Re: Cheap Cheap Database Recovery
From: Michael Austin <maustin_at_firstdbasource.com>
Date: Fri, 20 Feb 2009 20:25:27 -0600
Message-ID: <ECJnl.12928$W06.11618_at_flpi148.ffdc.sbc.com>
pitufo wrote:
> Hi,
> Having a extremely low budget for IT, I tried to design a very cheap
> recovery database system without having to pay for a standby database.
> This is the idea:
> 1 -use a duplex destination for the archive logs located in a remote
> server (nfs mounted).
> 2-every 5 minutes a cron job copy one of the control files to a local
> disk (very fast) and from that location to the nfs mounted fs (much
> slower)
>
> Why ? : I did not want to have an active copy of control files using a
> nfs filesystem, I am afraid of performance bottleneck because the
> network. But at the same time I wanted to have a pretty recent backup
> of the control file. The same reason why I don't do redologs in
> another nfs location.
>
> Once a week I do a full backup of PRODUCTION database and copy the
> file (tar) to the recovery database server.
>
> When I tried the system at first, I was able to to the following in
> the recovery database:
> --untar fullbackup copy and place it in the right paths.
> -get the latest copy of the control file (from step 2 above) and
> replace all backup control files with this copy. (so n ow the dbf
> files are old from the full backup, and the controlfile is pretty
> recent).
> - sqlplus "/ as sysdba"
> -startup mount
> -recover database using backup controlfile until cancel;
> --usually it will ask me for all archive logs necessary to
> advance until scn on controlfiles.
> --cancel (when I reach the last usefull archive log file)
> -alter database open resetllogs;
>
> ---I tried the above successfully two or three times, so I even
> created the documentation for the whole process. (I wanted to give the
> administration a pros/cons of the method versus a standby database, or
> RAC solution).
>
> But, today, as a test, I tried to duplicate the steps above and the
> database only takes one archive logs before telling me that media
> recovery is complete. I made sure that I was using a controlfile much
> more newer than the fullbackup datafiles.
>
>
> Is not this suppose to work ?
> Any help will be GREATLY appreciated.
>
>
Date: Fri, 20 Feb 2009 20:25:27 -0600
Message-ID: <ECJnl.12928$W06.11618_at_flpi148.ffdc.sbc.com>
pitufo wrote:
> Hi,
> Having a extremely low budget for IT, I tried to design a very cheap
> recovery database system without having to pay for a standby database.
> This is the idea:
> 1 -use a duplex destination for the archive logs located in a remote
> server (nfs mounted).
> 2-every 5 minutes a cron job copy one of the control files to a local
> disk (very fast) and from that location to the nfs mounted fs (much
> slower)
>
> Why ? : I did not want to have an active copy of control files using a
> nfs filesystem, I am afraid of performance bottleneck because the
> network. But at the same time I wanted to have a pretty recent backup
> of the control file. The same reason why I don't do redologs in
> another nfs location.
>
> Once a week I do a full backup of PRODUCTION database and copy the
> file (tar) to the recovery database server.
>
> When I tried the system at first, I was able to to the following in
> the recovery database:
> --untar fullbackup copy and place it in the right paths.
> -get the latest copy of the control file (from step 2 above) and
> replace all backup control files with this copy. (so n ow the dbf
> files are old from the full backup, and the controlfile is pretty
> recent).
> - sqlplus "/ as sysdba"
> -startup mount
> -recover database using backup controlfile until cancel;
> --usually it will ask me for all archive logs necessary to
> advance until scn on controlfiles.
> --cancel (when I reach the last usefull archive log file)
> -alter database open resetllogs;
>
> ---I tried the above successfully two or three times, so I even
> created the documentation for the whole process. (I wanted to give the
> administration a pros/cons of the method versus a standby database, or
> RAC solution).
>
> But, today, as a test, I tried to duplicate the steps above and the
> database only takes one archive logs before telling me that media
> recovery is complete. I made sure that I was using a controlfile much
> more newer than the fullbackup datafiles.
>
>
> Is not this suppose to work ?
> Any help will be GREATLY appreciated.
>
>
See the other comments...
I will add - no one has enough time or $$$ to do it right, but everyone has the time AND $$$ to fix it once it is broke. The previous much less costly than the latter. Do yourself and your company a favor and do it right! Received on Fri Feb 20 2009 - 20:25:27 CST