Can't start/stop mysql service

mysqladmin shutdown

should work to shutdown the server.

I see two likely possibilities:

1. MySQL has a problem and is refusing to shut down for some reason.
2. The previous admin did something strange. Either modified the init.d script or didn't bother using the Debian packages at all to install MySQL.

What does dpkg --list mysql\* say?

What does /var/log/mysql.err say? Or the other mysql logs?

EDIT:

So mysqladmin shutdown worked?

According to that, the mysql-server package is installed (mysql-server-5.0; the mysql-server package is probably just a stub). So they may have installed over it? Running debsums mysql-server-5.0 might tell you more. dpkg --listfiles mysql-server-5.0 could help, too...

What's actually in /etc/init.d/mysql? I haven't checked that specific version of the package, but it should try to use mysqladmin shutdown ... Maybe you're lucky and they only broke that...

Why this is happening

This is a common problem if you do a mysql import and overwrite the mysql database itself, such as when you might be restoring from a mysqldump -A backup.

This is a good thing: you probably want to back up all your mysql users, permissions, etc -- but it can wreak havoc with things like the debian-sys-maint user used to cleanly shutdown mysql.

Although this new database will possibly change both the root password and the debian-sys-maint password, of course it won't automatically change the expected debian-sys-maint password in /etc/mysql/debian.cnf. In fact, unless you also backed up that file, you probably don't even know what that password is anymore!

Resetting the mysql root password (optional)

First things first. If the mysql root password was different between old and new servers, you can use mysqladmin to fix it:

mysql -p -u root password 'newpassword'

However, when you apt-get installed mysql-server, it probably prompted you for the new mysql root password and you probably used the same one that you were using from before.

Fix the debian sys maint password.

So now look up the debian sys maint password that debian created for you when you installed it on the new server. (You need sudo because this should be a highly protected file.)

sudo cat /etc/mysql/debian.cnf

Now, log in to mysql using the root password you set above:

mysql -p -u root   # use your new password when prompted

Reset the password for the debian-sys-maint user and don't forget to flush privileges:

>  SET PASSWORD FOR 'debian-sys-maint'@'localhost' = PASSWORD('samepassword');
>  FLUSH PRIVILEGES;
>  QUIT

Test to be sure it works:

sudo /etc/init.d/mysql restart

Quick tip

If you ever need to reset the root password for the server without having to bring down the server, this user account has the authority to do it -- just cat the debian.cnf file and log in with that user. N.B. Protect this user account just like root.

pkill mysql

will definitely work

2 more hints:

sh -x /etc/init.d/mysql restart

This will show you the commands executed by the init script.

install the package debsums, and you can test which packages were modified (verify is also available for RPM, but IMHO works better).

Assuming the package is somewhat strange the problem could be the pid file. I suspect that the new packages or compiled install did not create /var/run/mysql/ or whatever is standard on Debian for the pid file to be written to or the init script is looking for the mysqld.pid file in another place. If you can fix the init/pid file mismatch things should probably work.

The mysql shutdown script uses the debian-sys-maint user to run 'mysqladmin shutdown', by reading the password for the user from /etc/mysql/debian.cnf. You should check that this file exists, and that you can run mysqladmin shutdown as this user.

You could technically end it with:

pkill -9 mysqld

But you might lose data?

You might be better off asking someone at http://www.serverfault.com

Using "pkill mysql" will also likely lose you data, particularly if invoked as "pkill -9" :(

I'd also recommend using 'sh -x' to see what the problem with the init script might be, and you can also go peek in the error logs for MySQL (/var/log/mysql or /var/lib/mysql, depending on config) to see if it is stuck on a really long running query or something and thus not willing to quit gracefully quite yet.

To follow up the comment on your question, I'll write down a full answer:

The issue is that the default socket is /tmp/mysql.sock with MySQL source, and /var/run/mysqld/mysqld.sock with Debian binaries.

The solution is to fix the socket path in /etc/mysql/debian.cnf, by providing the good socket=. Or by keeping it, but then change the one in /etc/mysql/my.cnf.

Here's how I found this out: in /etc/init.d/mysql when there the «failed» message, you have this line called:

echo -e "$ps_alive processes alive and '$MYADMIN ping' resulted in\n$ping_output\n" | $ERR_LOGGER -p daemon.debug

This pointed me to $MYADMIN ping, which is mysqladmin --defaults-file=/etc/mysql/debian.cnf ping. Running this very command ends on:

/usr/bin/mysqladmin: connect to server at 'localhost' failed

error: 'Can't connect to local MySQL server through socket '/var/run/mysqld/mysqld.sock' (2)'

Check that mysqld is running and that the socket: '/var/run/mysqld/mysqld.sock' exists!

So I had a look at /etc/mysql/debian.cnf and found out that was the bad socket.

use the following command :

$mysqladmin shutdown

this should be available in /usr/bin directory in your case.