DSXchange

You could experiment with the NLS Code page at the job and stage levels. See if UTF-16 will help.

Mike

RCP isn't going to do any type conversion. If your source is a ustring then it will propagate as a ustring. If it is a string then it will propagate as a string. If you need to convert between ustring and string, you can use a modify stage with a parameterized specification. $OSH_PRINT_SCHEMAS will ...

You've told the aggregator to expect sorted data. So have you partitioned and sorted the data by the keys that the aggregator is using? Have you set up your funnel stage to preserve the sort order if you have indeed sorted the data upstream?

Mike

Can't really say what your issue might be without actually looking over your shoulder. But in your example, the originally updated 16 byte surrogate key file should show 2004 2004 when viewed with the stage in view mode (if I remember that correctly). Running on 4 nodes, I would expect you to end up...

The Surrogate Key Generator stage results in a sequential operator, so the number of processing nodes shouldn't matter. Partitioning is not needed or wanted.

I still use my default parallel configuration, but all 3 stages in my generic utility job are set to run in sequential mode.

Mike

Then you did not successfully start with an empty state file. It will have a length of 0 when it is empty. After your update it should have a length of 16. Use an ls -l mystatefile to check file size. The state file is a binary file. It will have no record or field delimiters. Each record will be ma...

There are quite a few nuances to the surrogate key generator stage. It actually includes some interesting flexibility. You would have to spend some time experimenting to learn all of the nuances. I have a generic utility job that I use to sync a state file with the maximum key value for a table that...

Did you use round robin partitioning method?

Mike

I've gone with a 2 step approach for identifying version change candidates. Step 1 is to compare the modification timestamps using XMETA. These are the same modification timestamps that you would see with an advanced find using the Designer client. This does require read access to the XMETA reposito...

I haven't done an 11.5 install yet, but for all other versions that I have installed the required IBM java is part of the installation package. You just have to follow the installation instructions and make sure that there is no other java version to get in the way.

Mike

Nice job of narrowing down the cause.

It's good to know that a failed cast function in the ODBC Connector results in NULL.

You mentioned AS400, so is your underlying database DB2?

Mike

Unfortunately the report of schemas generally gets lost when the job aborts. 1) Verify that you have no column metadata defined in the ODBC connector. 2) Verify that the sql select statement doesn't have any function that might result in NULL. 3) Replace everything after the ODBC Connector with a co...

Unfortunately the job activity is lacking in flexibility when it comes to parameter sets. You are essentially stuck with 3 options: choose "As Predefined", choose a values file name from the drop-down list, set it to the controller's parameter set of the same name. The command line gives y...

As you see, bytes and characters have a variable ratio when you are dealing with UTF8. Characters will range in size from 1 to 4 bytes. Your 30 byte column in Oracle could potentially hold as few as 7 and as many as 30 complete characters. If you don't stick to character semantics, you run the risk ...

UTF8 and bounded varchars come with the sort of baggage that you're seeing. That's due to UTF8 having a variable number of bytes per character. You should switch to nvarchar (or the equivalent of adding Unicode extended property to your varchar). Also if you look closely at the output that you think...