R3ta, table splitting with Sapinst 620/640

[Former Member]

Dears,

I'm preparing a unicode conversion for an  R/3 47 Ent system, based on 640 kernel and Web As 620. I'm trying to manage the database export.

The Rdbms is Oracle 102.

I want to do both the table splitting and the export in customer order, apart the split of the STR files.

I'm tring to use the R3ta table splitting  , running the item "Table Splitting Preparation" proposed by the Sapinst 620 (latest available).

The R3ta is latest version, and is placed under the /sapmnt/SID/exe.

I managed the R3ta_hints.txt file in the proper way, (a lot of tables I want to split were missing) and the WHR clauses are generated in a reasonable time in a working directory togheter with the .STR files and a whr.txt file.

I run the "DB Export" phase, selecting also to extract the biggest tables from their original packages.

This selection contains among others all the tables I want to split. 

The DATA directory is created only when the Sapinst stop allowing me to change the export order, but here begins the problems.

First I move the .WHR files and the whr.txt from the R3ta working directory to the proper DATA directory where the DB export phase is supposed to write the export. 

Let's the VBAP one of the tables for which the WHR files where created; the several VBAP-1.WHR, VBAP-2.WHR and so on are under the DATA directory, togheter with the VBAP.STR

I want to start the export with VBAP-1, VBAP-2 ... and so on .

But I discovered this Sapinst does not generate the 'order_by.txt' file, where you can adjust the export order,  but appears only a screen table containing a row for any STR package and you have to change the order changing the default number 0.

So the table contains a row like :

|  0   |  SID_SID  |  VBAP.STR   |  an integer number  | path for the export  |  path for the DDLORA.TPL file  |  ......  |

I do not know in which way insert here my packages VBAP-1, VBAP-2 and so on.

If I leave the row as it is , just changing the 0 with -1 the VBAP export start as first, but it's exported by only one R3load.

So the .WHR clauses for this table are not taken in consideration.

Do I have to generate some file manually or I missing something ?

I thought the "Table Splitting Preparation" was supposed  to create all the things needed for the "DB export" step.

regards

[boobboo · Accepted solution]

Roberto,

The only way I have found the table splitting to be effective is to use the Migration monitor application - it allows you much more control over the aspects of the exports - every time I have talked to people having problems with table splitting they are all using SAPInst. When they move to Migration monitor, all their problems go away.

Also be careful when you use the orderby.txt file, because I have found that the distribution monitor and SAPInst get the scheduling reasonably right. In order to work out the most effective split takes a lot of analysis, all of which I have found to be on the wrong side of diminishing returns - it is easier just to throw more hardware at the problem which is a lot more reliable than human calculations on a process that as as many variables as the number of packages you divide your export into.

Thanks

Chris

[Former Member · Accepted solution]

Hi Chris,

Thanks for the feedback, I will try to introduce the migmon, in this Sapisnt version is not integrated , and beside it's not present at all neither into the /sapmnt/SID/exe or on Installation Master CD.

I have to try to manage it manually.

About the order_by.txt, as I described before this Sapinst 620/640 version does not generate it.

regards

[Former Member · Accepted solution]

I'm using Migmon, it works and I'm configuring the ddlMap option too to esport in unsorted way.

Using Migmon you can create the oerder_by.txt file and specify your order.

regards

[former_member189725 · Accepted solution]

Let me give you the format of an orderby file

======================================

[SWWCNTP0]

jobNum=3

SWWCNTP0-1

SWWCNTP0-2

SWWCNTP0-3

SWWCNTP0-4

SWWCNTP0-5

SWWCNTP0-6

=======================================

[SWWCNTP0]  -- This refers to a set of packages and for this 3 parallel R3load jobs/processes can . You can also specify a DDL file for this set using ddlMap option just below the jobNum option.

You need to place the .WHR files in the DATA directory .

Migration monitor is able to handle splitted tables and not SAPINST . In NW 640 systems , migmon was not integrated with SAPINST, so you need to invoke Migmon manually.

The orderby file you will have to create manually and update in the file export_monitor_cmd.properties

Regards

Ratnajit

[Former Member · Accepted solution]

Hi , yes it's as you says.

Neverthless I never used the set  packages feature.

I just prepare by myself the WHR file with the special splitter for oracle (R3ta it's to slow) and copy them under the DATA dir when Sapinst prompt to export the Database.

Then it takes some time to configure the migration monitor files, and to create manually the order_by.txt  and the ddlMap.txt files.

The real problem is, despite all this stuff of manual configuration, the export performances of the CDCLS remains very very bad.

Even if splitted , the first of the CDCLS's split end the export after more than 48 hours; that's definitely too long.

[former_member189725 · Accepted solution]

The export performance of CDCLS would be slow as this is a cluster table and the unicode conversion would definitely take a lot of time as it would do a sorted export. Also this phase of read ,concatenate ,convert and unload data into dump files is CPU intensive. The best way to minimize is to use parallel R3load processes as I suggested earlier and use as many parallel processes as possible. You can increase the jobNum dynamically in the migmon properties as well as in orderby file after checking the CPU utilization using the sar command. Since the SAP application is down , you can utilize the CPU to 90% during the export. Another way would be to use an application server with high CPU capacity .

Regards

Ratnajit

[Former Member]

Hi,

To my recollection:

In order_by.txt, you put package names, not split names.

VBAP, not VBAP-1 that is.

BR,

Samuli

[Former Member]

Hi,

as far I know If you split a big table and you want to start to download these pieces as first you have to put them on the top of the order_by file.

If you do the split for table , you create new small packages.

regards