URGENT - log switch problem

[farrokhp]

In our production Database we had only two redo log groups each with one 1 MB log member.
I created 3 new redo log groups each with a 20 MB log member in different disks and dropped old small group.
Very Surprisingly I see in the Alert log the log switch actually happens faster now !!!! Every 30 Seconds!!!
The log_checkpoint_interval = 9000000 and the log_checkpoint_timeout = 0 which I think it is correct.
Can you tell me what is wrong ??? I am stuck!

[aarroyob]

Which is the size of your log_buffer?

[farrokhp]

LOG_BUFFER = 409600 , Is that fine ?

[aarroyob]

If I'm not wrong, you will have a checkpoint too every 102400 (1/4 of the log_buffer). Probably It is for that you have now (after the recreate of redo logs) a switch often.

Correct me if I'm wrong.

Angel

[marist89]

Originally posted by farrokhp
In our production Database we had only two redo log groups each with one 1 MB log member.
I created 3 new redo log groups each with a 20 MB log member in different disks and dropped old small group.
Very Surprisingly I see in the Alert log the log switch actually happens faster now !!!! Every 30 Seconds!!!
The log_checkpoint_interval = 9000000 and the log_checkpoint_timeout = 0 which I think it is correct.
Can you tell me what is wrong ??? I am stuck!

log_checkpoint_interval and log_checkpoint_timeout have nothing to do with logswitch frequency.

Did you have a lot of "checkpoint not complete" messages in your alert.log before?


[Edited by marist89 on 03-25-2002 at 10:49 AM]

[farrokhp]

LOG_BUFFER = 409600 , Is that fine ?

[gpsingh]

Check out if u have any high DML during the fast log switch period.
Or too frequent commits.

[farrokhp]

Jeff and Others, We had checkpoint not complete when the redo log memebers was 1 MB not Now . But log switch happens even FASTER as I said!! We have lots of commit I guess but still I expect a better switch. My understanding is Commit flush log_buffer to log file members and if log files still not dull nommatter how many commit log switch doesn't happen! Is that right ?? Anyway what I have to do now ? The users says constantly that the system is slow.

[gpsingh]

check the system waits , where is the maximum waits ,
i believe the waits would be for LGWR parallel write , log file sync and
write complete waits and redo copy latch waits.

[farrokhp]

log file parallel wait is 1642113 and log file sync 671000 ( Total Wait ) , but my question is Why log switch happend too fast ? Log_buffer small ?? I don't think so!