We had done that, but are not really seeing any consistent issues with a
long running query in the normal sense. What we are seeing is that when the
system is running "well" (innodb_thread_concurrency at 1), we see little to
nothing for slow queries. However, when the server hits whatever the source
of the bottleneck is, every query being run slows to a crawl.

On Wed, Aug 27, 2008 at 8:00 PM, Jake Anderson <[EMAIL PROTECTED]> wrote:

> Have you tried turning on the slow query log and seeing what queries are
> causing the problem?
> Depending on your disk layout you can try moving some of the tables to
> different spindles which might improve performance.
>
>
> Rob Beglinger wrote:
>
>   I appreciate your observations, unfortunately I am stuck with Intel
> Quad-Core processors. :)
>
> We are having significant issues handling IMAP requests in a timely manner
> (12 seconds to load the Inbox in Thunderbird and an in-house Java-mail app)
> on our dbmail implementation and I believe the problem resides with the
> database configuration.  Is there anyone out there that is running dbmail
> 2.2.10 with a MySQL 5.0.X database with a large user group that can share
> their my.cnf?  Our user count is over 4000.  This is our my.cnf:
>
> # You can copy this file to
> # /etc/my.cnf to set global options,
> # mysql-data-dir/my.cnf to set server-specific options (in this
> # installation this directory is /usr/local/mysql/data) or
> # ~/.my.cnf to set user-specific options.
> #
> # In this file, you can use all long options that a program supports.
> # If you want to know which options a program supports, run the program
> # with the "--help" option.
>
> # The following options will be passed to all MySQL clients
> [client]
> #password       = your_password
> port            = 3306
> socket          = /var/lib/mysql/mysql.sock
>
> # Here follows entries for some specific programs
>
> # The MySQL server
> [mysqld]
> port                            = 3306
> socket                          = /var/lib/mysql/mysql.sock
> max_connections                 = 500
> max_connect_errors              = 999999
> skip-locking
> key_buffer                      = 384M
> max_allowed_packet              = 8M
> table_cache                     = 1024
> sort_buffer_size                = 16M
> read_buffer_size                = 16M
> read_rnd_buffer_size            = 8M
> myisam_sort_buffer_size         = 64M
> thread_cache_size               = 16
> query_cache_size                = 128M
> query_cache_limit               = 2M
> long_query_time                 = 10
>
> # These variables determine the max size of a memory temp table
> tmp_table_size                  = 32M
> max_heap_table_size             = 32M
>
> tmpdir = /tmp
> datadir = /var/lib/mysql
>
> # Try number of CPU's*2 for thread_concurrency
> thread_concurrency = 32
>
> # Don't listen on a TCP/IP port at all. This can be a security enhancement,
> # if all processes that need to connect to mysqld run on the same host.
> # All interaction with mysqld must be made via Unix sockets or named pipes.
> # Note that using this option without enabling named pipes on Windows
> # (via the "enable-named-pipe" option) will render mysqld useless!
> #
> #skip-networking
>
> # MySQL General Query Log
> #log=mysql.general.log
>
>
> # MySQL Binary Log
> log-bin=mysql-bin
> #expire_logs_days       = 1
>
>
> # MySQL Slow Query Log
> #log-slow-queries=/var/lib/mysql/mysql.slow.log
> #log-queries-not-using-indexes
>
> # required unique id between 1 and 2^32 - 1
> # defaults to 1 if master-host is not set
> # but will not function as a master if omitted
> server-id       = 1
>
>
> # InnoDB settings
>
> innodb_file_per_table
> innodb_data_home_dir            = /var/lib/mysql/
> innodb_data_file_path           = ibdata1:10M:autoextend
> innodb_log_group_home_dir       = /var/lib/mysql/
> innodb_log_arch_dir             = /var/lib/mysql/
> innodb_log_files_in_group       = 2
> innodb_buffer_pool_size         = 24576M
> innodb_additional_mem_pool_size = 20M
> innodb_log_file_size            = 256M
