Oracle recovery

[Vogra]

Hi All,
I hot-fbackup my oracle db: put all tablespaces in begin backup and fbackup all file systems. After that, put all tablespaces in end backup, switch logfile and backup controlfile to another location. A cron job put the archives on the tape.
The situation is: sometime one datafile is in use and it's not backed up. If it occur, and it is a datafile of RBS tablespace... in case of media crash, what I exactly have to do to recovery the system...?
Thanx in advance,
Note: We will istall a complete backup solution in one month.

[Yogeeraj_1]

hi,

You would be better served using RMAN. No OS
dependent scripts -- it can be as easy as "backup database" basically. Suggest
you look at RMAN.

If you need any further help, let us know.

Hope this helps!

Best Regards
Yogeeraj

[Julio Yamawaki]

Hi,

when you said that sometimes a datafile is in use... and it's not backed up, it's not true, if your database is in archivelog mode and you issued the sql command, alter database begin backup.
You always have access to backup a datafile in this case.
To recover from this situation in case of total crash:
. you have to backup the controlfiles together or at least, you have to have a script to re-create the controlfiles. You will have to have a instance created, recover the backup of datafiles and controlfiles and startup the instance with resetlogs at first and after that, recover the archivelogs until the time you want.
It's good to have a full offline backup, but you can recover totally with your online backup.

[Steven E. Protter · Accepted solution]

I regret to inform you.

You didn't back up anything.

Oracle describes an OS backup(fbackup) of an open database as a fuzzy backup.

They do not support recovering from such a backup. I have once in an utter emergency gotten away with making such a recovery. Its hard oracle won't help and you will get gray hairs from it.

Here is what you can do for the future.

As suggested:

rman
not hard to learn not hard to script or recovery

cold backup. Shut down the database take a backup. If its too big, you can break a disk mirror after a quick shutdown and restart the database quickly. On some databases this can leave you with less than a minute of downtime.

But let me make something clear to you. I'm a certified HP guy and I've got almost 5 years of Oracle experience. A fbackup of a database that is running is almost useless unelss there is no activity at all on the database. If activity is that low, you can shut the database and do a real copy style backup.

SEP

[Brian Crabtree]

Steve,

Not quite. Provided that the database has been placed in online backup mode, a standard filesystem backup works correctly. The database must be in archivelog mode, and the archivelogs for the period of the backup are required for recovery.

The problem most people experience is not placing the database into backup mode.

Brian

[Stan_17]

vogra,

foremost, do not put all tablespaces in 'begin backup' mode for hotbackups. This would lead to split or fractured blocks thereby generating too many redologs during backups.

so to avoid that, always backup one tablespace at a time. for e.g.

alter tablespace data begin backup;
do your host copy things here
alter tablespace data end backup;

Now to your problem, whether your datafile is in use or not, when you put the tablespace in 'begin backup' mode which basically freezes the datafiles headers until you release it with an 'end backup' mode. since the headers are frozen during begin backup mode, any checkpoints that occur during begin..coyp..end backup mode won't update recent SCN to these headers. howerver, when you do the recovery process, its knows what archived redologs to apply to completely recover these datafiles to bring up to the current SCN.

hence as far as you put the tablespace in begin backup..copy..end backup modes, then your are safe, provided you swtich the logfile at the end of backup process to make sure to archive all redos during a backup process.

hth,
Stan