percona/percona-toolkit
1
0
Fork 0
mirror of https://github.com/percona/percona-toolkit.git synced 2025-09-26 15:31:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove reminents of stuff copied in from pt-table-checksum. Handle errors properly in exec_nibble().

Browse Source
This commit is contained in:
Daniel Nichter
2012-03-27 19:06:35 -06:00
parent 3961c6d5c3
commit 4c542a71fc
2 changed files with 30 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

78
bin/pt-online-schema-change
View File

@@ -5080,7 +5080,7 @@ sub main {
} }
$msg .= "Please read the --check-replication-filters documentation " $msg .= "Please read the --check-replication-filters documentation "
. "to learn how to solve this problem."; . "to learn how to solve this problem.";
die ts($msg); die $msg;
} }
} }
@@ -5538,6 +5538,7 @@ sub main {
# Adjust chunk size. This affects the next chunk. # Adjust chunk size. This affects the next chunk.
if ( $chunk_time ) { if ( $chunk_time ) {
# Calcuate a new chunk-size based on the rate of rows/s.
$tbl->{chunk_size} = $tbl->{rate}->update( $tbl->{chunk_size} = $tbl->{rate}->update(
$cnt, # processed this many rows $cnt, # processed this many rows
$tbl->{nibble_time}, # is this amount of time $tbl->{nibble_time}, # is this amount of time
@@ -5549,20 +5550,19 @@ sub main {
$tbl->{chunk_size} = 1; $tbl->{chunk_size} = 1;
# This warning is printed once per table. # This warning is printed once per table.
if ( !$tbl->{warned_slow} && $o->get('quiet') < 2 ) { if ( !$tbl->{warned_slow} ) {
warn ts("Checksum queries for table " warn "Rows are copying very slowly. "
. "$tbl->{name} are executing very slowly. "
. "--chunk-size has been automatically reduced to 1. " . "--chunk-size has been automatically reduced to 1. "
. "Check that the server is not being overloaded, " . "Check that the server is not being overloaded, "
. "or increase --chunk-time. The last chunk " . "or increase --chunk-time. The last chunk "
. "selected $cnt rows and took " . "selected $cnt rows and took "
. sprintf('%.3f', $tbl->{nibble_time}) . sprintf('%.3f', $tbl->{nibble_time})
. " seconds to execute.\n"); . " seconds to execute.\n";
$tbl->{warned_slow} = 1; $tbl->{warned_slow} = 1;
} }
} }
# Update chunk-size based on rows/s checksum rate. # Update chunk-size based on the rate of rows/s.
$nibble_iter->set_chunk_size($tbl->{chunk_size}); $nibble_iter->set_chunk_size($tbl->{chunk_size});
} }
@@ -6349,11 +6349,6 @@ sub drop_triggers {
return $exit_status; return $exit_status;
} }
# BARON: in this subroutine we still have a lot of references to checksums, so
# we can get error messages like this:
# Error copying rows from `sbtest`.`sbtest` to `sbtest`.`_sbtest_new`: Error
# executing checksum query: Checksum query for table sbtest.sbtest caused MySQL
# error 1364:
sub exec_nibble { sub exec_nibble {
my (%args) = @_; my (%args) = @_;
my @required_args = qw(Cxn tbl NibbleIterator Retry Quoter OptionParser); my @required_args = qw(Cxn tbl NibbleIterator Retry Quoter OptionParser);
@@ -6395,7 +6390,7 @@ sub exec_nibble {
wait => sub { return; }, wait => sub { return; },
try => sub { try => sub {
# ################################################################### # ###################################################################
# Start timing the checksum query. # Start timing the query.
# ################################################################### # ###################################################################
my $t_start = time; my $t_start = time;
@@ -6411,13 +6406,13 @@ sub exec_nibble {
my $t_end = time; my $t_end = time;
# ################################################################### # ###################################################################
# End timing the checksum query. # End timing the query.
# ################################################################### # ###################################################################
# How many rows were inserted this time. Used for auto chunk sizing. # How many rows were inserted this time. Used for auto chunk sizing.
$tbl->{row_cnt} = $sth->{nibble}->rows(); $tbl->{row_cnt} = $sth->{nibble}->rows();
# Check if checksum query caused any warnings. # Check if query caused any warnings.
my $sql_warn = 'SHOW WARNINGS'; my $sql_warn = 'SHOW WARNINGS';
PTDEBUG && _d($sql_warn); PTDEBUG && _d($sql_warn);
my $warnings = $dbh->selectall_arrayref($sql_warn, { Slice => {} } ); my $warnings = $dbh->selectall_arrayref($sql_warn, { Slice => {} } );
