Adding new PDB to a CDB with Standby

On Oracle 12c we got a new feature on RMAN which allows you to restore database files, over the network, from a physical standby database by using the FROM SERVICE clause of the RESTORE command.

So in a Data Guard environment in case you add a new Pluggable database (PDB) you don’t need to duplicate the entire Container database (CDB) again. You can restore only the new PDB to the physical standby. Cool right, less work, faster work. I would prefer this as RMAN would do all its magic and we would not need to manual copies to renames.

On the practical side, this would be done in case you need to add a new PDB in your Data Guard configuration.

Stop the archive log apply on the standby using DG broker

validate database verbose PRIMARY;
validate database verbose STANDBY;
edit database 'STANDBY' set state='apply-off';
On the standby database restore the new PDB

rman target /
run{
allocate channel prim1 type disk;
set newname for pluggable database NEW_PDB to new;
restore pluggable database NEW_PDB from service PRIMARY;
switch datafile all;
}

– Enable recovery on the new PDB, if you running an Active DataGuard you need to stop it and start the standby database in mount, also if you get ORA-1113 make sure you are in MOUNT stage.

sqlplus / as sysdba
shtudown immediate
startup mount
alter session set container=NEW_PDB;
alter pluggable database enable recovery;
exit

– Enable archive log apply

edit database 'STANDBY' set state='apply-on';
validate database verbose PRIMARY;
validate database verbose STANDBY;

More information you can find on the MOS Note Making Use Deferred PDB Recovery and the STANDBYS=NONE Feature with Oracle Multitenant (Doc ID 1916648.1)

Hope it helps,
Elisson Almeida

Unplug/Plug PDB between different Clusters

Everyone test, write and show how to move pluggable databases between containers (CBDs) in the same Cluster, but a little more than a few write/show about move pluggable databases between different clusters, with isolated storage. So, let’s do that:

OBS: Just to stay easy to understand, this post is about migration of a Pluggable Database (BACENDB) from a cluster named ORAGRID12C and a Container Database named INFRACDB to the Cluster CLBBGER12, into Container CDBBGER.
(Click on images to get it bigger)

1. Access the container INFRACDB (Cluster GRID12C) and List the PDBs: 1

2. Shutdown BACENDB:
2
(of course it does’n worked with a normal shutdown. I don’t know what I was thinking… haha) 3

3. Unplug BACENDB (PDB) to XML (must be done from Pluggable, as you see…) 4
4. Created an ACFS (180G) to use as “migration area” mounted on “/migration/” in ORAGRID12C cluster:
5

5. Copy Datafiles and Tempfiles for the “/migration” through ASMCMD cp 6

6. ACFS exported and mounted as NFS on destination (CLBBGER12): 7
8

7. Pluggable created (Plugged) on new Cluster (CDBBGER), using “MOVE” FILE_NAME_CONVERT, to send the files to diskgroup +DGCDBBGER:

9

7.1 How it looks like on alert.log?

10

7.2 How about the Datafiles?

11

7.3 Checking database by remote sqlplus:

13

8. Creating the services as needed:

12

9. Dropping Pluggable from INFRACDB:

14

That’s Okey? Of course there is a few other ways to copy the files from an infra to another, like scp rather than mount.nfs, RMAN Copy, or other possibilities…

By the way, one of the restrictions of pluggable migration is to use the same endian format. Buut it’s possible to use RMAN Convert Plataform and convert datafiles to a filesystem, isn’t?
So, I guess it’s not a necessary limitation. Must to test an write another post… haha

About the post, this link helped, but, again, don’t mention about “another” cluster/infra/storage.

Matheus.