alexblyth.net

Ultra-fast PDB thin clones on Exadata Exascale

Written by

Alex

in

, , , ,

Want to see how quick it is to create a clone of an Oracle 23ai PDB on Exadata Exascale?

First a little background information.

I have a “simple” database in the example, with a PDB – named DEMOPDB1. 

show pdbs

SQL> show pdbs

   CON_ID CON_NAME    OPEN MODE     RESTRICTED
_________ ___________ _____________ _____________
        2 PDB$SEED    READ ONLY     NO
        3 PDB1        READ WRITE    NO
        4 DEMOPDB1    READ WRITE    NO

It’s approximately 2.4 TB in size (temp excluded).

SQL> select p.name, sum(f.bytes)/1024/1024/1024 as total_gb
  2  from v$pdbs p, v$datafile f
  3  where p.con_id = f.con_id
  4* group by p.name;

NAME                        TOTAL_GB
___________ ________________________
PDB$SEED          10.625396728515625
PDB1              14.718536376953125
DEMOPDB1       2403.8817901611328125

To create a clone of this PDB on Exascale, all we need to do is

SQL> set timing on; <- so you can see how fast it is

SQL> create pluggable database demopdb2 from demopdb1 snapshot copy;

Pluggable database DEMOPDB2 created.

Elapsed: 00:00:11.105 <- ~2.4 TB in under 12 seconds!!

The keywords ‘SNAPSHOT COPY’ are important here. They’re instructing the database to use the space-efficient, thin-provisioned, redirect-on-write capabilities of Exascale to create

Then, we’ll open it up.

SQL> alter pluggable database DEMOPDB2 open instances=all;

Pluggable database DEMOPDB2 altered.

Elapsed: 00:00:10.897 <- opening the PDB on 2 instances in this case

Check the size of the new PDB – spoiler alert, its the same size as the PDB I was cloning!

SQL> select p.name, sum(f.bytes)/1024/1024/1024 as total_gb
  2  from v$pdbs p, v$datafile f
  3  where p.con_id = f.con_id
  4* group by p.name;

NAME                        TOTAL_GB
___________ ________________________
PDB$SEED          10.625396728515625
PDB1              14.718536376953125
DEMOPDB1       2403.8817901611328125
DEMOPDB2       2403.8817901611328125 <- new PDB!

And that’s it – less than 30 seconds and you have a thin clone of a PDB.

How cool, and easy, is that!

More posts