Original subtitles

In this lesson, we're going to discuss complete recoveries

with RMAN.

Now, a complete recovery is going

to be a recovery where all the data is recovered up

to the point of failure.

So what we're going to do in this example

is actually simulate a failure and perform a recovery.

So it goes without saying that in order to perform a recovery,

we must have a good backup.

So we ensure that we have a good backup before we attempt

to simulate a media failure.

We need to remember that when we do a recovery,

it must be done in the mount state.

So the database must be in mount.

We cannot do a database recovery while the database is in no

mount state or in open state.

And indeed, if we have a media failure,

it's not very likely that the database

would be in an open state, at least in any functional way.

When we do complete recoveries, we

can do them much in the same way as backups.

So we can do them at the tablespace level, the data file

level, or the entire database.

And recovery involves two steps--

restore and recover.

Restore is restoring the backup, the file from a backup.

And recover is the act of rolling it forward, applying

all the changes that have occurred that are in the redo

logs and archive logs.

So restore and recover.

So in order to simulate this, log into RMAN.

Connect target.

And shut down the database.

In this case, we need to shut down the database

because Windows will prevent us from actually

doing what I'm about to do.

That is to change the name on a data file simulating its loss,

because it locks the files when the database is open.

However, on Linux and Unix, you are

free to go and destroy a file, one of the data

files of the database, without the operating system stopping

you.

The database is shut down.

And so I'm going to go to the directory that

holds the data files.

Oracle Base directory, there'll be

an oradata, the name of the database, and then data file.

And so I'm going to target the data file that belongs

to the user's tablespace.

And just for slight safety's sake,

I'm just going to rename this, rather than delete it,

to dot old.

However, the effect is the same, because when

the database attempts to start, it

will have a problem because it will not find this data file

that it expects to be there.

The data file locations are written in the control file,

and it expects all of those to be there when

the database startup occurs.

I'm going to exit out of here.

sqlplus as sysdba.

And now, we're going to just type startup

and see how far we get.

So notice that the instance started and gave us

the information about the memory.

And the database made it to mount state.

But it couldn't get any farther than that.

It says I cannot identify our locked data file number 4

and then gives us that file.

And there it is, the data file that belongs

to the user's tablespace.

And so it cannot see it, so it's as if we just went ahead

and deleted it.

So now, we are already in the mount state,

so we're in a position that we can restore and recover

this data file.

So we say restore data file 4.

And that, again, is the number of the data file

that's indicating is missing.

And we need to be in RMAN to do this.

The RMAN connect target, restore data file 4.

So what it's doing right now is reading the backed up

data for data file number 4 from our backup,

and its restoring it.

And if we were to see, notice that it

has restored a data file for the user's tablespace.

But we're not finished, because we

would need to recover all of the data that

was lost between the time that the database was shut down

and when the data file was lost.

So we say recover data file 4.

Starting media recovery.

Media recovery complete.

And now, our database is ready to be opened.

So we do alter database open.

So what would we do in the event of a complete loss

of the database?

So what if we lost all of the data files in the database

if I'd actually just remove those files that are here

under the data file directory?

Well, the process would be very, very similar.

We would simply type restore database and then recover

database.

And following the activity that occurred,

our database would be fully recovered and restored.