topic Commit for Asynchronous Function Module in Application Development and Automation Discussions

Commit for Asynchronous Function Module

Former Member — Tue, 20 Sep 2011 09:38:46 GMT

Halo Experts,

I am working on an Perfomance Optimization project . I am trying to optimize the program by scehduling the individual methods inside the program via asynchronous function modules . By this way I can run this independent methods parallely ( and not sequentailly) . for eg

if the program has

 
 me->m1( ).
 
me->m2( ).
 
me->m3( )

if m2 and m3 are independent of each other I schedule them parallely by

me->run_m2_m3_parallely( ) see the below code

 

CALL FUNCTION 'ZM2'
    STARTING NEW TASK 'FUNC1'
    DESTINATION 'NONE'
    CALLING ZCLASS=>set_flag_4_m2 ON END OF TASK.
 
CALL FUNCTION 'ZM3'
    STARTING NEW TASK 'FUNC2'
    DESTINATION 'NONE'
    CALLING ZCLASS=>set_flag_4_m3 ON END OF TASK.
 
 
 
 
  WAIT UNTIL l_m2_flag = 'X' AND
             l_m3_flag = 'X'.

the FMs ZM2 and ZM3 are updating Z Database tables( say ZM2 and ZM3 ) . Do I need an explicit commit work for Individual FMs or a single commit work for both the FMs.

How can I roll back both the updates to the database if a dump occurs ?

How should the commit work statement be in this case ?

Regards

Arshad

Re: Commit for Asynchronous Function Module

SuhaSaha — Tue, 20 Sep 2011 09:53:20 GMT

Hello Arshad,

When you're using asynch-RFC(aRFC) they are run in separate RFC contexts which don't share a common LUW. You need to maintain an explicit commit to store the data in the DB tables.

How can I roll back both the updates to the database if a dump occurs?

What exactly is your requirement, can you elaborate?

BR,

Suhas

PS: If you're using a transactioncal RFC(tRFC), they are registed in a common LUW.

Re: Commit for Asynchronous Function Module

arseni_gallardo — Tue, 20 Sep 2011 10:21:56 GMT

I think you can solve your issue by using the KEEPING TASK addition.

In your ZCLASS=>set_flag_4_m2 method try this:


METHOD set_flag_4_m2.
  RECEIVE RESULTS FROM FUNCTION 'ZM2 ' KEEPING TASK.
 
ENDMETHOD.
METHOD set_flag_4_m3.
  RECEIVE RESULTS FROM FUNCTION 'ZM3 ' KEEPING TASK.
ENDMETHOD.

And after your wait you can try to do this:


WAIT UNTIL l_m2_flag = 'X' AND
             l_m3_flag = 'X'.

* In set_func2_commited you must receive results from BAPI_tRANSACTION_COMMIT and set global flag g_func2_commit
CALL FUNCTION 'BAPI_TRANSACTION_COMMIT'
    STARTING NEW TASK 'FUNC2'
    DESTINATION 'NONE'
    CALLING ZCLASS=>set_func2_commited ON END OF TASK.

* In set_func3_commited you must receive results from BAPI_tRANSACTION_COMMIT and set global flag g_func3_commit
CALL FUNCTION 'BAPI_TRANSACTION_COMMIT'
    STARTING NEW TASK 'FUNC3'
    DESTINATION 'NONE'
    CALLING ZCLASS=>set_func3_commited ON END OF TASK.
WAIT UNTIL g_func3_commit EQ 'X' and g_func2_commit EQ 'X'.

Re: Commit for Asynchronous Function Module

Former Member — Tue, 20 Sep 2011 20:27:36 GMT

Halo Suhas,

My aim is to run independent methods parallely so that I can optimise the perfomance of the program

My question regarding roll back was if some exception happens and dumps in Zm3 , will the M2 update be rolled back also?( because we are having individual commits inside Zm2 and Zm3 )

I tried this parallel execution , it is succesfully updating table m2 ( same as the number of entries of sequential run ) and table m3 it is not updating.

But the strange thing is if I put user defined break point in Zm3 and executing the program it is correctly updating m3 table .

Regards

Arshad

Re: Commit for Asynchronous Function Module

SuhaSaha — Wed, 21 Sep 2011 04:48:09 GMT

Hello Arshad,

My question regarding roll back was if some exception happens and dumps in Zm3 , will the M2 update be rolled back also?( because we are having individual commits inside Zm2 and Zm3 )

No, because you're running the updates in parallel, if you want to achieve this requirement you'll have to bundle the updates in a single LUW.

I tried this parallel execution , it is succesfully updating table m2 ( same as the number of entries of sequential run ) and table m3 it is not updating.

Are you issuing a COMMIT WORK in the asych task 'M3'?

BR,

Suhas

Re: Commit for Asynchronous Function Module

Former Member — Wed, 21 Sep 2011 08:35:36 GMT

Yes I am using commit work inside task M3 also.

Regards

Arshad

Re: Commit for Asynchronous Function Module

Former Member — Wed, 21 Sep 2011 09:07:27 GMT

Hi,

As said by Suhas, bundle into one LUW. then it will work..

Instead of upadating tables directly in FM Z2 , and Z3 , recieve the results back to main internal session and use UPDATE Function module to update the tables..

RECEIVE RESULTS FROM FUNCTION z2 
                parameter_list 

RECEIVE RESULTS FROM FUNCTION z3 
                parameter_list

Regards,

Ravi.

Re: Commit for Asynchronous Function Module

Former Member — Fri, 23 Sep 2011 13:53:40 GMT

Halo Arseni,

Thanks for your reply . I tried the same but still the table M3 is not updated . While going in debugging its still updated . Dont know what really is the issue here ?

Regards

Arshad

Re: Commit for Asynchronous Function Module

Former Member — Fri, 23 Sep 2011 13:55:56 GMT

Halo rshankar,

I cant change the existing logic of the program . So the table updates takes place in method m2 and m3 . and I need to put the same inside Async FM ZM2 and ZM3

Regards

Arshad