Page 1 of 1

MQ and Server Shared container

Posted: Mon Aug 13, 2012 2:19 pm
by dodda
I have a job design like

MQ--Transformer--Server Shared Container(with Folder Stage in it).
I am reading around 4 messages from MQ at a time and creating each XML for each message and storing in a FOlder stage. each xml is of around 1.2 MB size. I increade the APT_DEFAULT_TRASPORT_BLOCK_SIZE to higher number as well.
I am getting the below error when i am reading the file from MQ and storing it in a FOlder on the server.

SCLoadXML,0: row is to big to fit in shared memory buffer

Can some one help me on this

Posted: Tue Aug 14, 2012 1:03 am
by ArndW
It has been 2 or more years since I did a DataStage with MQ project, but seem to recall having hit the same problem as you are seeing. What OS are you working on? We were on AIX and I believe we did something AIX-specific. There was also an interrelationship between the default and max block sizes in the DSParams as well.

Posted: Tue Aug 14, 2012 12:27 pm
by dsoupaty
Hi
I am using AIX 6.1. I also played aroung with
APT_BUFFER_MAXIMUM_MEMORY
APT_BUFFER_DISK_WRITE_INCREMENTAPT_BUFFER_FREE_RUN

environemntal vairables as well. All the APT variables i used i changed them at the job level.
but no use.

Can you tell me the relation ship you are talking about.

Thanks

Posted: Tue Aug 14, 2012 5:05 pm
by ArunaDas_Maharana
what is transformer used for? you are reading XML from MQ and creating xml in transformer?

Posted: Tue Aug 14, 2012 8:15 pm
by dsoupaty
I am using transformer for concatinating messageputime data element and message date data element from mq and generating a disint file name so that each message goes to the single distinct file in the folder stage.

Posted: Tue Aug 14, 2012 8:17 pm
by eostic
Skip the EE Job altogether. Real time --- do you "really" need the deep parallelism that you can accomplish with EE? Sometimes you do...but often Server will be just fine in a real time paradigm -- maybe even better. Unless this is quality Stage, or you can really justify the need for massive parallelism, drop EE as the host and use a Server Job all the way...the barrier to shipping large rows across the Container "boundary" go away --- and Server is better at handling huge, totally undefined varchar strings anyway.....

Ernie

Posted: Wed Aug 15, 2012 8:17 am
by dsoupaty
Thanks ernie. Will try that and let you know

Posted: Wed Aug 15, 2012 11:54 am
by ArunaDas_Maharana
what is the datatype you were storing xml and passing through transformer --> folder stage through shared container?