Migration from HPUX to Linux

From: pier paolo Bruno <pbrunoster_at_gmail.com>
Date: Wed, 23 Oct 2024 23:12:33 +0200
Message-ID: <CA+dM1yM9-P4CiLt1JBwwqm44OcQBT1sr6haNmB-5uxPNTcSM5w_at_mail.gmail.com>



Hi,
I have to move a database of 1,5gb from hpux 11.2.0.1 to a linux machine with 19.24 with oracle restart (different endianess ) . The database is formed by 500gb of data and lob, a single lob of 700b and indexes. The single lob table is not on separated tablespace. I discarded the use of transportable tablespace because for installing 11.2.0.4 on OL 8.10 i need a patchset and a patch that i can not download :
Requirements for Installing Oracle Database/Client 11.2.0.4 on OL8 or RHEL8 64-bit (x86-64) (Doc ID 2988626.1). It would also be a serious problem to put the source in readonly for the time of the datafile copy, it would last at least 8 hours and for a health system it is a problem (source system is old and slow ) . I could ask a window once , but we evaluated that with datapump i can do all the test i want without giving impact, in the window 16pm 7 am . I can do the migration in 2 steps ,so our idea is to export everything without the big lob table and to do it afterwords. Applications can survive with this table empty for what we know till now. The hpux machine is quite slow but in an initial test i managed to export with datapump and import everything in 13 hours and for me it is ok. The problem is how to proceed with the 700gb lob. Exporting on filesystem seems quite slow, i had to kill the export because i could i had a window from 16pm to 7am and only 200 were exported . I know that when i will do the migration i will be able t o let it going but i was asking if it was the best way to do the task. Lobs are basicfiles.
I also try a test like this :
create global temporary table tb1 ;
insert into tb1 select * from table_at_remote where rownum< 5000; insert into final_table select * from tb1 ; It worked but it took 20 minutes for 5gb . Surely i could study a way to cluster values of primary key in plsql and to try to build a manual "parallelism" mechanism, but i don't know if at the end i go faster and i have not a test envirorment . I can not modify the source envirorment because it is a system of pubic health service and so we can touch anything . I was wondering to try impdp with network link but i am not sure if it works with basicfile lob ( if i can i try tomorrow but i am allocated on more than one project ) .
Have you a different idea to test ?
Any hint is appreciated

Regards
Paolo

--
http://www.freelists.org/webpage/oracle-l
Received on Wed Oct 23 2024 - 23:12:33 CEST

Original text of this message