From: Adam Powell
Date: March 24 1999 12:46am
Subject: Fw: locking
List-Archive: http://lists.mysql.com/mysql/868
Message-Id: <001201be758f$ce7ca180$19937ed4@laptop>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit


Right... here is the details.  Basically doing a DELETE FROM last_seen_1
crashes our database.... as follows.

bash$ my
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 8069 to server version: 3.22.20a

Type 'help' for help.

mysql> delete from last_seen_1;
ERROR 2013: Lost connection to MySQL server during query
mysql> quit

It happens every damn time, and there are variably around 30,000 to 150,000
records in the database.... but there used to be a bigger database when we
used 3.21.16 and there was no problems doing it then.  Each of the last_seen
databases is accessed about 3 to 4 times a second, but I dont see why this
should cause a problem.  We are using --skip-locking, and there is only one
server accessing the tables at any one time.  This is running on an Ultra
Sparc 5 with 300 of its 512 Meg RAM free.

Hmm, any ideas, im confused!

Adam



>-----Original Message-----
>From: Michael Widenius <monty@stripped>
>To: Adam Powell <adam@stripped>
>Cc: sinisa@stripped <sinisa@stripped>
>Date: Tuesday, March 23, 1999 4:21 PM
>Subject: Re: locking
>
>
>>>>>>> "Adam" == Adam Powell <adam@stripped> writes:
>>
>>Adam> Thanks a lot, this has helped, however now another problem has
reared
>its
>>Adam> ugly head, and I suspect it could be a bug with mySQL is it happens
>every
>>Adam> time.  I have split up our IP table into 8 sub tables (this is
>working very
>>Adam> well - we are getting less records and correspondingly less load).
>>
>>Adam> I run the command 'DELETE FROM <ip_table>' on a PHP script using a
>lynx
>>Adam> browser set up to go off every day.  This just clears the table for
>the day
>>Adam> and lets us start afresh.  However every time this runs mySQL
>crashes.  This
>>Adam> problem never occurred with 3.21, but since we have upgraded doing a
>'DELETE
>>Adam> FROM x' where is there a large number of rows (some of the tables
>have
>>Adam> 150,000 rows - but then again its not that much is it?) crashes the
>entire
>>Adam> system and leaves me with a phone call from irate people and having
>to get
>>Adam> up at 5am to fix it :(  It crashes when I use mysqladmin
>>
>>Adam> I will look into some way of taking the table offline before I try
to
>do
>>Adam> anything major to it... but has anybody found a good way of getting
>around
>>Adam> this?
>>
>>Adam> Adam
>>
>>Hi!
>>
>>Which MySQL version are you now using ?
>>
>>I am not aware of any problem with <DELETE FROM> in recent MySQL 3.22
>>versions.
>>
>>Can you reproduce this anyway ?  If yes, I would really like to get a
>>test that I can run to test this!  Our standard scripts does a LOT of
>>deletes without any problem!
>>
>>Take a look at the 'fork_test.pl' script.  This should be easy to
>>modify to test things in a multi-user environment!
>>
>>If you are now using --skip-locking, I assume you have now only one MySQL
>>server that access the tables.
>>
>>Regards,
>>Monty
>>
>
>