Hi,

i have to move a 10.2.0.3 database from solaris sparc to red hat
linux. I see 4 possibilities

1. exp/imp
2. expdp/impdp
3. Transportable Tablespaces
4. Oracle Streams

The first approach was expdp and was not succesful, because there are
400000 database objects in a 100GByte database. When i'm using expdp
there, is no progress after one hour. I think exp will take between 30
and 35 hours. Any idea is welcome.

regards
thomas

On Sun, 20 Jun 2010 09:51:25 -0700, tb wrote:

That is strange. How are you doing expdp? Did you specify the "parallel"
clause? Have you tried pre-creating the structures and then doing the
DATA_ONLY export/import? As for the transportable tablespaces, that is
probably the fastest option. You will, however, need to convert the data
files.



--
http://mgogala.byethost5.com

tb:

I would expect ( with reasonable bandwidth between the systems ) that
you could use a link and impdp and parallel and complete the process
in 8 hours or less. Unless the IO characteristics of the new linux
system really suck.

On 20.06.2010 22:33, John Hurley wrote:
That's what I'd attempt, too. I think there are also some constraints
on this IIRC charsets must match. Otherwise you get error messages that
are not really revealing.

Cheers

robert

--
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/

tb wrote,on my timestamp of 21/06/2010 2:51 AM:

That must be indeed a very slooooow Solaris sparc system!
I regularly export 100GB with expdp in AIX in much less than 2 hours.
Admittedly: a lot less than 400k db objects. Around 75K.

On Jun 20, 12:51*pm, tb <t.be... (AT) web (DOT) de> wrote:
Hi,

What is it waiting on?

Regards,

Steve

On Jun 21, 12:24*am, Noons <wizofo... (AT) yahoo (DOT) com.au> wrote:
See http://www.orafaq.com/usenet/comp.da...11/15/0883.htm

expdp springs memory leaks (see bug 5095025 for too little
information), and Solaris just generally seems to have idiosyncratic
performance issues, which I say based on the posts I've seen in
various places, my personal knowledge of it is way old, but it was
true then, too.

See http://jonathanlewis.wordpress.com/2.../07/pga-leaks/ and
Tanel's stuff referenced there and the interesting comments for some
things to look at.

For me, 50G (export size) direct path exp takes 45 minutes, full rman
about an hour on hp-ux, low end san.

Some people have complained about LOB and domain indexes expdp
performance.

jg
--
@home.com is bogus.
"Any sufficiently optimistic statement is indistinguishable from
sarcasm." - Eric Smith

joel garry wrote,on my timestamp of 22/06/2010 3:09 AM:

Yikes! And this is now Oracle's own OS...
I'm truly glad we decided to ditch all Solaris stuff a few years ago and gone
all AIX/Looneeks! Long before it all became Oracle.


Ball park what I get as well with AIX, 10.2.0.3 patched to the ears.



Haven't had a problem. In fact if anything, I'm impressed with datapump's
performance and capability in AIX! Use it all the time now, instead of good ole
exp/imp.
The only problem so far was a glitch that stopped sub-partitioned tables from
being imported in both data mode and transportable ts mode. Quickly fixed by
application of a patch that changed the XML skeleton used by the datapump mechanism.