Re: How to recreate accidentally deleted redo.logs

"Alex Filonov" <afilonov_at_pro-ns.net> wrote in message news:95la0t$v23$1_at_nnrp1.deja.com...
> Create mirrored log files. Oracle calls them redo log members. Put
> mirror copies onto different disks. Same goes for control files.
>
> In article <jiam7t89fmqt9uq0ntbgk11978lcqfdf79_at_4ax.com>,
> tbf_at_nospam.stam.dk wrote:
> > I made a stupid error: deleted all control files AND all redologfiles
> > by a foolish script.
> > I had a recent 'backup control file to trace' which I tried to use,
> > but it errs because of the missing log files.
> > I cannot create new log files because the database is not mounted, and
> > I cannot mount because of the missing control file.
> > Vicious circle!
> > Copying some old log files from tape didn't work.
> >
> > I had a very recent backup of the database and it is recovered now
> > with only minimal data loss, but it worries me that the redo.logs
> > seems to be that important on a closed database (close down normal).
> > With archiving on the log files should not contain any necessary data.
> >

What does that last statement mean? Archiving means you've copied logs which have been filled up and switched away from. There will always be at least one log (the one currently being written to) that hasn't been archived yet, and contains the details of the most recent transactions. Accordingly, the CURRENT redo log is absolutely critical to complete database recovery. And shutting down a database doesn't make a CURRENT log suddenly get archived.

All a normal shutdown buys you is a guarantee that a restored entire database won't need Instance Recovery before it can be used, since only committed data is in the data files, and no transactions took place after the time of the last checkpoint.

It's also the case that LGWR can throw a spas attack from time to time and introduce corruption into the redo logs. If you've only the one member per log group, you're stuffed: ARCH won't copy corrupt online logs. With multiple members, if ARCH finds corruption in one member, it will move onto the next member, until a clean archive can be constructed.

Redo -all of it, both online and archived- is the only thing standing between you and loss of data. You look after ALL of it well, and mirror like crazy (and I hope you at least use log_archive_suplex_dest to mirror your archives).

Regards
HJR
> > Can anyone tell me what I should have done (apart from dropping that
> > foolish script of course).
> >
> > (W2K, Oracle 8.1.6.1 Standard edition.)
> >
> > Yours Hans Erik Busk
> > tbf_at_cn.stam.dk
> >
>
>
> Sent via Deja.com
> http://www.deja.com/
Received on Mon Feb 05 2001 - 02:04:49 CST