very Frequent redo switch

Printable View

11-15-2004, 12:42 PM
newcomer

very Frequent redo switch

Hello Everybody,

see the time below, for each redo switch from server

MINUTES
----------
13.9
9.95
5.93
5.7
6.17
6.17
5.87
6.17
6.93
7.05
6.35
6.83

The size of each archive is also 1G.

1048574976 Nov 15 07:12 _8431.arc
1048574976 Nov 15 07:29 _8432.arc
1048568832 Nov 15 07:47 _8433.arc
1048574976 Nov 15 08:07 _8434.arc
1048572928 Nov 15 08:23 _8435.arc
1048574976 Nov 15 08:39 _8436.arc
1048574976 Nov 15 08:56 _8437.arc
1048572928 Nov 15 09:14 _8438.arc

My redo log size is 1G. The log block size is 1024 on HP. In my DB log_checkpoint_timeout is set to 86400, but Oracle recommends to set log_checkpont_interval which should be calculated as log block size.

1) I would like to set my checkpoint_interval = (1G*1024*1024*1024)/1024 is that fine?
2) I am thinking to increase the size of redo log size to 2G.

Please suggest me what should I do?

Regards
11-15-2004, 12:48 PM
marist89

I'm not sure what "MINUTES" is, but your archive activity doesn't look bad to me. Do you get any "Checkpoint not complete" messages in your alert.log?
11-15-2004, 01:02 PM
newcomer

Minutes = each redo switch time difference.Not always, but very often the time diff is only 3 -4 minutes.

No I did not get any "checkpoint not complete" error.

But I am thinking about Load of redo switch. Frequent redo switch has lot of overheads and performance can degrade since each datafile will be frozen during each redo switch.

What do you suggest?

Regards
11-15-2004, 01:07 PM
davey23uk

tune your code to produce less redo - simple
11-15-2004, 01:11 PM
newcomer

The application has to do lot of updates.
11-15-2004, 02:04 PM
marist89

Quote:

Originally posted by newcomer
The application has to do lot of updates.

Then the database will have lots of redo.
11-15-2004, 02:13 PM
Axr2

If it's a warehouse or something, you might want to look into NOLOGGING operations.
11-15-2004, 02:50 PM
Mr.Hanky

That does not look bad to me.
Is this consistant or were they running a large job?
Has DB performance suffered recently?
I have seen them less than 30 seconds apart, but it does not last for too long so no cause for alarm.
11-15-2004, 03:25 PM
slimdave

Quote:

Originally posted by Axr2
If it's a warehouse or something, you might want to look into NOLOGGING operations.

That's a pretty big decision to make -- wouldn't help with updates of course.
11-15-2004, 03:28 PM
Axr2

Quote:

Originally posted by slimdave
That's a pretty big decision to make -- wouldn't help with updates of course.

I was just providing him/her food for thought, that's all.
11-15-2004, 04:46 PM
slimdave

Quote:

Originally posted by Axr2
I was just providing him/her food for thought, that's all.

I was just adding some garnish :)