Michael Widenius wrote:
>
> >>>>> "Plamen" == Plamen Petkov <plamendp@stripped> writes:
>
> Plamen> 1. I got the following error as an output from safe_mysqld:
> Plamen> ===
> Plamen> [~/mysql] UserName@sun00404(pts/3)
> Plamen> [5:41am]# bin/safe_mysqld
> Plamen> Starting mysqld daemon with databases from
> /usr/sites/bgstore.com/mysql/var
> Plamen> mysqld daemon ended
> Plamen> [~/mysql] UserName@sun00404(pts/3)
> Plamen> [5:42am]#
> Plamen> ===
>
> Plamen> 2. This is from <hostname>.err log file:
>
> Plamen> ===
> Plamen> ...
> Plamen> 990705 5:42:04 Can't start server: Bind on TCP/IP port: Address already
> in
> Plamen> use
> Plamen> 990705 5:42:04 Do you already have another mysqld server running on
> port:
> Plamen> 3306 ?
> Plamen> 990705 5:42:04 Aborting
> Plamen> ===
>
> Plamen> NOTE: nobody on this server is running another mysql on port 3306. I've
> Plamen> checked it out.
>
> Plamen> 3. netstat -a | grep mysql
> Plamen> f66c8158 stream-ord f697b090 0
> Plamen> /usr/sites/bgstore.com/mysql/mysql.sock
> Plamen> f66c8d58 stream-ord f676a9c0 0
> Plamen> /usr/sites/bgstore.com/mysql/mysql.sock
>
> Plamen> But there are NO socket files existing.
>
> Plamen> 4. netstat -a | grep 3306
>
> Plamen> *.3306 *.* 0 0 0 0 LISTEN
>
> Plamen> This is a result from 'my' mysql daemon. But I shot it down succesfully
> ??
> Plamen> I this is still here ?
>
> Plamen> Well, I am not a 'network guru'. How can I clean this mess ?
>
> Plamen> ---
> Plamen> Plamen D. Petkov, plamendp@stripped
> Plamen> ICQ 2214327
> Plamen> FBiS First Bulgarian Internet Store
> Plamen> http://www.bgstore.com
> Plamen> ---
>
> Hi!
>
> mysqld gets the above error comes from bind() unix call. I am not
> aware of any possible error source than that there is someone that is
> using this socket.
>
> Try starting mysqld directly (without safe_mysqld).
>
> libexec/mysqld
>
> Check also the default socket mysqld is trying to use with:
>
> libexec/mysqld --help
>
> and
>
> libexec/mysqld --print-defaults
>
> Regards,
> Monty
>
You can get this error if mysqld gets killed while there is an open
connection to it. Then the socket might go to FIN_WAIT2, while the
client for some reason (he probably does not realize that the connection
is dead) does not send FIN_ACK to finanlize the socket shutdown. In that
situation, if possible, make all the clients exist - they will send a
FIN_ACK, the socket will be closed, and the port will be available
again. If that is not possible or does not help, just wait a few minutes
- most systems will time out FIN_WAIT2 connections withing 5 minutes.
And of course, if someone else is listening on port 3306, you are in
trouble!
--
Sasha Pachev
http://www.sashanet.com
