Restoring Your Confidence in Oracle Backups
John Ouellette
Wouldn't it be great to be able to rely on your backups and have full confidence
that they will work properly during a disaster or a routine restore operation?
Unfortunately, making assumptions about your restores can lead to delays in
projects, wasted time, and ultimately data loss. While it may be fairly easy
to configure backups and assume the restore process will work as long as the
backups have run successfully, the reality is that each process is different
and involves its own set of challenges.
On a basic hardware level, backups write to tape, and restores perform a read
operation on the tape hardware. As the following example illustrates, success
of one process does not guarantee success of the other.
The Power of Testing
When testing our database restores, I set the read-only tab on the tapes to
avoid any surprises or accidental tape overwrites. The backup logs looked fine
with no errors. Unfortunately, the particular tape drive selected for the restore
had a bug in its firmware and rendered the drive unusable with the tab set.
If we hadn't tested, we would have incurred much lost time and much aggravation
chasing down a solution when a real problem occurred on a production database.
Testing not only can increase your confidence in the success of restores,
but also allow you to estimate more accurately when the restore will finish.
It may seem reasonable to estimate a database that is backed up in an hour can
be restored in an hour, but that is not always the case.
For example, it is common for database administrators to back up multiple
databases to one tape drive at a time, whether it's to a locally attached tape
or a SAN.
|
