Re: Redo log question

Rob,
  Not sure if I'm completely on the right track, but checkpoints and log switches are two different things. You can have a log switch every time it fills up (20M in your case), but also have a checkpoint that flushes the buffers. This will reduce your recovery time. Checkpoints can be set up to occur by number of blocks or time and the init.ora parameters are log_checkpoint_interval and log_checkpoint_time. However, to see these checkpoints in your alert file you'll have to set log_checkpoing_to_alert = TRUE. I have just finished playing with my redo logs and we had quite a discussion about it at work. Anyways, I might not be 100% correct but I believe I'm close.

Hope this helps,

Charley O

RobW95 wrote:

> I recently created a new instance with three groups of redo logs, two members
> each. Each of the redo logs are 20mb each. The log_buffer parameter is set at
> 160k. Checkpoint is not turned on. This instance is basically a repository for
> inquiry so the logs will mostly be used for the weekly batch loads. We have
> been loading data into the instance successfully but I have noticed some
> messages in the alert log that concern me. The logs seem to be switching
> successfully most of the time but occasionally I see the message 'Can't
> allocate new sequence, checkpoint not complete'. The log doesn't seem to switch
> but instead seems to fill itself up again and then next time switches to the
> next log. This message is not an ORA message and I've looked in the reference
> manuals and I can't find anything related to it. There does not seem to be any
> pattern to it. Has anyone seen this before?
>
> Thanks in advance. Rob
Received on Thu Jan 08 1998 - 00:00:00 CST