How do you hot-backup ORACLE?

[Randy Hagedorn]

Yes, these are the tablespaces that we are altering to backup mode.

Another thought I had was to use our MirrorUX to do a split-mirror backup, off of the mirrored copy of our data.

I think we would still need to alter the tablespace into backup mode prior to the split, so that we would get an ORACLE checkpoint.

[T G Manikandan]

Recovery manager(RMAN) is a great tool for Oracle backup's.
It has several features like

1.you can look for block corruption.
2.Incremental backup
3.It can recover files without prompting you any date ot file.It has its own catalog.
4.The information in the catalog can be easily queried using reports and lists.

It can be integrated with the third party media management software like Legato manager to get wonderful results.


Thanks

[T G Manikandan]

I just forgot to put in the main point

RMAN does not backs up only the used space of the database.
Also RMAN does not put the tablespace in backup mode thus saving redo overhead.
It re-reads the database blocks until it gets a consistent image of that.

Also parallel backups are possible with RMAN.

[T G Manikandan]

Sorry for the last answer.

---It just backs up the used space on the database.
-----
It is late night here.infact early morning.
I have posted a query and waiting for replies from the great admins.

Sorry

Thanks

[Mark Seaman · Accepted solution]

Starting with 8i, Oracle introduced 'ALTER SYSTEM SUSPEND/RESUME' to freeze I/O on the server. You put the tablespace(s) in backup mode, execute this command, wait for a while, and then perform your backup. Since all I/O is frozen, this technique should only be used for some very quick method of backup like split mirror or snapshot. I say 'wait for awhile' because there is currently no way to tell that I/O has ceased.
Another option might be the Freeze/Thaw ioctl()'s for vxfs file systems. These flush buffers and freeze I/O synchronously, so you are guaranteed to have no I/O when they return.
Contrary to what others have said, Oracle continues to write to datafiles during backup mode. The SCN in the datafile header is frozen during hot backup, to ensure redo is applied when the recover command is issued. As long as you have the redo generated for the entire time the tablespace was in backup mode, you are guaranteed to be able to recover a consistent copy of your datafile. This is true as long as you as you have not turned off the _log_blocks_during_backup initialization parameter.