Re: archive logs to NFS storage

Remigiusz Boguszewicz wrote:

> Hi,
>
> Environment:
> Solaris 9 on SPARC
> Oracle 9i EE
>
> We do on our production system one nightly cold backup. In case the
> machine with storage goes into flames 1 minute before a backup we loose
> the whole day of work.
>
> I was thinking of a simple solution to add an extra log_archive_dest_x
> parameter which points to an NFS share from PC with Linux.
>
> With redo archive files and last backup I am able to recover and lost
> only what was in not archived redo log, or what has not been coppied.
> That is a lot better tha a whole day.
>
> I thought that it would be simple to implement and whould do no harm to
> database. With log_archive_min_succeed_dest=1 even if my Linux NFS
> server disappears it should not have any bad impact on database.
>
> This is where I was wrong. Even with NFS mount with soft; or with
> soft,retry=2. Database goes untill it fills all available redo logs. The
> ARCH process does not release the redo logs and database stops. This is
> not good.
>
> Anyone with similar problem? How to make NFS "umount" itself if things
> go wrong, I think now about rsync maybe?
>
> The perfect solution whould be standby database but it requires the same
> architecture (SPARC) which is expensive. This is not an option now.
>
> Thanks for your comments.
> Remigiusz Boguszewicz

First comment, I'm guessing it has been a while since you actually tested your recovery process, so that should be the first order of business, to see what your current configuration will support, and to clarify with management what degree of data loss risk they are currently paying to avoid.

Check the STATUS column of V$ARCHIVE_DEST. Is the destination you defined valid? For more help, please post the relevant startup parameters and corresponding messages from the alert log.

You can also set up an archive log repository under Solaris which only uses a standby control file, not a complete database. Not sure what the licensing requirement is for this, but the doc says "This avoids most of the storage and processing expense of another fully-configured standby database."

http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96653/log_transport.htm#1065274

Lastly, if you want to do user-managed off-server backup of archived redo logs without NFS, send me a note off-line and I'll forward a Perl script which I have used without a problem in numerous production environments for years under HP/UX, Solaris, and Linux.

-Mark Bole Received on Fri Feb 25 2005 - 10:15:22 CST