DSXchange

Read 1st file, and write into a sequential file(file_1), and in the output of the file_1 read the 2nd file, and append to file_1 , again in the output read 3rd file and append to File_1

Server job it is certainly possible and easy to implement. Parallel , possible, tricky. and probably not very useful. i don't have PX environment now so can't test. try this take the last records from 1st file as refrence input to the second file using lookup stage, so that the second file will wait...

Assuming you need to do some data validation of each file, and that's the reason you can't combined 3 files into 1 (using cat command etc.) .

you can have a single job with multi instance, to append into a single file, from the three(or more) instance of job.

are you using user defined Query?? can you post the query.

is it throughing any error or warning??

Recomended

NO

search is good option.

http://it.toolbox.com/blogs/dw-soa/ibm- ... data-19417

can you put some example of source1 and source2 data. and what is you are expecting in the output

In the source table the column type is decimal[28,0],
whereas in the OEE stage, the coulmn is defined as Integer.

not sure if it can be done in Parallel.
but in a server job, using link partition(hash partition on that column) will do.

for large volume of data to join, more number of nodes.

can you post the content of configuration your are using for this job?

check the Mload Log file for detail error.

the second job should also have four node to read the dataset..
so if you want to process the data in 2 node in second job, then make the other two node in "read only" pool.

may be the scratch area is not enough to hold the input data in join stage.

in the grid queue , all the compute node is busy.. command qstat will show how many jobs already in queue... so delete the jobs from grid queue using qdel .......

by disabling the grid, you are by passing the queue, and overloading the compute node.