Test::DBIx::Class::SchemaManager::Trait::Testmysqld - deploy to a test mysql instance


This trait uses Test::mysqld to auto create a test instance of mysql in a temporary area. This way you can test against mysql without having to create a test database, users, etc. Mysql needs to be installed (but doesn't need to be running) as well as DBD::mysql. You need to install these yourself.

Please review Test::mysqld for help if you get stuck.


This trait supports all the existing features but adds some additional options you can put into your inlined configuration files. These following additional configuration options basically map to the options supported by Test::mysqld and the docs are adapted shamelessly from that module.

For the most part, if you have mysql installed in a normal, findable manner you should be able to leave all these options blank.


Returns directory under which the mysqld instance is being created. If you leave this unset we automatically create a place in the temporary directory and then clean it up later. Unless you plan to roundtrip to the same database a lot you can just leave this blank.

Please note if you set this to a particular area, we will delete it unless you specifically use the 'keep_db' option. SO be care where you point it!

Here's an example use. I often want the test database setup in my local testing directory, that makes it easy for me to examine the logs, etc. I do:

        BASE_DIR=t/tmp KEEP_DB=1 prove -lv t/my-mysql-test.t

Now I can roundtrip the test as often as I want and in between tests I can review the logs, start the database manually and login (see the 'keep_db' section below for an example of how to do this). Next time I run the tests the framework will automatically clean it up and rest the schema for testing.

You may need to do this if you are stuck on a shared host and can't write anything to /tmp. Remember, you can also put the 'base_dir' option into configuration instead of having to type it into the commandline each time!


A hashref containing the list of name=value pairs to be written into "my.cnf", which is the primary configuration file for the mysql instance. Again, unless you have some specific needs you can leave this empty, since we set the few things most needed to get a server running. You will need to review the documentation on the Mysql website for options related to this.


This is the port that will be used when starting mysqld. We check that this port is available for use before starting mysqld. If it is not available we increment by 1 and try again. We use the first free port found.

By default this is a random port between 8000 and 10000. The randomness is an attempt to avoid race condition issues when running tests in parallel, between checking the availability of a port and actually starting the server. Spreading the "first port" numbers used greatly reduces the chance of these issues occuring.

mysql_install_db or mysqld

If your mysqld is not in the $PATH you might need to specify the location to one of there binaries. If you have a normal mysql setup this should not be a problem and you can leave this blank.

For example, I often use MySQL::Sandbox to setup various versions of mysql in my local user directory, particularly if I am on a shared host, or in the case where I don't want mysql installed globally. Personally I think this is really your safest option (and there will probably be a trait based on this in the future)


The following are notes regarding the way this trait alters or extends the core functionality as described in the basic documentation.


Since it is always safe to use the 'force_drop_table' option with mysql, we set the default to true. We recommend you leave it this way, particularly if you want to 'roundtrip' the same test database.


If you use the 'keep_db' option, this will preserve the temporarily created database files, however it will not prevent Test::mysqld from stopping the database when you are finished. This is a safety measure, since if we didn't stop a test generated database instance automatically, you could easily end up with many databases running at once, and that could bring your server or testing box to a halt.

If you use the 'keep_db' option and want to start and log into the test generated database instance, you can start the database by noticing the diagnostic output that should be generated at the top of your test. It will look similar to:

        # Starting mysqld with: /usr/local/mysql/bin/mysqld --defaults-file=/tmp/KHKfJf0Yf6/etc/my.cnf --user=root

If you have specified the base_dir to use, this output will not be displayed by default. You can force it's display by setting tdbic_debug to true. eg.

        TDBIC_DEBUG=1 BASE_DIR=t/tmp KEEP_DB=1 prove -lv t/my-mysql-test.t

You can then start the database instance yourself with something like:

        /usr/local/mysql/bin/mysqld --defaults-file=/tmp/WT0P0VutAe/etc/my.cnf \
        --user=root &
        [1] 3447
        090827 15:06:16  InnoDB: Started; log sequence number 0 78863
        090827 15:06:16 [Note] Event Scheduler: Loaded 0 events
        090827 15:06:16 [Note] /usr/local/mysql/bin/mysqld: ready for connections.
        Version: '5.1.37'  socket: '/tmp/WT0P0VutAe/tmp/mysql.sock'  port: 0  MySQL Community Server (GPL)

There will be some additional output to the term and then the server will go into the background. If you don't like the extra output, you can just redirect it all to /dev/null or whatever is similar for your OS.

You can now log into the test generated database instance with:

        mysql --socket=/tmp/WT0P0VutAe/tmp/mysql.sock -u root test

You may need to specify the full path to 'mysql' if it's not in your search $PATH.

When you are finished you can then kill the process. In this case our reported process id is '3447'

        kill 3447

And then you might wish to 'tidy' up temp

        rm -rf /tmp/WT0P0VutAe

All the above assume you are on a unix or unixlike system. Would welcome document patches for how to do all the above on windows.


John Napiorkowski <>


Copyright 2009, John Napiorkowski <>

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.