DSXchange

Don't know, you didn't post the SQL so we can't tell. If your reference lookup was

then the issue might be that UniqKey is padded, or :1 contains spaces. If that wasn't your SQL, then please share.

Follow these steps, you must be root or dsadm: cd `cat /.dshome` . ./dsenv bin/uvsh From the command prompt, type in: > LOGTO UV Type in > LIST.READU Look at the USERNO column and get the number associated to your locked job. Type in > UNLOCK USER yournumber ALL Your job should be released. This ALW...

Hmmm, oraoci8.so (Oracle 8i) used with Oracle 9.2 client. Methinks your version of Oracle client was changed on your server.

PX has no update-able reference data set capability. For PX, you're going to have to either write a custom buildop or use the obscure Make Vector stage to do what you need.

Verify no trailing spaces exist using TRIM and make sure case sensitivity is handled. Just like SQL, data must be exact matches.

Finding no rows for update will not generate any error message or return code. Therefore, you must separate inserts from updates during transformation when you have a chance to act on that result, prior to loading, when you have less functionality.

You're opening a can of worms. Just FTP the file local and then use OPENPATH. NFS accessed hash files are not a good idea.

A purposeful reject link will generate a yellow warning message in the job log if the row count on the reject link is non-zero. Other than that, warning messages on Server jobs tell you your metadata is wrong, or your job has spewed information into the phantom capture file indicating non-numeric va...

Use reference lookups to determine the criteria for update and pass the appropriate row to the target OCI stage.

You could consider in the initialization of a stage variable using an ExecSH/ExecDOS call to a database command line program like sqlplus to issue SELECT MAX(... SQL statement on your table and return the answer as the initialization value. Increment your stage variable from there with each use.

The DS APIs are documented in your manuals, consider DSGetLogSummary and such. The command line dsjob is also well documented. How about you experiment with both? dsjob is really is to use, but the APIs give you more flexibility but require more programming. dsjob is easily used with a ksh or perl s...

Export from 3.0, import to 7.5. Then start making tons of changes to fix: link ordering, hashed file caching, row buffering, function argument modification, etc.

bmadhav wrote:Just to confirm, i need to read and write into the same hash file, correct?

Yes, which means you need to have this file created in advance of referencing it.

If you have a fixed number of reference columns, then switch your hashed reference file to have the same number of columns. The derivation for each column is the same except for the offset value. Something like: If inlink.YourAttributeIndex = 1 Then inlink.YourDataValue Else reflink.DataValue. For r...

You need to give a lot more information:
1. What's your commit rate?
2. Are there triggers on the tables?
3. Indexes on the tables?
4. What's the SQL like?
5. Describe the job design
6. How many inserts, how many updates?