We continue our MySQL server setup for CentOS by looking at configuration options to try and ensure the server doesn’t just run, but runs smoothly.
Beyond the defaults
In the previous article we covered a basic MySQL server setup on Red Hat Linux. We set the root password, created a database, and created a user for the database. Now let's look at MySQL in a little more detail so we can tweak its configuration and be ready in case something goes wrong.
Finding the config files
By default you'll find MySQL's configuration file at:
If it's not there, however, you can ask mysqld where it looks for its config. Run the command:
/usr/libexec/mysqld --help --verbose
You'll get a flood of text back. The first part describes the options you can send to the server when you launch it. The second part is all the configuration stuff that was set when the server was compiled.
What we're looking for shows up near the start of the output. Find a couple lines that look like:
Default options are read from the following files in the given order: /etc/my.cnf /etc/mysql/my.cnf /usr/etc/my.cnf ~/.my.cnf
And there we are. The server works down that list until it finds a configuration file.
With the location in hand, open the my.cnf file and have a look inside.
Any lines starting with "#" are comments, and they mostly document what the different settings are for. They're good to read through. You'll find details like the location of log files and where the database files are kept.
There are lines in the config file that just contain a word in square brackets, like "[client]" or "[mysqld]". Those are "config groups" and they tell the programs that read the configuration file which parts they should pay attention to.
See, while we've been focusing on the server part of MySQL, it's technically a collection of tools. That includes the server (mysqld), the client (mysql), and some other tools we'll talk about in a bit. Those programs look in my.cnf to see how they should behave.
There's a bit more to it, but basically: The "client" config section controls the mysql client, and the "mysqld" section controls the server config.
If something does go wrong the best place to start troubleshooting any program is its logs. By default MySQL stores its log files in the directory:
You may need to use sudo to get a listing of the files in that directory.
If you don't find the MySQL logs in the default directory you'll need to check MySQL's config. Look in the my.cnf file and look for a "log_error" line, as in:
log_error = /var/log/mysql/error.log
If you don't see a line like that, create one in the "mysqld" section so MySQL will use its own error log. We recommend using the location in the example, creating the "/var/log/mysql" directory if it doesn't already exist. Then restart MySQL to make the change.
Make sure the log directory you choose can be written to by the user controlling the MySQL process (usually "mysql"). The user running the process will be defined in the "user" config value for mysqld in my.cnf.
There might be a "port" setting under both the client and server config sections. The port under the server section controls what port the server will listen to. By default that's 3306 but you can change it to anything you'd like.
The port in the client section tells the client what port to connect to by default. You'll generally want the two port settings to match up.
If you don't see the port entries in the config file that just means they're using the default. If you want to change the port you would add the lines in the appropriate categories:
[client] port = 3306 [mysqld] port = 3306
The other network setting to look for is the "bind-address" value. That usually gets set to the address for localhost, 127.0.0.1. By binding to localhost the server makes sure no one can connect to it from outside the local machine.
If you're running your MySQL server on a different machine from your application you'll want to bind to a remotely-accessible address instead of localhost. Change the bind-address setting to match your public IP address (or, better, a backend IP address on a network that fewer machines can access).
If you don't see a "bind-address" entry you should put one into the "mysqld" category to help control access to the server:
[mysqld] bind-address = 127.0.0.1
Remember to account for the client's hostname when you set up your database users and to poke a hole in your firewall if you're running iptables.
mysqld and mysqld_safe
Behind the scenes there are actually two versions of the MySQL server, "mysqld" and "mysqld_safe". Both read the same config sections. The main difference is that mysqld_safe launches with a few more safety features enabled to make it easier to recover from a crash or other problem.
Both mysqld and mysqld_safe will read config entries in the "mysqld" section. If you include a "mysqld_safe" section, then only mysqld_safe will read those values in.
By default the mysql service launches "mysqld_safe". That's a good thing, and you should only look to change that if you really know what you're doing.
The mysqladmin tool lets you perform some administrative functions from the command line. We won't talk much about it here because we're just trying to get you up and running with enough basics to get by. It's worth looking at the tool in more depth later to see what it can do, particularly if you need to build scripts that perform functions like checking the status of the server or creating and dropping databases.
You have a few options when it comes to making backups of your databases apart from the usual "back up the whole machine" approach. The main two are copying the database files and using mysqldump.
By default MySQL creates a directory for each database in its data directory:
Once you've found the data directory, hold off a moment before making a copy of it. When the database server is active it could be writing new values to tables at any time. That means if it writes to a table halfway through your copy some files will change and lead to a corrupt backup. Not a good thing if you're trying to plan for disaster recovery.
