DSXchange

Try changing the key value that you are using. For example, in job "A" use KeyMgtGetNextKey("A") and in job B use KeyMgtGetNextKey("B"). Note that the sequences in SDK.SEQUENCES are never reset (unless you do so manually or in separate code). Could it be that you'd already processed 1000 rows using ...

Beware, however, that lookups over a network are abominably slow and should be avoided. Vincent's model is good if the Oracle instance is on the same box as the DataStage server, but slow if not.

This question really belongs on the ETL thread, as it's not a data quality issue. You posted this question there a couple of days ago, and got a good answer there. Why did you post here? The simple answer - assuming you're constructing a server job - is that you import the stored procedure definitio...

IF your DataStage is running on a Windows platform then you can create functions and deliver them via an ActiveX control. Import these into DataStage (in Manager, Import External Function Definitions) then they're ready to use. Note, however, that performance is not all that good. What's wrong with ...

The message is telling you that you're generating SQL in a UV stage but that the UniVerse server is having problems opening a UV table called "SnapshotDate". Check that you have the spelling and casing of this name correct. For example, is it actually "SnapShotDate"? You might also check the operati...

Use the DataStage security mechanism from your Administrator client. You can make one UNIX group developers (all access) and another group operators (run only, limited access to log files). There is a check box for giving operators full access to the log files. Use Version Control to mark jobs read-...

I never use DSLogFatal, or Abort, or the DS.ERRFATAL key with DSAttachJob() in DataStage jobs or in routines called from them. This way I always retain control. I can detect what happed from log entries (for example using DSGetNewestLogId function) or from the controlled job's status (for example us...

If you check execution status of a job from that job's after-job subroutine, the answer will always be "running".

Ray Wurlod
Education and Consulting Services
ABN 57 092 448 518

Beyond that, if you feel like modifying the code generated by compiling your job sequence, there is a function called DSSetJobLimit that you can use prior to executing the job (but after it is attached). You can find this function in on-line help from your Designer. Ray Wurlod Education and Consulti...

Check that there's an entry in the uvodbc.config file for the data source. This entry must have DBMSTYPE = ODBC. Check that you've properly configured the data source in ODBC.INI and that that file is visible as .ODBC.INI. Test (using the utilities provided with the driver) that the driver manager i...

The difficulty is knowing exactly which processes to kill. The job runs as one process, and active stages (such as Transformer stages) run as child processes. Any use of DSExecute, ExecSH, ExecTCL and so on may also start child processes. You can, of course, determine the job's pid in an after-job s...

Another solution is to maintain a list of keys that you have seen, and search this list. For a sufficiently small list (up to, say, a few thousand), a dynamic array in a variable declared to be in COMMON would suffice. For larger lists, a more efficient search strategy would be needed, such as a bin...

Be aware that records in hashed files are not date-stamped. Therefore, unless one of the fields within each data record contains a date-stamp of some kind, there is no mechanism for identifying the records that you might want to delete. For example, the DataStage job logs are hashed files but, in th...

That is really odd. If you examine the code of KeyMgtGetNextValue routine, you will not find any 1000 offset. The two different argument values will result in two different records existing in a hashed file called SDK.SEQUENCES. Could it be that "someone" has modified KeyMgtGetNextValue on your site...

You can archive the job logs to text file before compiling. Techniques for doing this have been published - check the archives here, search all forums (so that you pick up the Oliver list). Essentially the algorithm is: (1) determine name of log file (RT_LOGnn, where nn is job number) (2) open log f...