Re: Shutdown Immediate Command

Not addressing the particular issue of this post, I can think of one particularly weird scenario that makes a shutdown abort a potential problem. Since it simply rips up the Instance, there is a possibility that you issue a commit statement, and if the abort command kicks in just before LGWR finishes writing to the redo logs, it is possible that you end up with a lost committed transaction.

Multiplexing wouldn't help here: the problem is the premature termination of LGWR, not whether your redo log files are safe from corruption or loss.

An immediate shutdown would allow LGWR to finish before gracefully terminating the Instance. With an Immediate, there can be no possibility of lost committed transactions.

Your timing would have to be awfully unfortunate, but it is possible.

Regards
HJR <jeremiah_at_wolfenet.com> wrote in message news:944mko$8f1$1_at_nnrp1.deja.com...
> raas.no_spam_at_interware.hu wrote:
> > With shutdown abort, you lose the redundancy in having commited
> > transactions both in (archive) logs and in the database files. Imagine
> > some disks failing.
>
> Ah, at last, a technical argument instead of a political/fear-based one.
> Your point is an important one, but this is why most people "multiplex'
> their redologs. In the very unlikely event that you lose a redo disk
> between the time you shutdown abort and start up, crash recovery will
> just use one of the other log group members. Note that your argument
> requires an exceedingly unlikely (but still possible) problem: a disk
> failure duing a brief time interval (between shutdwn and startup), with
> no multiplexed redologs. With multiple members per redolog group, there
> should be no problem using shutdown abort, even if a disk containing one
> of the log members is lost in the intervening moments.
>
> Thank you for a logical response, though. Everyone else just says, "cuz
> Oracle sez."
>
> --
> Jeremiah
>
>
> Sent via Deja.com
> http://www.deja.com/
Received on Thu Jan 18 2001 - 02:23:27 CST