mk-parallel-restore(1) Load files into MySQL in parallel.


mk-parallel-restore /path/to/files
mk-parallel-restore --tab /path/to/files


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.

mk-parallel-restore is not a backup program! It is only for fast data imports, for purposes such as loading test data into a system quickly. Do not use mk-parallel-restore for backups. mk-parallel-restore inserts data unless you use the ``--dry-run'' option.

At the time of this release, there is a bug that prevents huge statements from being printed when an error is encountered and a bug applying ``--[no]foreign-key-checks'' when truncating tables.

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: <>.

See also ``BUGS'' for more information on filing bugs and getting help.


mk-parallel-restore is a way to load SQL or delimited-file dumps into MySQL in parallel at high speed. It is especially designed for restoring files dumped by mk-parallel-dump. It automatically detects whether a file contains SQL or delimited data from the filename extension, and either shells out to "mysql" or executes "LOAD DATA INFILE" with the file. On UNIX-like systems, it will even make a FIFO to decompress gzipped files for "LOAD DATA INFILE".

By default it discovers all files in the directory you specify on the command line. It uses the file's parent directory as the database name and the file's name (up to the first dot) as the table name. It can deal with files named like the following:


It is also happy with files that look like this, where "EXT" is one of the extensions just listed.


By default, it loads "SQL" files first, if they exist, then loads "CSV" or "TXT" files next, in order of the numbers in the filename extension as just shown. This makes it easy for you to reload a table's definition followed by its data, in case you dumped them into separate files (as happens with "mysqldump"'s "--tab" option). See mk-parallel-dump for details on how data is dumped.

Exit status is 0 if everything went well, 1 if any files failed, and any other value indicates an internal error.


Output depends on verbosity. When ``--dry-run'' is given, output includes commands that would be executed.

When ``--verbose'' is 0, there is normally no output unless there's an error.

When ``--verbose'' is 1, there is one line of output for the entire job, showing how many tables were processed, how many files were loaded with what status, how much time elapsed, and how much time the parallel load jobs added up to. If any files were skipped, the filenames are printed to the output.

When ``--verbose'' is 2, there's one line of output per table, showing extra data such as how many threads were running when each table finished loading:

  sakila   language             2  0.07      0       2
  sakila   film_actor           2  0.07      0       2
  sakila   actor                2  0.06      0       2
  sakila   payment              2  0.07      0       2
  sakila   transport_backup     2  0.05      0       2
  sakila   country              2  0.08      0       2
  sakila   film                 2  0.05      0       2
  sakila   rental               2  0.07      0       2


User-contributed benchmarks are welcome. See <> for one user's experiences.


Prompt for a password when connecting to MySQL.
default: yes

Treat chunks as atomic when resuming restore.

By default "mk-parallel-restore" resumes restoration from the first chunk that is missing all its rows. For dumps of transactionally-safe tables (InnoDB), it cannot happen that a chunk is only partially restored. Therefore, restoring from the first missing chunk is safe.

However, for dumps of non-transactionally safe tables, it is possible that a chunk can be only partially restored. In such cases, the chunk will wrongly appear to be fully restored. Therefore, you must specify "--no-atomic-resume" so that the partially restored chunk is fully restored.

type: string

Directory where FIFO files will be created.

default: yes

Restore the biggest tables first for highest concurrency.

default: yes

Enable binary logging ("SET SQL_LOG_BIN=1").

Restore operations are replicated by default (SQL_LOG_BIN=1) except for ``--tab'' restores which are not replicated by default (SQL_LOG_BIN=0). This prevents large loads from being logged to the server's binary log.

The value given on the command line overrides the defaults. Therefore, specifying "--bin-log" with ``--tab'' will allow the ``--tab'' restore to replicate.

type: int

Set bulk_insert_buffer_size before each "LOAD DATA INFILE".

Has no effect without ``--tab''.

short form: -A; type: string; default: BINARY

Sets the connection, database, and "LOAD DATA INFILE" character set.

The default is "BINARY", which is the safest value to use for "LOAD DATA INFILE". Has no effect without ``--tab''.

default: yes

Commit after each file.

type: Array

Read this comma-separated list of config files; if specified, this must be the first option on the command line.

Create databases if they don't exist.
default: yes

Create tables.

See also ``--[no]drop-tables''.

Files are in CSV format (implies ``--tab'').

Changes ``--tab'' options so the following "LOAD DATA INFILE" statement is used:

   LOAD DATA INFILE <filename> INTO TABLE <table>
short form: -D; type: string

Load all files into this database.

Overrides the database which is normally specified by the directory in which the files live. Does not specify a default database for the connection.

short form: -d; type: hash

Restore only this comma-separated list of databases.

type: string

Restore only databases whose names match this regex.

type: string; default: gzip -d -c

Command used to decompress and print .gz files to STDOUT (like zcat).

short form: -F; type: string

Only read mysql options from the given file. You must give an absolute pathname.

default: yes

Execute "ALTER TABLE DISABLE KEYS" before each MyISAM table.

This option only works with MyISAM tables.

default: yes

Execute "DROP TABLE IF EXISTS" before creating each table.

Print commands instead of executing them.
Do InnoDB plugin fast index creation by restoring secondary indexes after data.

This option only works with InnoDB tables and the InnoDB plugin.

default: yes

Stream files into a FIFO for ``--tab''.

Load compressed tab-separated files by piping them into a FIFO and using the FIFO with "LOAD DATA INFILE", instead of by decompressing the files on disk. Sets ``--umask'' to 0.

default: yes


Show help and exit.
short form: -h; type: string

Connect to host.

Adds the "IGNORE" modifier to "LOAD DATA INFILE".
type: Hash

Ignore this comma-separated list of databases.

type: Hash

Ignore this comma-separated list of table names.

