inodba switch_log causes severe slowness

On a small database (<10GB) with few users (<10) it took 11 minutes to do a log switch for our incrementals:

19-JUN-2007 11:00:00 INODST1751: Start of Tamino DBA tool version 4.2.1.8(811) on Windows
Tamino server online
Log file successfully changed

19-JUN-2007 11:11:35 INODSI1634: Successful completion

A full backup only takes ~3 minutes.

Anyone have any recommendations on what to look at to figure out why?

(this is a batch script executed every hour except when full backups are taken - server is Win2003 w/ 16GB of RAM and Dual 4-core Xeons).

This may sound funny, but the more traffic there is on the database, teh faster the log switch is. In your case, it seems the database has not been used for updates for several minutes.

This is a characteristic of teh log switch: it coincides with an update. no update (or other activity causing a log write, e.g. a timeout) - no log switch. So to make it faster: keep your database busy :wink:

regards

Harald

I failed to mention my main concern -

during those 11 minutes the database was completely unresponsive. Any thoughts on that?

bump to make sure I don’t get overlooked :slight_smile:

Hi,

before the log switch is executed, all transactions are quiesced. This migth explain the experienced behavior. For what purpose do you do the log switch?

Regards

Harald

I am doing the log switch to have an incremental backup. It is my company’s policy that on enterprise systems we do hourly incremental backups. So every hour I have a script that dumps the database keys and savepoints out of the registry and initiates a log_switch (except on hours where we do a full backup which is every 6 hours).

That said, when I tried to turn on incrementals it held up our application that uses the database for 11 minutes.