Re: best way to convert from unix to linux

From: Norman Dunbar <>
Date: Fri, 25 May 2012 15:31:48 +0100
Message-ID: <>

Hi Dustin,

On 25/05/12 12:14, Dustin Hayden wrote:
> I have a 3TB database that needs to be moved from hp-ux to linux.

I did this recently with a 6.5 TB database or three. It just worked. I used transportable tablespaces to do it.

When I got it on Linux, I had to carry out a ROWS=NO import to get the PL/SQL, triggers etc - TTs don't take those across.

Once that has been done, I needed to run DBMS_STATS.UNLOCK_SCHEMA_STATS on any schema that had been imported by the imp ROWS=NO which, for some unknown reason, locks stats and doesn't bother to mention it. (Except in the manual.)

The process is well documented in the manuals, but basically you:

  • Run a few tests to make sure that there are no dependencies etc accross tablespaces you wish to transport and others that you are not transporting.
  • Put the tablespaces into read only mode.
  • Copy the data files making up the tablespace(s) to a safe place.
  • Run an exp (or expdp) in transportable tablespace mode.
  • Run an exp (or expdp) in ROWS=NO mode.
  • Take the tablespaces back out of read only mode.

Get the files over to the Linux server and copy them to "a/safe/place" which is not where they will end up going when the database is finished.

  • Build a new database, but WITHOUT the tablespaces you are transporting.
  • Create all the users in the HP database in the Linux database. Point their default_tablespace at a temporary one which is not coming from HP.
  • Run up RMAN with target set to the Linux database.
  • run the following command in RMAN:

convert datafile='a/safe/place/data_file_1.dbf',

'a/safe/place/data_file_2.dbf', ...
from_platform='HP-UX (64-bit)'
to_platform='Linux x86 64-bit'


So if your files were in /u01/app/SID1/data on HP and are going into /srv/app/SID1/data on Linux, you need:


The files will be copied from /a/safe/place to /srv/app/SID1/data and on the way, will be converted from HP-UX to Linux format.

  • When all files are converted, exit from RMAN and run an IMP or IMPDP in transportable tablespace mode.
  • Login to the database and put the tablespaces into read-write mode.
  • Run an IMP or IMPDP in ROWS=No mode to get the PL/SQL etc.
  • Login to the database and unlock schema stats.
  • Alter your user's default_tablespace to point at the corrrect one(s) that you have just TT'd into the database.


  • The exporting database must be 10203 Enterprise or higher.
  • The importing database can be SE or EE as required.
  • You might have trouble with Spatial Indexes.
  • The exporting database needs to have compatible set to or higher.
  • If you have to change it, do it while the tablespaces are read-write!
  • Both exporting and importing databases must use the same character set(s).
  • Sequences, triggers, procedures, functions, packages need to be imported by IMP or IMPDP. The TT process will not copy them across.
  • Stats will be locked for any schema imported with IMP and ROWS=NO. They must be unlocked afterwards.

Have fun.


PS. I have a very rough and ready document on this very subject if you want a copy, I can modify the names and such like to protect the innocent and let you have a copy.

Norman Dunbar
Dunbar IT Consultants Ltd

Registered address:
Thorpe House
61 Richardshaw Lane
West Yorkshire
United Kingdom
LS28 7EL

Company Number: 05132767
Received on Fri May 25 2012 - 09:31:48 CDT

Original text of this message