mysqlhotcopy

开启mysql的远程连接的方法 mysql-noinstall-5.1.6-alpha-win32.zip

	$dbh->do( qq{ replace into $table_name 
			  set host=?, log_file=?, log_pos=?, 
                          master_host=?, master_log_file=?, master_log_pos=? }, 
		  undef, 
		  $hostname, $file, $position, 
		  $master_host, $log_file, $log_pos  );
	
    };
    
    if ( $@ ) {
	warn "Failed to store master position: $@\n";
    }
}

sub get_row {
  my ( $dbh, $sql ) = @_;

  my $sth = $dbh->prepare($sql);
  $sth->execute;
  return $sth->fetchrow_array();
}

sub get_row_hash {
  my ( $dbh, $sql ) = @_;

  my $sth = $dbh->prepare($sql);
  $sth->execute;
  return $sth->fetchrow_hashref();
}

sub scan_raid_dir {
    my ( $r_db_files, $data_dir, @raid_dir ) = @_;

    local(*RAID_DIR);
    
    foreach my $rd ( @raid_dir ) {

	opendir(RAID_DIR, "$data_dir/$rd" ) 
	    or die "Cannot open dir '$data_dir/$rd': $!";

	while ( defined( my $name = readdir RAID_DIR ) ) {
	    $r_db_files->{"$rd/$name"} = $1 if ( $name =~ /(.+)\.\w+$/ );
	}
	closedir( RAID_DIR );
    }
}

sub get_raid_dirs {
    my ( $r_files ) = @_;

    my %dirs = ();
    foreach my $f ( @$r_files ) {
	if ( $f =~ m:^(\d\d)/: ) {
	    $dirs{$1} = 1;
	}
    }
    return sort keys %dirs;
}

sub get_list_of_tables {
    my ( $db ) = @_;

    # "use database" cannot cope with database names containing spaces
    # so create a new connection 

    my $dbh = DBI->connect("dbi:mysql:${db}${dsn};mysql_read_default_group=mysqlhotcopy",
			    $opt{user}, $opt{password},
    {
	RaiseError => 1,
	PrintError => 0,
	AutoCommit => 1,
    });

    my @dbh_tables = eval { $dbh->tables() };

    ## Remove quotes around table names
    my $quote = $dbh->get_info(29); # SQL_IDENTIFIER_QUOTE_CHAR
    if ($quote) {
      foreach (@dbh_tables) {
        s/^$quote(.*)$quote$/$1/;
        s/$quote$quote/$quote/g;
      }
    }

    $dbh->disconnect();
    return @dbh_tables;
}

sub quote_names {
  my ( $name ) = @_;
  # given a db.table name, add quotes

  my ($db, $table, @cruft) = split( /\./, $name );
  die "Invalid db.table name '$name'" if (@cruft || !defined $db || !defined $table );

  # Earlier versions of DBD return table name non-quoted,
  # such as DBD-2.1012 and the newer ones, such as DBD-2.9002
  # returns it quoted. Let's have a support for both.
  $table=~ s/\`//g;
  return "`$db`.`$table`";
}

__END__

=head1 DESCRIPTION

mysqlhotcopy is designed to make stable copies of live MySQL databases.

Here "live" means that the database server is running and the database
may be in active use. And "stable" means that the copy will not have
any corruptions that could occur if the table files were simply copied
without first being locked and flushed from within the server.

=head1 OPTIONS

=over 4

=item --checkpoint checkpoint-table

As each database is copied, an entry is written to the specified
checkpoint-table.  This has the happy side-effect of updating the
MySQL update-log (if it is switched on) giving a good indication of
where roll-forward should begin for backup+rollforward schemes.

The name of the checkpoint table should be supplied in database.table format.
The checkpoint-table must contain at least the following fields:

=over 4

  time_stamp timestamp not null
  src varchar(32)
  dest varchar(60)
  msg varchar(255)

=back

=item --record_log_pos log-pos-table

Just before the database files are copied, update the record in the
log-pos-table from the values returned from "show master status" and
"show slave status". The master status values are stored in the
log_file and log_pos columns, and establish the position in the binary
logs that any slaves of this host should adopt if initialised from
this dump.  The slave status values are stored in master_host,
master_log_file, and master_log_pos, and these are useful if the host
performing the dump is a slave and other sibling slaves are to be
initialised from this dump.

The name of the log-pos table should be supplied in database.table format.
A sample log-pos table definition:

=over 4

CREATE TABLE log_pos (
  host            varchar(60) NOT null,
  time_stamp      timestamp(14) NOT NULL,
  log_file        varchar(32) default NULL,
  log_pos         int(11)     default NULL,
  master_host     varchar(60) NULL,
  master_log_file varchar(32) NULL,
  master_log_pos  int NULL,

  PRIMARY KEY  (host) 
);

=back


=item --suffix suffix

Each database is copied back into the originating datadir under
a new name. The new name is the original name with the suffix
appended. 

If only a single db_name is supplied and the --suffix flag is not
supplied, then "--suffix=_copy" is assumed.

=item --allowold

Move any existing version of the destination to a backup directory for
the duration of the copy. If the copy successfully completes, the backup 
directory is deleted - unless the --keepold flag is set.  If the copy fails,
the backup directory is restored.

The backup directory name is the original name with "_old" appended.
Any existing versions of the backup directory are deleted.

=item --keepold

Behaves as for the --allowold, with the additional feature 
of keeping the backup directory after the copy successfully completes.

