From kde  Sat Sep 25 09:29:33 2010
From: Thorsten Schnebeck <thorsten.schnebeck () gmx ! net>
Date: Sat, 25 Sep 2010 09:29:33 +0000
To: kde
Subject: Re: [kde] disabling Akonadi
Message-Id: <201009251129.34025.thorsten.schnebeck () gmx ! net>
X-MARC-Message: https://marc.info/?l=kde&m=128540703202488

Hi,

Am Freitag, 24. September 2010, um 19:49:47 schrieb Christian 
Mikovits:
> What i found out:
> 
> The logfiles are quite big (2x64m) and the datafiles itself pretty small.
> So an adjustment of mysql is neccessary to get rid of that:
> 
> -rw-rw---- 1 gaelic users  64M Sep 24 18:58 ib_logfile0
> -rw-rw---- 1 gaelic users  64M Sep 21 21:12 ib_logfile1
> 
> in mysql.conf of the akonadi dir just change
> 
> innodb_log_file_size=64M
> 
> to
> 
> innodb_log_file_size=8M
> 
> I hope no problems will arise with that change ...

Hmm, AFAIK these logfiles are transactions logs for disk management of 
the database (and not something like a syslog)

"Usually the redo log should be large enough and never fill up. 
Consequently your Innodb_log_waits counter should be 0 or at least not 
move when you look at it twice. If you experience Innodb_log_wait 
events one of two situations exists: Your server has write bursts larger 
than your redo log - the redo log is too small and must be extended. Or 
your server has persistent high write load and the redo log will overflow 
no matter how large you make it. In this case, but more disks or choose 
other ways to distribute the write load to more spindles.

By default the redo log consists of two files (innodb_log_files_in_group), 
each of which is 5M in size (innodb_log_file_size), for a total of 10M. This 
is usually much to small. Ideally you should have two files which are 64M 
to 256M in size, resulting in a total redo log of 128M to 512M. In any 
case the redo log cannot be larger than 4096M = 4G, even if you are on 
a 64 bit box."

(from: http://mysqldump.azundris.com/archives/78-Configuring-InnoDB-An-InnoDB-tutorial.html )

If space does not matter like on any modern PC and you want a high 
speed database: stay with the 2x 64M. Its more like a cache and will not 
grow. If you have GBytes of DIMAP data on your disk then 128MB is irrelevant.

But if you only use some contacts with Akonadi InnoDB seems to be 
way too powerful. On the N900 Akonadi for kmobile-kontact is defined as

> innodb_buffer_pool_size=8M
> innodb_log_file_size=2M
> innodb_flush_log_at_trx_commit=2

HTH

  Thorsten
___________________________________________________
This message is from the kde mailing list.
Account management:  https://mail.kde.org/mailman/listinfo/kde.
Archives: http://lists.kde.org/.
More info: http://www.kde.org/faq.html.