DSXchange

Have you tried adding CHAR(10) to the end of your string?

Then add a reject link to your Oracle stage, make that go through a transform stage whose output is set to abort if more than 0 rows go through the reject link. The abort rolls back the incomplete transaction.

I assume then that you are using the upsert option rather than the load option. In that case, if you don't specify either the APT_ORAUPSERT_COMMIT_ROW_INTERVAL or the APT_ORAUPSERT_COMMIT_TIME_INTERVAL then the whole job run will be one transaction.

if the free() command fails then you might have corrupted your memory pointers - perhaps by having run out of same. If you monitor physical and virtual memory usage during the job run, do you see a large amount of memory used or constant growth until out of memory?

It does depend upon which file system the AIX administrators have created. Is it JFS2, JFS or another? In the past we've worked around this limitation by RESIZEing all RT_STATUSnnn files to static type 2 hashed files. Since Type 30 (Dynamic) files consist of a subdirectory with 2 visible and a hidde...

I don't have MQ installed at this client, but I think that this is exactly what setting segmentation does.

What stage are you using to write to Oracle, the answer depends upon how you are doing this. Also, is this a Parallel or Server job (I am assuming PX due to the environment variable, but since you posted in 'General' we cannot tell)

I knew I posted the wrong APT variable. Colin should have known that I really meant APT_ORACLE_PRESERVE_BLANKS when I wrote APT_STRING_PADCHAR

Since thet source for ICONV and OCONV are in c, it would be much easier for IBM to adapt existing code than to rewrite the complete functionality. Albeit the time and date handling (with full NLS support) would be a great feature to add to PX and is currently sorely lacking.

What database are you loading with which method and do you NLS turned on. What is the character set of the database table and, if it is oracle, what are your environment NLS settings for it?

If you replace your DB2 output stage with a dataset, does your error persist? Does your DB2 stage have a reject link (because the score indicates 2 operators)?

The optimal array size is so that one fetch fits inside one block (which I think is a default of 2048, but that might be version/platform/install dependant). As Craig said, 1 is the slowest but 100 might be even slower, so testing various values will be your best bet to tuning this stage using the a...

Windows DataStage servers come with the MKS Toolkit, which also includes the 'sed' command.

If the after-job script returns an error code then the job's status is set to aborted.

The Oracle VarChar2 strips trailing spaces, if I recall correctly. I think an insert of "Hello World " into a Varchar2 column results in an automatic rtrim. This might be at the heart of your problems.