - inject new slave
# - ignore after timeout
diag("Setting slave delay to $delay seconds");
$slave_dbh1->do('STOP SLAVE');
$slave_dbh1->do("CHANGE MASTER TO MASTER_DELAY=$delay");
$slave_dbh1->do('START SLAVE');
# Run a full table scan query to ensure the slave is behind the master
# There is no query cache in MySQL 8.0+
reset_query_cache($master_dbh, $master_dbh);
# Update one row so slave is delayed
$master_dbh->do('UPDATE `test`.`pt178` SET f2 = f2 + 1 LIMIT 1');
$master_dbh->do('UPDATE `test`.`pt178` SET f2 = f2 + 1 WHERE f1 = ""');
# This is the base test, just to ensure that without using --check-slave-lag nor --skip-check-slave-lag
# pt-online-schema-change will wait on the slave at port 12346
my $max_lag = $delay / 2;
# We need to sleep, otherwise pt-osc can finish before slave is delayed
sleep($max_lag);
# Basic test: we check if pt-osc fails if replica restarted while it is running with default options
sub base_test {
my ($args) = @_;
#diag("pid: $tmp_file_name");
my ($fh, $filename) = tempfile();
my $pid = fork();
if (!$pid) {
open(STDERR, '>', $filename);
open(STDOUT, '>', $filename);
exec("$trunk/bin/pt-online-schema-change $args");
}
sleep($max_lag + $max_lag/2);
# restart slave 12347
diag(`/tmp/12347/stop >/dev/null`);
sleep 1;
diag(`/tmp/12347/start >/dev/null`);
waitpid($pid, 0);
my $output = do {
local $/ = undef;
<$fh>;
};
unlink $filename;
return $output;
}
sub crash_test {
my ($args) = @_;
#diag("pid: $tmp_file_name");
my ($fh, $filename) = tempfile();
my $pid = fork();
if (!$pid) {
open(STDERR, '>', $filename);
open(STDOUT, '>', $filename);
exec("$trunk/bin/pt-online-schema-change $args");
}
sleep($max_lag + 10);
# restart slave 12347
diag(`/tmp/12347/start >/dev/null`);
waitpid($pid, 0);
my $output = do {
local $/ = undef;
<$fh>;
};
unlink $filename;
return $output;
}
sub error_test {
my ($test, $pattern, $query) = @_;
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_action='rewrite'");
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_pattern='$pattern'");
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_query='$query'");
my $args = "$master_dsn,D=test,t=pt178,A=utf8 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5";
my $output = `$trunk/bin/pt-online-schema-change $args 2>&1`;
unlike(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc fails with error if replica returns error when $test",
);
$args = "$master_dsn,D=test,t=pt178,A=utf8 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5 --nofail-on-stopped-replication";
$output = `$trunk/bin/pt-online-schema-change $args 2>&1`;
unlike(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc fails with error if replica returns error when $test and option --nofail-on-stopped-replication is specified",
);
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_pattern=''");
$slave_dbh2 = $sb->get_dbh_for('slave2');
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_pattern='$pattern'");
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_action='abort'");
$args = "$master_dsn,D=test,t=pt178,A=utf8 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5";
$output = crash_test($args);
unlike(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc fails with error if replica disconnects when $test",
);
$slave_dbh2 = $sb->get_dbh_for('slave2');
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_pattern='$pattern'");
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_action='abort'");
$args = "$master_dsn,D=test,t=pt178,A=utf8 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5 --nofail-on-stopped-replication";
$output = crash_test($args);
like(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc finishes succesfully if replica disconnects when $test and option --nofail-on-stopped-replication is specified",
);
$slave_dbh2 = $sb->get_dbh_for('slave2');
$slave_dbh2->do("SET GLOBAL simple_rewrite_plugin_action='rewrite'");
}
diag("Starting base tests. This is going to take some time due to the delay in the slave");
my $output = base_test("$master_dsn,D=test,t=pt178 --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5");
unlike(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc fails when one of replicas is restarted",
);
# pt-osc doesn't fail if replica is restarted and option --nofail-on-stopped-replication specified
$output = base_test("$master_dsn,D=test,t=pt178 --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5 --nofail-on-stopped-replication");
like(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc completes successfully when one of replicas is restarted and option --nofail-on-stopped-replication is specified",
);
$output = base_test("$master_dsn,D=test,t=pt178 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5");
unlike(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc fails with recursion-method=dsn when one of replicas is restarted",
);