To make sure the database files are copied cleanly you can shut the MySQL server down entirely before the copy. That's safe but isn't always ideal.
Another approach you can take is to lock the database as read-only for the duration of the copy. Then when you're done, release the lock. That way your applications can still read data while you're backing up files.
Lock the databases to read-only by running, from the command line:
mysql -u root -p -e "FLUSH TABLES WITH READ LOCK;"
To unlock the database when you're done, run:
mysql -u root -p -e "UNLOCK TABLES;"
We're using a new option with the mysql client, "-e". That tells the client to run the query in quotes as if we'd entered it in the mysql shell proper.
Note that if you're setting these commands up in a script you can put the password in quotes right after "-p" with no space between the two, as in:
mysql -u root -p"password" -e "FLUSH TABLES WITH READ LOCK;" mysql -u root -p"password" -e "UNLOCK TABLES;"
Just make sure you set the permissions on that file to restrict read access. We don't want just anyone to be able to see that password.
Another approach to backing up your database is to use the "mysqldump" tool. Rather than copying the database files directly, mysqldump generates a text file that represents the database. By default the text file contains a list of SQL statements you would use to recreate the database, but you can also export the database in another format like CSV or XML. You can read the man page for mysqldump to see all its options.
The statements generated by mysqldump go straight to standard output. You'll want to specify a file to redirect the output to when you run it. For example:
mysqldump -u root -p demodb > dbbackup.sql
That command will tell mysqldump to recreate the "demodb" database in SQL statements and to write them to the file "dbbackup.sql". Note that the username and password options function the same as the mysql client, so you can include the password directly after "-p" in a script.
Restore from mysqldump
Restoring a mysqldumped database looks similar to what was used to create it, but we use plain old "mysql" instead of "mysqldump":
mysql -u root -p demodb < dbbackup.sql
We also change from a greater-than to a less-than sign. That switches the command from redirecting its output to telling it to read its input from the existing file. That input is sent to the "mysql" command, causing the mysqldumped instructions to recreate the database.
Note that by default the SQL statements generated would just add to existing database tables, not overwrite them. If you're restoring a backup over an existing database you should drop the database's tables first, or drop and recreate the database itself. You can change that behavior by using the "--add-drop-table" option with the command that creates the mysqldump. That causes mysqldump to add a command to the backup files it writes that will drop tables before recreating them.
The last concept we'll talk about here is that of the "database engine". The engine is the process that's churning away behind the scenes, writing to and reading data from files. You won't usually need to know anything other than that it's there, but sometimes you'll want to run an application that's been optimized for a particular database engine.
The engine type is set when a table is created. Tables are usually created by the application that's going to use them, which is why we aren't going to get into that syntax here.
To see the engine used by your database's tables you can run the following command in the MySQL shell:
SHOW TABLE STATUS FROM demodb;
Change "demodb" to the name of your database.
Choosing an engine
Ideally you won't need to choose an engine. If you're not very familiar with MySQL that's certainly the safest way to go - let the application do its thing, and if you're writing the application, use the default engine until you're more comfortable with your options.
If you have a database administrator, do whatever he or she says. They're smart people, they know what they're talking about.
The two database engines used most often with MySQL are "MyISAM" and "InnoDB". The default database engine for MySQL version 5.1 and earlier is MyISAM, while InnoDB is the default database engine starting with MySQL version 5.5.
Because MyISAM has been the default in MySQL for a while it's the most compatible choice of the two main engines. Certain types of searches perform better on MyISAM than InnoDB. Just because it's the older of the two doesn't mean it can't be the best for a given application type.
InnoDB is more fault-tolerant than MyISAM and handles crashes and recovery with a much smaller chance of database corruption. This is a good thing.
The main trouble with InnoDB is that for best performance it requires a lot of tweaking for your environment and access patterns. If you have a DBA that's no problem, but if you're a developer who just wants a database up and running for a test server you probably won't want to deal with tuning InnoDB.
It's possible you may be running an application that requires InnoDB, and if you're using MySQL 5.1 or earlier there might not be any settings already in the my.cnf config file. That can be a problem if you're running on a server that doesn't have an abundance of memory.
Some settings to get you started with InnoDB on a shared server with 256 megs of RAM are:
innodb_buffer_pool_size = 32M innodb_log_file_size = 8M innodb_thread_concurrency = 8 innodb_file_per_table
Add those to the [mysqld] section of the config file. Again, those are only rough guides - enough to get you running, but definitely not optimized. For that you'll probably want a DBA, or at least to experiment with incremental changes over time.
Now you should have MySQL configured for your environment, and might even have accounted for the database engine being used by your tables.
In the next article we'll look at some basic tasks you can run in the mysql shell to manipulate and observe your database at a high level.