Table names may be qualified with the database name.

Uses the "LOCAL" option to "LOAD DATA INFILE".

If you enable this option, the files are read locally by the client library, not by the server.

Lock tables before "LOAD DATA INFILE".
default: yes


Restore only to empty databases.

By default mk-parallel-restore will restore tables into a database so long as it exists (or is created by ``--create-databases''). This option is a safety feature that prevents any tables from being restored into a database that already has tables even if those tables are the same ones being restored. If you specify this option, every database must have zero tables.

This implicitly disables ``--[no]resume''. ``--create-databases'' will work if the database doesn't already exist and it creates it.

The databases are checked after all filters (``--databases'', etc.)

short form: -p; type: string

Password to use when connecting.

type: string

Create the given PID file. The file contains the process ID of the script. The PID file is removed when the script exits. Before starting, the script checks if the PID file already exists. If it does not, then the script creates and writes its own PID to it. If it does, then the script checks the following: if the file contains a PID and a process is running with that PID, then the script dies; or, if there is no process running with that PID, then the script overwrites the file with its own PID and starts; else, if the file contains no PID, then the script dies.

short form: -P; type: int

Port number to use for connection.

Display progress messages.

Progress is displayed each time a table finishes loading. Progress is calculated by measuring the size of each file to be loaded, and assuming all bytes are created equal. The output is the completed and total size, the percent completed, estimated time remaining, and estimated completion time.

short form: -q

Sets ``--verbose'' to 0.

Adds the "REPLACE" modifier to "LOAD DATA INFILE".
default: yes

Resume the restore from a previously incomplete restore.

By default, "mk-parallel-restore" checks each table's chunks for existing rows and restores only from the point where a previous restore stopped. Specify --no-resume to disable restore resumption and fully restores every table.

Restore resumption does not work with tab-separated files or dumps that were not chunked.

type: string; default: wait_timeout=10000

Set these MySQL variables. Immediately after connecting to MySQL, this string will be appended to SET and executed.

short form: -S; type: string

Socket file to use for connection.

Load tab-separated files with "LOAD DATA INFILE".

This is similar to what "mysqlimport" does, but more flexible.

The following options are enabled unless they are specifically disabled on the command line:


And the following options are disabled ("--no-bin-log", etc.) unless they are specifically enabled on the command line:

short form: -t; type: hash

Restore only this comma-separated list of table names.

Table names may be qualified with the database name.

type: string

Restore only tables whose names match this regex.

type: int; default: 2

Specifies the number of parallel processes to run.

The default is 2 (this is mk-parallel-restore after all --- 1 is not parallel). On GNU/Linux machines, the default is the number of times 'processor' appears in /proc/cpuinfo. On Windows, the default is read from the environment. In any case, the default is at least 2, even when there's only a single processor.


This will delete all rows from a table before loading the first tab-delimited file into it.

type: string

Set the program's "umask" to this octal value.

This is useful when you want created files (such as FIFO files) to be readable or writable by other users (for example, the MySQL server itself).

default: yes


short form: -u; type: string

User for login if not current user.

short form: -v; cumulative: yes; default: 1

Verbosity; can specify multiple times.

Repeatedly specifying it increments the verbosity. Default is 1 if not specified. See ``OUTPUT''.

Show version and exit.
short form: -w; type: time; default: 5m

Wait limit when server is down.

If the MySQL server crashes during loading, waits until the server comes back and then continues with the rest of the files. "mk-parallel-restore" will check the server every second until this time is exhausted, at which point it will give up and exit.


These DSN options are used to create a DSN. Each option is given like "option=value". The options are case-sensitive, so P and p are not the same option. There cannot be whitespace before or after the "=" and if the value contains whitespace it must be quoted. DSN options are comma-separated. See the maatkit manpage for full details.
  • A

    dsn: charset; copy: yes

    Default character set.

  • D

    dsn: database; copy: yes

    Default database.

  • F

    dsn: mysql_read_default_file; copy: yes

    Only read default options from the given file

  • h

    dsn: host; copy: yes

    Connect to host.

  • p

    dsn: password; copy: yes

    Password to use when connecting.

  • P

    dsn: port; copy: yes

    Port number to use for connection.

  • S

    dsn: mysql_socket; copy: yes

    Socket file to use for connection.

  • u

    dsn: user; copy: yes

    User for login if not current user.


You can download Maatkit from Google Code at <>, or you can get any of the tools easily with a command like the following:


Where "toolname" can be replaced with the name (or fragment of a name) of any of the Maatkit tools. Once downloaded, they're ready to run; no installation is needed. The first URL gets the latest released version of the tool, and the second gets the latest trunk code from Subversion.


The environment variable "MKDEBUG" enables verbose debugging output in all of the Maatkit tools:

   MKDEBUG=1 mk-....


You need Perl, DBI, DBD::mysql, and some core packages that ought to be installed in any reasonably new version of Perl.


For list of known bugs see <>.

Please use Google Code Issues and Groups to report bugs or request support: <>. You can also join #maatkit on Freenode to discuss Maatkit.

Please include the complete command-line used to reproduce the problem you are seeing, the version of all MySQL servers involved, the complete output of the tool when run with ``--version'', and if possible, debugging output produced by running with the "MKDEBUG=1" environment variable.


This program is copyright 2007-2010 Baron Schwartz. Feedback and improvements are welcome.


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.


Baron Schwartz


This tool is part of Maatkit, a toolkit for power users of MySQL. Maatkit was created by Baron Schwartz; Baron and Daniel Nichter are the primary code contributors. Both are employed by Percona. Financial support for Maatkit development is primarily provided by Percona and its clients.


This manual page documents Ver 1.0.23 Distrib 6652 $Revision: 6550 $.