$output = base_test("$master_dsn,D=test,t=pt178 --recursion-method=dsn=D=test_recursion_method,t=dsns,h=127.0.0.1,P=12345,u=msandbox,p=msandbox --execute --chunk-size 10 --max-lag $max_lag --alter 'engine=INNODB' --pid $tmp_file_name --progress time,5 --nofail-on-stopped-replication");
like(
$output,
qr/Successfully altered `test`.`pt178`/s,
"pt-osc completes successfully with recursion-method=dsn when one of replicas is restarted and option --nofail-on-stopped-replication is specified",
);
# Errors that happen while pt-osc executes SQL while checking slave availability.
# We check few scenarios.
# - Error not related to connection: pt-osc aborted regardless option --nofail-on-stopped-replication
# - Error, related to connection: pt-osc behavior depends on option --nofail-on-stopped-replication
# We work only with replica with port 12347 here.
diag("Starting replica lost and error tests");
SKIP: {
$slave_dbh2 = $sb->get_dbh_for('slave2');
eval { $slave_dbh2->do("install plugin simple_rewrite_plugin soname 'simple_rewrite_plugin.so'") };
if ( $EVAL_ERROR && $EVAL_ERROR !~ m/Function 'simple_rewrite_plugin' already exists/) {
skip 'These tests require simple_rewrite_plugin. You can get it from https://github.com/svetasmirnova/simple_rewrite_plugin';
}
my @res = $slave_dbh2->selectrow_array("select count(*) from information_schema.plugins where plugin_name='simple_rewrite_plugin' and PLUGIN_STATUS='ACTIVE'");
if ( $res[0] != 1 ) {
skip 'These tests require simple_rewrite_plugin in active status';
}
# get_dbh sets character set connection
$master_dbh->do("UPDATE test_recursion_method.dsns SET dsn='D=test_recursion_method,t=dsns,P=12346,h=127.0.0.1,u=root,p=msandbox,A=utf8' WHERE id=1");
$master_dbh->do("UPDATE test_recursion_method.dsns SET dsn='D=test_recursion_method,t=dsns,P=12347,h=127.0.0.1,u=root,p=msandbox,A=utf8' WHERE id=2");
error_test("setting character set", '.*(SET NAMES) "?([[:alnum:]]+)"?.*', '$1 $2$2');
$master_dbh->do("UPDATE test_recursion_method.dsns SET dsn='D=test_recursion_method,t=dsns,P=12346,h=127.0.0.1,u=root,p=msandbox' WHERE id=1");
$master_dbh->do("UPDATE test_recursion_method.dsns SET dsn='D=test_recursion_method,t=dsns,P=12347,h=127.0.0.1,u=root,p=msandbox' WHERE id=2");
# get_dbh selects SQL mode
error_test("selecting SQL mode", 'SELECT @@SQL_MODE', 'SELEC @@SQL_MODE');
# get_dbh sets SQL mode
error_test("setting SQL_QUOTE_SHOW_CREATE", 'SET @@SQL_QUOTE_SHOW_CREATE.*', 'SE @@SQL_QUOTE_SHOW_CREATE = 1');
# get_dbh selects version
error_test("selecting MySQL version", 'SELECT VERSION.*', 'SELEC VERSION()');
# get_dbh queries server character set
error_test("querying server character set", "SHOW VARIABLES LIKE \\'character_set_server\\'", "SHO VARIABLES LIKE \\'character_set_server\\'");
# get_dbh sets character set utf8mb4 in version 8+
if ($sandbox_version ge '8.0') {
error_test("setting character set utf8mb4", "SET NAMES \\'utf8mb4\\'", "SET NAMES \\'utf8mb4utf8mb4\\'");
}
# recurse_to_slaves asks for SERVER_ID
error_test("selecting server id", 'SELECT @@SERVER_ID.*', 'SELEC @@SERVER_ID');
$slave_dbh2 = $sb->get_dbh_for('slave2');
$slave_dbh2->do("uninstall plugin simple_rewrite_plugin");
}
# #############################################################################
# Done.
# #############################################################################
diag("Cleaning");
$slave_dbh2 = $sb->get_dbh_for('slave2');
diag("Setting slave delay to 0 seconds");
$slave_dbh1->do('STOP SLAVE');
$slave_dbh2->do('STOP SLAVE');
$master_dbh->do("RESET MASTER");
$slave_dbh1->do('RESET SLAVE');
$slave_dbh2->do('RESET SLAVE');
$slave_dbh1->do('START SLAVE');
$slave_dbh2->do('START SLAVE');
diag(`mv $cnf.bak $cnf`);
diag(`/tmp/12347/stop >/dev/null`);
diag(`/tmp/12347/start >/dev/null`);
diag("Dropping test database");
$master_dbh->do("DROP DATABASE IF EXISTS test");
$sb->wait_for_slaves();
$sb->wipe_clean($master_dbh);
ok($sb->ok(), "Sandbox servers") or BAIL_OUT(__FILE__ . " broke the sandbox");
done_testing;

2
t/pt-online-schema-change/slave_lag.t
View File

@@ -184,7 +184,7 @@ $output = `$trunk/bin/pt-online-schema-change $args 2>&1`;
unlike(
$output,
qr/Replica lag is \d+ seconds on .* Waiting/s,
qr/Replica lag is \d+ seconds on .*:12346. Waiting/s,
"--skip-check-slave-lag is really skipping the slave",
);

4
t/pt-table-sync/bidirectional.t
View File

@@ -187,7 +187,7 @@ is(
is(
$err,
"# Cannot resolve conflict WHERE `id`='3': `ts` values do not differ by the threhold, 30m.
"# Cannot resolve conflict WHERE `id`='3': `ts` values do not differ by the threshold, 30m.
",
'Warns about conflict'
);
@@ -266,7 +266,7 @@ is(
# syncing other tables.
is(
$err,
"# Cannot resolve conflict WHERE `id`='3': `ts` values do not differ by the threhold, 30m. while doing bidi.t on 127.1
"# Cannot resolve conflict WHERE `id`='3': `ts` values do not differ by the threshold, 30m. while doing bidi.t on 127.1
",
'Die/warn about conflict'
);