[TYPO3-english] Saving pages in BE locks complete database

Christopher Lörken lists at bytro.com
Wed Jan 27 15:00:31 CET 2010

Am 27.01.2010 13:46, schrieb Dmitry Dulepov:
> Hi!
> On 2010-01-27 13:44:45 +0200, Christopher Lörken said:
>> But my question is:
>> Do I have to run it a second time since I ran it already once.
>> (I do not know if the magic has to be tickled more...)
> I wrote: "I know that "pages" is already InnoDB at your installation.
> Just run the query and see if it helps". I think it clearly suggests
> running query for the second time (AFTER running it to convert tables to
> InnoDB).

I am sorry that I understood you wrong, but in my eyes knowing that my 
pages table is _now_ InnoDB is different from knowing that I have 
executed that exakt statement a couple of days before I faced that 
problem again... Of course I should have assumed that you read my post 
carefully enough that you actually know that fact. Sorry.

Anyway, I googled for InnoDB fragmentation as you have suggested in your 
other post and it indeed looks to me like a possible problem. We are 
using some other quite large InnoDB tables (> 2 milion rows) which 
perform quite a lot of inserts / updates (no deletes). Since InnoDB uses 
one file to store all that data it might actually influence the 
performance of changing some data of flexible length in the middle of 
this file.

Well the SYS_LASTCHANGED time itself is fixed length, but who knows...

I've executed the statement again. It took ~ 11 seconds for a <200 rows 
table so it seems to have done something...

I will keep an eye on this, executing the alter table command the next 
time when there has past some time and we want to save again. I will 
report my observations here.

Thanks for your feedback and time, it is highly appreciated!


More information about the TYPO3-english mailing list