> innodb_log_buffer_size          = 16M
> innodb_flush_log_at_trx_commit  = 1
> innodb_lock_wait_timeout        = 50
> innodb_thread_concurrency       = 1
> innodb_thread_sleep_delay       = 0
> innodb_flush_method             = O_DIRECT
> transaction-isolation           = READ-COMMITTED
> #innodb_sync_spin_loops         = 20
> innodb_concurrency_tickets      = 1500
> innodb_support_xa               = 0
> innodb_open_files               = 1000
> #skip_innodb_doublewrite
> #skip_innodb_checksums
>
> [mysqldump]
> quick
> max_allowed_packet = 16M
>
> [mysql]
> no-auto-rehash
> # Remove the next comment character if you are not familiar with SQL
> #safe-updates
>
> [isamchk]
> key_buffer              = 256M
> sort_buffer_size        = 256M
> read_buffer             = 2M
> write_buffer            = 2M
>
> [myisamchk]
> key_buffer              = 256M
> sort_buffer_size        = 256M
> read_buffer             = 2M
> write_buffer            = 2M
>
>
> We are running MySQL on SLES 10 SP2 kernel version is 2.6.16.60-0.21-smp
> 32GB of RAM on a 4xQuad-Core Intel server.  We've changed
> innodb_thread_concurrency on the fly to see if we had any performance
> increase, but the latest tests show no improvement with decreased
> performance at innodb_thread_concurrency at 4 or higher.  I have been given
> until Friday (2 days) to resolve the issue before I am forced to go back to
> a flat file email system, so I greatly appreciate any and all help.
>
> If there are any other areas that I should be looking at, I will gladly
> take any/all suggestions.
>
> Thank you very much,
>
> Rob
>
>
>
>
>
> On Fri, Aug 22, 2008 at 5:15 AM, Vladimir Likhachev <
> [EMAIL PROTECTED]> wrote:
>
>> Sorry, of course, Sun bought MySQL AB.
>> Large server memory (more then 4 GB) usage details is the main reason of
>> AMD64 usage and its suggestion by Oracle, I think... Not details of
>> InnoDB format or MySQL server realisation... At whole, for each (heavy?
>> strongly?) loaded server program large memory usage is very important.
>> The next reason is non-independent level 2 cache memory in Intel Core Quad
>> processors.
>> Intel Em64T technology is (a very simple view) memory over 4 GB mapping in
>> a window 128 or 256 MB width laying below 4 GB in UMA address space.
>>
>> All this is my opinion only, based on Intel SR/SH (4*Xeon-MP) and Intel
>> platforms 4*Xeon 2*Core processors, 1*Intel CoreQuad workstations usage. No
>> use Intel CoreQuad for "serious" servers in companies where I work. Up to
>> 120 GB InnoDB bases in DBMail 2.xx.
>>
>> ------------------------------
>>
>> Date: Fri, 22 Aug 2008 10:07:10 +1000
>> From: [EMAIL PROTECTED]
>> To: [email protected]
>> Subject: RE: [Dbmail] Query to show mailbox folders running slow
>>
>>
>> On Thu, 21 Aug 2008 08:01:43 +0000, Vladimir Likhachev  wrote:
>>
>> > Oracle (current owner of MySQL AB) suggests AMD64 processors.
>>
>> Minor correction. Sun bought MySQL AB. Oracle bought InnoDB. I'm not too 
>> impressed with either move frankly. How many people think Oracle bought 
>> InnoDB to incorporate their technology into the next version of Oracle?
>>
>>
>> ------------------------------
>> Explore the seven wonders of the world Learn 
>> more!<http://search.msn.com/results.aspx?q=7+wonders+world&mkt=en-US&form=QBRE>
>>
>> _______________________________________________
>> DBmail mailing list
>> [email protected]
>> https://mailman.fastxs.nl/mailman/listinfo/dbmail
>>
>>
> ------------------------------
>
> _______________________________________________
> DBmail mailing [EMAIL PROTECTED]://mailman.fastxs.nl/mailman/listinfo/dbmail
>
>
>
> --
>
> Vapour Forge
>
> Jake Anderson
>
> Project Manager
>
> Mobile:       0412 897 125
>
> Email:         [EMAIL PROTECTED]
>
> Web Page:  www.vapourforge.com
>
> Your source for custom IT services
>
> _______________________________________________
> DBmail mailing list
> [email protected]
> https://mailman.fastxs.nl/mailman/listinfo/dbmail
>
>
_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail

Reply via email to