This is a very complex question. You can't use snapshots because AIX
volume manager and JFS2 are not compatible with Linux. It would be
practically impossible to mount JFS2 file systems created on top of AIX
volume manager devices on a Linux server. The most realistic option is
bonding 4 10GB Ethernet devices into a 40Gbit/sec link, capable of
transferring 12 TBytes/hour and use NFS mount. With such a fast NFS
mount, you can do rman convert in approximately 2 hours. You can then
re-create the control file and open the database. That would give you
Oracle 11g database on Linux. Upgrading it should also take around 1
hour, depending on the speed of the Linux disk subsystem, which would
require around 5 hours of downtime.
Data pump would be a very tricky option because it would very likely
mess up PL/SQL objects, views, synonyms and grants. As a matter of fact,
it could be done, a team that included Arup Nanda and myself has done
something like that with a classic export/import in 2001. The project
was finalized during the Labor Day on 2001 and I was dealing with the
remaining issues late into evening of Monday, September the 10th 2001.
The next morning I came to the office late and, boy, was I in for a
shock. It was a beautiful late summer morning of Tuesday, September
11th, 2001. I will never, ever forget that day.
However, such project requires careful planning, numerous dry runs, lots
of scripting and a very competent DBA personnel. Our databases were
much smaller, the LAN speed was 100MBit/sec and we were moving database
from Oracle 7.3.5 on a Pyramid Nile system to Oracle 8.3 on a HP 9000/N
4-way RAC. The largest table has just exceeded 1 GB and the entire
database was around 700 GB, which was considered huge in 2001. So, you
should count on additional hardware expenses and around 6 months of time
for the project. This is not something that can be achieved by using
Agile project management methodology.
On 1/31/20 9:22 AM, ahmed.fikri@xxxxxxxxxxx wrote:
Hi all,
we are planning to migrate a 16 Terabyte database from 11g on aix machine to 12c on linux. In the target db is Dataguard used. The DB has about 50 Schemas the biggest one is about 11 TB the second is 2.6 TB then four with each one 1 TB the rest is each one less than 1 TB. Unfortunately all Schemas share the table spaces.
Wich approach could we use with less downtime?
My idea was to move the schemas to separate tablespaces an migrate the Schemas using transportable ts.
Or somehow copying the metadata to do new instance in such way the new db use the old data files and then copy them separately one by one.
Or even copying the Schemas separately using dblink and data pump.
Any idea please?
Regards
Ahmed Fikri
------------------------------------------------------------------------
Gesendet mit der Telekom Mail App <https://kommunikationsdienste.t-online.de/redirects/email_app_android_sendmail_footer>
