The mysqlpump Utility

mysqlpumpIn this blog, we’ll look at the
mysqlpump utility.

mysqlpump is a utility that performs logical backups (which means backing up your data as SQL statements instead of a raw copy of data files). It was added in MySQL Server version 5.7.8, and can be used to dump a database or a set of databases to a file and then loaded on another SQL server (not necessarily a MySQL server).

Its usage is similar to
mysqldump, but it includes a new set of features. Many of the options are the same, but it was written from scratch to avoid being limited to
mysqldump compatibility.

The Main Features Include:

  • To make the dump process faster, it allows parallel processing of databases and objects within databases.
  • There are more options to customize your dumps and choose which databases and objects to dump (tables, stored programs, user accounts), using the
    include* and 
    exclude* parameters.
  • User accounts can be dumped now as
    CREATE USER and
    GRANT statements, instead of inserting directly to the MySQL system database.
  • Information between the client and the server can be compressed using the
    compress option. This feature is very useful for remote backups, as it saves bandwidth and transfer time. You can also compress the output file using
    compressoutput, which supports ZLIB and LZ4 compression algorithms.
  • It has an estimated progress indicator. This is really useful to check the current status of the dump process. You can see the total amount of rows dumped and the number of databases completed. It also reports an estimate of the total time to complete the dump.
  • Creation of secondary indexes for InnoDB tables happens after data load for shorter load times.

Exclude/Include:

This feature provides more control over customizing your dumps, and filter the data that you need. Using this feature, you can be more selective with the data you want to dump (databases, tables, triggers, events, routines, users) and save file size, process time and transferring time while copying/moving the file to another host.

Keep in mind that there are some options that are mutually exclusive: e.g., if you use the
alldatabases option, the
excludedatabases  parameter won’t take effect. By default,
mysqlpump will not dump the following databases unless you specify them using the
includedatabases option:
INFORMATION_SCHEMA,
performance_schema,
ndbinfo  and
sys.

Values for these options need to be declared by comma-separated listing. Using a “%” as a value for any of the exclude/include options acts as a wildcard. For example, you can dump all databases starting with “t” and “p” by adding the option
includedatabases=t%,p%  to the command line.

For users, routines, triggers and events,
mysqlpump has
include* and
exclude* options with similar usage. Some specific notes:

  • Triggers are dumped by default, but you can also filter them using the
    includetriggers/
    excludetriggers options
  • Routines and events are not dumped by default, and need to be specified in the command line with
    routines and
    events, or the corresponding
    include and 
    exclude options
  • Keep in mind that if a stored procedure and a function have the same name, then include/exclude applies to both

Parallel Processing:

This feature allows you to process several databases, and tables within the databases, in parallel. By default,
mysqlpump uses one processing queue with two threads. You can increase the number of threads for this default queue with
defaultparallelism. Unless you create additional queues, all the databases and/or tables you elect to dump go through the default queue.

To create additional queues you can use the 
parallelschemas option, which takes two parameters: the number of threads for the queue and the sub-set of databases this queue processes.  As an example, you could run:

so that schemas c, d, e, f, g and h are processed by the default queue (which uses three threads), and then tables from schemas a and b are processed by a separate queue (that uses four threads). Database names should be included as a comma-separated list:

User Accounts:

User accounts can be dumped using this tool. Here’s a comparison of our Percona Tool
ptshowgrants versus
mysqlpump to check their differences.

By default,
mysqlpump doesn’t dump user account definitions (even while dumping the MySQL database). To include user accounts on the dump, you must specify the
users option.

Here’s an example on how use
mysqlpump to get only user accounts dumped to a file:

As you can see, above the tool makes sure the session uses known values for timezone and character sets. This won’t affect users, it’s part of the dump process to ensure correctness while restoring on the destination.

Comparing it with
ptshowgrants from Percona Toolkit, we can see that 
mysqlpump dumps the
CREATE USER  information as well. The statements produced by
mysqlpump are the right thing to run to recreate users (and should be the preferred method), especially because of the
sql_mode NO_AUTO_CREATE_USERS. If enabled, it renders
ptshowgrants useless.

Here’s an example of
ptshowgrants usage:

Some Miscellaneous Notes:

  • One of the differences with
    mysqldump is that
    mysqlpump adds 
    CREATE DATABASE statements to the dump by default, unless specified with the
    nocreatedb option.

    • There’s an important difference on the dump process that is closely related: it includes the database name while adding the
      CREATE TABLE statement. This causes a problem when trying to use the tool to create a duplicate.

via Planet MySQL
The mysqlpump Utility