on 01-14-2013 7:34 AM
Hi Experts,
We have a urgent scenario to reduce downtime for migration. We are currently working on a project where customer is asking for guarantee for no more than a weekend downtime. We need to look into the possibility of a parallel export/import as well as other techniques to reduce downtime. Kindly suggest for the same.
HPUX to RHEL6
Oracle to Oralce
Datasize: 5TB
Thanks in advance,
-Alok
Hi,
How many test exports and imports you have done ??
How many hours it took export and import? In my environment, we performed a migration of 9 TB and we exported and imported it in 18 hours 20 minutes.
YOu need to perform the following activities to download the downtime.
1. Check what are the tables of more than 1 GB and split them into mulitple smaller packages. May be 1 GB of Table can be splitted into 2 mulitple packages.
2. Set the parameters recommended in the Note 936441 - Oracle settings for R3load based system copy
3. Use Migration Monitor for export and import.
4. Use Parallel export and import.
5.Based on the CPU ulitization and disk usage, you can dynamically change the R3load.
6. Always use the latest R3loads may be 730 Kernels.
7. Make sure you have ran the update the oracle stats just before the 24 hours of export.
8. Try to have atleast 2 or 3 parallel export and import test runs and try to minimze the export time after every run.
Hope it helps.
Thanks and Regards,
Vimal
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello Joesph,
Since you have minimized the time frame for such a huge database, will you please provide me more inputs on how did you prepare your "order_by" file?
Since I am going to perform this kind of activity(export SAP 4.7 non-unicode for unicode conversion and import as unicode system), I would like to have more clarity on the same.
Also I would like to know few more things
1) did you use the table splitting tool available in SAPINST ?
2) did you use migration monitor as separate tool or as part of SAPINST
Looking forward for your response.
Thanks & Regards,
Victor
Dear Vimal,
We are planning a OS/DB migration of 7 TB DB2 database. Obviously, our challenge is to perform this at a minimum downtime and your post looks very promising. Can you please let me know what was the storage (SAS or SSD) in place when you did the export of 9 TB DB in 18 hours. Have you done any optimization in the storage to speed up the export? Also it will help me if you share how many cores and RAM was used for this export.
Thanks
R Srinath
Hi,
For very large databases, you can contact SAP to use Minimize Downtime service (MDS) using Incremental Migration (IMIG) method, its a special service which is provided by SAP for large databases and requires intervention from SAP during the migration.
Regards,
Nilesh Jagtap
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Alok,
I have did sismilar kind of migration from HPUX to RHEL with 4 TB dataabase, you have two options - either you can use Distribution Monitor ( with minimum 4 nodes ) or contact SAP for zero-downtime services. However being a OSDB migration consualtant we would like to go with option one.
Definetly Migration Monitor is not right tool for bigger database migration ( untill you have hardware resource constraint - infrustructure team can not give you more machine or App Server to use for DistMon node)
As a stating point ( for your first POC) you can see my blogs - how to optimize migration using Distribution Monitor.
http://scn.sap.com/docs/DOC-34138
http://scn.sap.com/docs/DOC-34753
For Oracle specific settings check the OSS note : Note 936441.
If you have any question let me know.
Regards,
Dipam
Hi Dipam,
The document that shows how to optimze is very very useful. Thanks a lot. If you can share a document with screenshot on how to perform the distribution monitor
But I feel that exporting 4 TB within 2 days ( 48 hours) is pretty much easier with running 2 or 3 optimizations .As I said earlier, I exported as well as around 9 TB of ECC system within 18 hours. As well as our BW system which is of 2.8 TB has been exproted and imported in 8 hours appro....
We didnt use Distribution Monitor nor approached SAP for their help.
We just optimized DB ( oracle ) , used latest kernel version & was playing with order.txt file to start the export of larger tables .
We were dynamically changing the number of R3loads and order.txt file and exported and imported the ECC system.
Initial run was round 28 hours , Second run 20 hours. And finally 18 hours.
During every test run, we were closing monitoring the time taken for each package and optimized the long running table further and reduced the export time.
Thanks and Regards,
Vimal
could please elaborate on how to change the R3load number of process dynamically?
As we speak i am exporting a 7 TB BW system after table splitting and STR splitting with 670 packages. I am using 16 Parallel process and i can see CPU is 55% occupied...
It would be helpful if i can change the R3load dynamically?
Thanks and Regards,
Durairaj Premsundar
Hi Alok,
have you performed Mock Run for Migration.. ? How much time export took ?
R3Load Export-Import run time can be reduced by:
1. Parallel Export Import
2. Splitting large table
3. Package Splitting.
4. Optimizing large tables export / import order
5. Increasing parallel running R3Load export / import processes(depends on Hardware)
6. Distribution Monitor.
Best regards
ashish
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Ashish,
Migration has not been started yet, before that we have to give a proposal that how we are going to migrate 5TB of data in 2 days downtime,during weekend ( including export & import). Do you think it is possible using methods mentioned by you or is there any advance method to reduce the downtime.
Regards
-Alok
User | Count |
---|---|
87 | |
10 | |
10 | |
10 | |
7 | |
6 | |
6 | |
5 | |
5 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.