@@ -6433,24 +6428,20 @@ sub exec_nibble {
|| $message =~ m/$warn_code{$code}->{pattern}/) ) || $message =~ m/$warn_code{$code}->{pattern}/) )
{ {
if ( !$tbl->{"warned_code_$code"} ) { # warn once per table if ( !$tbl->{"warned_code_$code"} ) { # warn once per table
if ( $o->get('quiet') < 2 ) { warn "Copying rows caused a MySQL error $code: "
warn "Checksum query for table $tbl->{db}.$tbl->{tbl} " . ($warn_code{$code}->{message}
. "caused MySQL error $code: " ? $warn_code{$code}->{message}
. ($warn_code{$code}->{message} : $message)
? $warn_code{$code}->{message} . "\nThis MySQL error is being ignored and further "
: $message) . "occurrences of it will not be reported.\n";
. "\n";
}
$tbl->{"warned_code_$code"} = 1; $tbl->{"warned_code_$code"} = 1;
$tbl->{checksum_results}->{errors}++;
} }
} }
else { else {
# This die will propagate to fail which will return 0 # This die will propagate to fail which will return 0
# and propagate it to final_fail which will die with # and propagate it to final_fail which will die with
# this error message. (So don't wrap it in ts().) # this error message.
die "Checksum query for table $tbl->{db}.$tbl->{tbl} " die "Copying rows caused a MySQL error $code:\n"
. "caused MySQL error $code:\n"
. " Level: " . ($warning->{level} || '') . "\n" . " Level: " . ($warning->{level} || '') . "\n"
. " Code: " . ($warning->{code} || '') . "\n" . " Code: " . ($warning->{code} || '') . "\n"
. " Message: " . ($warning->{message} || '') . "\n" . " Message: " . ($warning->{message} || '') . "\n"
@@ -6465,49 +6456,34 @@ sub exec_nibble {
my (%args) = @_; my (%args) = @_;
my $error = $args{error}; 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/ if ( $error =~ m/Lock wait timeout exceeded/
|| $error =~ m/Query execution was interrupted/ || $error =~ m/Query execution was interrupted/
) { ) {
# These errors/warnings can be retried, so don't print # These errors/warnings can be retried, so don't print
# a warning yet; do that in final_fail. # a warning yet; do that in final_fail.
return 1; return 1; # try again
} }
elsif ( $error =~ m/MySQL server has gone away/ elsif ( $error =~ m/MySQL server has gone away/
|| $error =~ m/Lost connection to MySQL server/ || $error =~ m/Lost connection to MySQL server/
) { ) {
# The 2nd pattern means that MySQL itself died or was stopped. # The 2nd pattern means that MySQL itself died or was stopped.
# The 3rd pattern means that our cxn was killed (KILL <id>). # The 3rd pattern means that our cxn was killed (KILL <id>).
eval { $dbh = $cxn->connect(); }; $dbh = $cxn->connect(); # connect or die trying
return 1 unless $EVAL_ERROR; # reconnected, retry checksum query return 1; # reconnected, try again
$oktorun = 0; # failed to reconnect, exit tool
} }
# At this point, either the error/warning cannot be retried, # At this point, either the error/warning cannot be retried,
# or we failed to reconnect. So stop trying and call final_fail. # or we failed to reconnect. Don't retry; call final_fail.
return 0; return 0;
}, },
final_fail => sub { final_fail => sub {
my (%args) = @_; my (%args) = @_;
my $error = $args{error}; # This die should be caught by the caller. Copying rows and
# the tool will stop, which is probably good because by this
if ( $error =~ /Lock wait timeout exceeded/ # point the error or warning indicates that something is wrong.
|| $error =~ /Query execution was interrupted/ die $args{error};
) {
# These errors/warnings are not fatal but only cause this
# nibble to be skipped.
if ( $o->get('quiet') < 2 ) {
warn "$error\n";
}
return; # skip this nibble
}
# This die will be caught by the eval inside the TABLE loop.
# Checksumming for this table will stop, which is probably
# good because by this point the error or warning indicates
# that something fundamental is broken or wrong. Checksumming
# will continue with the next table, unless the fail code set
# oktorun=0, in which case the error/warning is fatal.
die "Error executing checksum query: $args{error}\n";
} }
); );
} }

3
t/pt-online-schema-change/basics.t
View File

@@ -28,6 +28,9 @@ my $slave_dbh = $sb->get_dbh_for('slave1');
if ( !$master_dbh ) { if ( !$master_dbh ) {
plan skip_all => 'Cannot connect to sandbox master'; plan skip_all => 'Cannot connect to sandbox master';
} }
elsif ( !$slave_dbh ) {
plan skip_all => 'Cannot connect to sandbox slave';
}
else { else {
plan tests => 55; plan tests => 55;
} }