=item --addtodest

Don't rename target directory if it already exists, just add the
copied files into it.

This is most useful when backing up a database with many large
tables and you don't want to have all the tables locked for the
whole duration.

In this situation, I<if> you are happy for groups of tables to be
backed up separately (and thus possibly not be logically consistant
with one another) then you can run mysqlhotcopy several times on
the same database each with different db_name./table_regex/.
All but the first should use the --addtodest option so the tables
all end up in the same directory.

=item --flushlog

Rotate the log files by executing "FLUSH LOGS" after all tables are
locked, and before they are copied.

=item --resetmaster

Reset the bin-log by executing "RESET MASTER" after all tables are
locked, and before they are copied. Useful if you are recovering a
slave in a replication setup.

=item --resetslave

Reset the master.info by executing "RESET SLAVE" after all tables are
locked, and before they are copied. Useful if you are recovering a
server in a mutual replication setup.

=item --regexp pattern

Copy all databases with names matching the pattern

=item --regexp /pattern1/./pattern2/

Copy all tables with names matching pattern2 from all databases with
names matching pattern1. For example, to select all tables which
names begin with 'bar' from all databases which names end with 'foo':

   mysqlhotcopy --indices --method=cp --regexp /foo$/./^bar/

=item db_name./pattern/

Copy only tables matching pattern. Shell metacharacters ( (, ), |, !,
etc.) have to be escaped (e.g. \). For example, to select all tables
in database db1 whose names begin with 'foo' or 'bar':

    mysqlhotcopy --indices --method=cp db1./^\(foo\|bar\)/

=item db_name./~pattern/

Copy only tables not matching pattern. For example, to copy tables
that do not begin with foo nor bar:

    mysqlhotcopy --indices --method=cp db1./~^\(foo\|bar\)/

=item -?, --help

Display helpscreen and exit

=item -u, --user=#         

user for database login if not current user

=item -p, --password=#     

password to use when connecting to the server. Note that you are strongly
encouraged *not* to use this option as every user would be able to see the
password in the process list. Instead use the '[mysqlhotcopy]' section in
one of the config files, normally /etc/my.cnf or your personal ~/.my.cnf.
(See the chapter 'my.cnf Option Files' in the manual)

=item -h, -h, --host=#

Hostname for local server when connecting over TCP/IP.  By specifying this
different from 'localhost' will trigger mysqlhotcopy to use TCP/IP connection.

=item -P, --port=#         

port to use when connecting to MySQL server with TCP/IP.  This is only used
when using the --host option.

=item -S, --socket=#         

UNIX domain socket to use when connecting to local server

=item  --noindices          

Don\'t include index files in copy. Only up to the first 2048 bytes
are copied;  You can restore the indexes with isamchk -r or myisamchk -r
on the backup.

=item  --method=#           

method for copy (only "cp" currently supported). Alpha support for
"scp" was added in November 2000. Your experience with the scp method
will vary with your ability to understand how scp works. 'man scp'
and 'man ssh' are your friends.

The destination directory _must exist_ on the target machine using the
scp method. --keepold and --allowold are meaningless with scp.
Liberal use of the --debug option will help you figure out what\'s
really going on when you do an scp.

Note that using scp will lock your tables for a _long_ time unless
your network connection is _fast_. If this is unacceptable to you,
use the 'cp' method to copy the tables to some temporary area and then
scp or rsync the files at your leisure.

=item -q, --quiet              

be silent except for errors

=item  --debug

Debug messages are displayed 

=item -n, --dryrun

Display commands without actually doing them

=back

=head1 WARRANTY

This software is free and comes without warranty of any kind. You
should never trust backup software without studying the code yourself.
Study the code inside this script and only rely on it if I<you> believe
that it does the right thing for you.

Patches adding bug fixes, documentation and new features are welcome.
Please send these to [e-mail].

=head1 TO DO

Extend the individual table copy to allow multiple subsets of tables
to be specified on the command line:

  mysqlhotcopy db newdb  t1 t2 /^foo_/ : t3 /^bar_/ : +

where ":" delimits the subsets, the /^foo_/ indicates all tables
with names begining with "foo_" and the "+" indicates all tables
not copied by the previous subsets.

newdb is either another not existing database or a full path to a directory
where we can create a directory 'db'

Add option to lock each table in turn for people who don\'t need
cross-table integrity.

Add option to FLUSH STATUS just before UNLOCK TABLES.

Add support for other copy methods (eg tar to single file?).

Add support for forthcoming MySQL ``RAID'' table subdirectory layouts.

=head1 AUTHOR

Tim Bunce

Martin Waite - added checkpoint, flushlog, regexp and dryrun options
               Fixed cleanup of targets when hotcopy fails. 
	       Added --record_log_pos.
               RAID tables are now copied (don't know if this works over scp).

Ralph Corderoy - added synonyms for commands

Scott Wiersdorf - added table regex and scp support

Monty - working --noindex (copy only first 2048 bytes of index file)
        Fixes for --method=scp

Ask Bjoern Hansen - Cleanup code to fix a few bugs and enable -w again.

Emil S. Hansen - Added resetslave and resetmaster.

Jeremy D. Zawodny - Removed depricated DBI calls.  Fixed bug which
resulted in nothing being copied when a regexp was specified but no
database name(s).

Martin Waite - Fix to handle database name that contains space.

Paul DuBois - Remove end '/' from directory names