DSXchange

Execute "ls -f /my/path/to/directory". Use the output, which is comma separated, as a list to parse within a loop, extracting one file name per iteration. This filename is passed to a DataStage job as a parameter, but is also used as the instance name for that job call.

Don't add and set it at the project level, since by raising this value the working memory of all stages is increased (significantly) and performance is affected (significantly). Only set it for the job where you would need bigger blocks.

I am assuming that each file name is distinct, so if an instance exists then the job is already running or has completed.

Can you do a View Data on the source from the Designer?

Which confuses me, since the OP stated that it was working before the server jobs were activated and installed OCI9. What stage was it using prior to that?

Replace your target with a sequential file stage (in a copy of your job). Does the error still occur? This way we can narrow it down to the source stage - DB2 or SQL Server. Can you do a "View Data" on the source?

I would write a sequence that gets the directory contents using an execute command stage inside an outer loop. It would then loop through each file found and check to see if there is already an instance of the multiinstance job running for that filename (I'd use the filename as the instance name). I...

You can raise the size of a block read from sequential files by increasing the environment variables $APT_DEFAULT_TRANSPORT_BLOCK_SIZE and $APT_MAX_DELIMITED_READ_SIZE. Note you will need to raise this value to 1 more than the next higher power-of-two to the longest record length; i.e. if your maxim...

Thanks for that piece of information regarding the PX-only install. I had no idea that it was possible to do this, I had always understood that buying a PX license implicitly gave one a Server license. The underlying engine is always installed and just about the only way to stop one from designing s...

What datatype did you specify in Oracle? Could it be a long type or blob? Also, use "VarChar2(1000 char)"

Hmmm... that object should be a 32 bit one, in my opinion. I am working on 8.0.1 so can't answer this problem, which looks like a 8.1 specific one. On my system I only have "liborchbuildopaix3.so" in that subdir, no "liborchbuildopaix64.so"

Varchar2 is the appropriate type for string values up to 4000 characters. I used "32" as an example - you might need a longer string - what is your input string maximum length?

No, a routine just returns one value. But you can always code the routine so that it returns one string, but with separators between discrete values that you can parse out in your job.

For NLS applications you should always make sure that, in Oracle, you declare strings using VarChar2(32 Char) so that 32 characters can be used. If you use the default, then just 32 bytes would be reserved and you couldn't store 32 multibyte characters into that strnig.

The type and level of documentation for DataStage is entirely dependant upon what the client or user wishes. There a capabilities at all levels to add documentation to jobs - in fact there are far too many places where one can add information and using all of them would be very tedious. At Version 7...