PT-1585 Added --force-nibbling in pt-tc

bin/pt-online-schema-change

@@ -5838,6 +5838,7 @@ sub can_nibble {
    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}
@@ -5852,6 +5853,16 @@ sub can_nibble {
       die "There is no good index and the table is oversized.";
    }
 
+   if ($o->has('force-nibbling') && $o->get('force-nibbling')) {
+       my @nibbling_tables = split(/,/, $o->get('force-nibbling'));
+       my @table_in_list = grep (/$tbl->{tbl}/, @nibbling_tables);
+       if (@table_in_list && $index && $row_est >= 2) {
+           PTDEBUG && _d("Disabling one nibble for the entire table");
+           $one_nibble = 0;
+       }
+   }
+
+
    my $pause_file = ($o->has('pause-file') && $o->get('pause-file')) || undef;
    
    return {

bin/pt-table-checksum

@@ -6689,6 +6689,7 @@ sub can_nibble {
    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}
@@ -6703,6 +6704,16 @@ sub can_nibble {
       die "There is no good index and the table is oversized.";
    }
 
+   if ($o->has('force-nibbling') && $o->get('force-nibbling')) {
+       my @nibbling_tables = split(/,/, $o->get('force-nibbling'));
+       my @table_in_list = grep (/$tbl->{tbl}/, @nibbling_tables);
+       if (@table_in_list && $index && $row_est >= 2) {
+           PTDEBUG && _d("Disabling one nibble for the entire table");
+           $one_nibble = 0;
+       }
+   }
+
+
    my $pause_file = ($o->has('pause-file') && $o->get('pause-file')) || undef;
    
    return {
@@ -13262,6 +13273,16 @@ the values are converted to strings by the CONCAT() function, and MySQL chooses
 the string representation.  If you specify a value of 2, for example, then the
 values 1.008 and 1.009 will be rounded to 1.01, and will checksum as equal.
 
+=item --force-nibbling
+
+type: string
+
+Comma sepatered list of tables on which we want to force nibbling instead of
+checksuming the whole table in only one chunk,
+In some cases, MySQL returns a wrong estimate about number of rows in a table 
+and NibbleIterator is trying to process huge tables in a single chunk.
+This parameter disable one nibble checksums for the specified tables.
+
 =item --function
 
 type: string

lib/NibbleIterator.pm

@@ -495,6 +495,8 @@ sub can_nibble {
    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');
 
    # Special case: we're resuming and there's no boundaries, so the table
@@ -512,6 +514,23 @@ sub can_nibble {
       die "There is no good index and the table is oversized.";
    }
 
+   # In some cases, MySQL returns a wrong estimate about number of rows in a table 
+   # and NibbleIterator is trying to process huge tables in a single chunk.
+   # This parameter disable one nibble.
+   # See https://jira.percona.com/browse/PT-1585
+   # This function should work on all scenarios so, to prevent trying to nibble on empty
+   # (or really really small) tables, we need to ensure tha table has an index and it has
+   # at least 2 rows, otherwise there is no way to split only one row into chunks.
+   if ($o->has('force-nibbling') && $o->get('force-nibbling')) {
+       my @nibbling_tables = split(/,/, $o->get('force-nibbling'));
+       my @table_in_list = grep (/$tbl->{tbl}/, @nibbling_tables);
+       if (@table_in_list && $index && $row_est >= 2) {
+           PTDEBUG && _d("Disabling one nibble for the entire table");
+           $one_nibble = 0;
+       }
+   }
+
+
    # The table can be nibbled if this point is reached, else we would have
    # died earlier.  Return some values about nibbling the table.
    my $pause_file = ($o->has('pause-file') && $o->get('pause-file')) || undef;

t/pt-table-checksum/pt-1584.t

@@ -0,0 +1,75 @@
+#!/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 English qw(-no_match_vars);
+use Test::More;
+
+use PerconaTest;
+use Sandbox;
+use SqlModes;
+require "$trunk/bin/pt-table-checksum";
+
+my $dp  = new DSNParser(opts=>$dsn_opts);
+my $sb  = new Sandbox(basedir => '/tmp', DSNParser => $dp);
+my $dbh = $sb->get_dbh_for('master');
+
+if ( !$dbh ) {
+   plan skip_all => 'Cannot connect to sandbox master';
+}
+else {
+   plan tests => 3;
+}
+
+diag("loading samples");
+#$sb->load_file('master', 't/pt-table-checksum/samples/pt-226.sql');
+$sb->load_file('master', 't/pt-table-checksum/samples/pt-1585.sql');
+my $num_rows = 2000;
+diag(`util/mysql_random_data_load --host=127.0.0.1 --port=12345 --user=msandbox --password=msandbox test t1 $num_rows`);
+diag("$num_rows sample rows loaded");
+
+# The sandbox servers run with lock_wait_timeout=3 and it's not dynamic
+# so we need to specify --set-vars innodb_lock_wait_timeout=3 else the tool will die.
+# And --max-load "" prevents waiting for status variables.
+my $master_dsn = $sb->dsn_for('master');
+
+my @args       = ($master_dsn, "--set-vars", "innodb_lock_wait_timeout=50", "--force-nibbling"); 
+my $output;
+my $exit_status;
+warn @args;
+
+$ENV{PTDEBUG} = 1;
+# Test #1 
+$output = output(
+   sub { $exit_status = pt_table_checksum::main(@args) },
+   stderr => 1,
+);
+
+delete $ENV{PTDEBUG};
+
+is(
+   $exit_status,
+   0,
+   "PT-226 SET binlog_format='STATEMENT' exit status",
+);
+
+like(
+    $output,
+    qr/1\s+100\s+0\s+1\s+0\s+.*test.joinit/,
+    "PT-226 table joinit has differences",
+);
+
+$dbh->do('SET GLOBAL binlog_format="STATEMENT"');
+
+# #############################################################################
+# Done.
+# #############################################################################
+$sb->wipe_clean($dbh);
+ok($sb->ok(), "Sandbox servers") or BAIL_OUT(__FILE__ . " broke the sandbox");
+exit;

t/pt-table-checksum/samples/pt-1585.sql

@@ -0,0 +1,9 @@
+DROP DATABASE IF EXISTS test;
+CREATE DATABASE test;
+
+USE test;
+
+CREATE TABLE `test`.`t1` (
+    id INT AUTO_INCREMENT PRIMARY KEY NOT NULL,
+    f1 INT
+) Engine=InnoDB CHARSET=utf8;