Re: Loss of Both current redo logs

Hi,

Several ideas:

1.) If restarting your database at morning is an option for you,

    you could use a different INIT.ORA at daytime.     In this file, change LOG_CHECKPOINT_INTERVAL to some     smaller number (counted in OS-blocks).     Or, if you want 15 minute intervals, use     LOG_CHECKPOINT_TIMEOUT = 900, which results in checkpoints     every 15 minutes.
    I personally prefer xxx_INTERVAL instead of xxx_TIMEOUT.     This parameter will cause your DB to issue a checkpoint after     the specified amount of redo information has been written to your logs.     It will *not* force a log switch. In a crash scenario, use     RECOVER DATABASE UNTIL CANCEL or ALTER DATABASE RECOVER CANCEL.     Oracle will most probably ask for the lost online redo log     (since your last checkpoint refers to an scn in this file).     Cancel the recovery and issue ALTER DATABASE OPEN.

2.) If you implement your 15 minute job, let it issue

    ALTER SYSTEM CHECKPOINT instead of forcing a log switch.     This way, you minimize your log volume. Proceed as described above.

Hope this helps, Robert.

In article <6ovjm6$sss$1_at_nova.thezone.net>,   fseymour_at_nfpower.nf.ca wrote:
> Hi
>
> I have a question on the loss of current REDO log files and recovery that I am
> hoping someone might like to comment on.
>
> It is my understanding that if we were to lose both current redo log files
that
> we would have to restore our database and recover up to just before the
> creation of those lost redo logs.
>
> For us this could mean the loss of allot of critical data. Here is why...
>
> We have fairly large redo log files (100m) to prevent frequent switching
> during our nightly batch processing that does a lot of large updates. This
> means that during our online day, where we have *alot* of small update
> transactions (customer requests), our log files do not switch at all. This
> implies that if we were to loss both current redo logs near the end of our
> online day we would loss all transactions for that day (we have no paper trail
> - no way to re-create these transactions). I am thinking that we should
> schedule a job to run every 15 minutes or so during our online day that would
> force a switch of our redo logs. That way, if we were to be so unlucky that we
> lost both of our online redo logs (yes they are on different disks and
> controllers), we not lose allot.
>
> What negative side effects would forcing a switch every 15 minutes imply? Is
> this a bad idea? Am I understand this correctly? Am I overly paranoid?
>
> Thanks in Advance for any input...
> Flora
>
>

-----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum Received on Wed Jul 22 1998 - 02:55:18 CDT