topic Asynchronous function in Application Development and Automation Discussions

Asynchronous function

FredericGirod — Fri, 25 Feb 2005 08:00:16 GMT

Hello,

I need to add a line in a table (c) with data of two tables (a) (b). I have only one user-exit and it reads the data of only one table (a). I need both data of (a) (b) to add a line in (c). The user-exit is at the commit of the table (a), the commit of the table (b) comming just after.

I think to create an asynchronous function, call it in the user-exit and wait the commit of the table (b).

My user-exit is already call in an asynchronous function.

So, do you think it's dangerous to have two asynchronous function launch in the same time ? (process problem)

How would you do to wait a commit ? wait n seconds ? or there is a way to wait a commit on a table ?

Regards

Frédéric

Re: Asynchronous function

christian_wohlfahrt — Fri, 25 Feb 2005 08:33:53 GMT

Hi Frédéric!

You should search for different options.

Statement 'call function ... starting new task / in background' may trigger an implicit commit work -> this has to be avoided (at all costs), since later rollback work won't clear up whole transactions.

How about writing change pointers in exit for (a) and (b) - can be done without commit; only message (WE81) has to be customized, pointer activated (BD50) and CHANGE_POINTERS_CREATE_DIRECT can be used. Later you can read both entries 'asynchron' by a job.

Regards,

Christian

Re: Asynchronous function

FredericGirod — Fri, 25 Feb 2005 08:49:42 GMT

Hi Christian,

Thanks for you repply,

Your solution looks very interesting, I never play with things like that. Do you know where I could find documentation ? I look in help.sap.com but I don't find a real doc.

I work in MSEG & MSEGO2 (OIL) this is not too big for this solution ?

Frédéric

Message was edited by: Frédéric Girod

Re: Asynchronous function

christian_wohlfahrt — Fri, 25 Feb 2005 09:04:31 GMT

Hi!

it's no problem - with appropriate server - to have 1 million change pointers per day (and jobs handling this).

Menu points for this settings are behind:

workbench - ALE - ALE development - IDOC - change service (free translated, but you should find it anyway). Documentation is linked to this key words, too.

But ALE-scenarios are much more sophisticated and have nothing to do with my process description.

Start was already described.

Your job should contain following:

FM CHANGE_POINTERS_READ

Your (a), (b), (c) handling

FM CHANGE_POINTERS_STATUS_WRITE

Daily or weekly BD22 'archiving'.

If you plan this job in higher frequency then your runtime (every 5 minutes, sometimes runtime of job is longer -> two active jobs), move ...status_write in the beginning (+ commit work) -> then second job only reads new pointers, not same again.

Regards,

Christian

Re: Asynchronous function

FredericGirod — Fri, 25 Feb 2005 09:10:39 GMT

Thank a lot for this answer Christian, I can stop borring SAP OSS now

Regards

Frédéric