DSXchange

Hello Jorge and welcome to DSXChange. It would seem that you havne't read the FAQ as requested, and thus have missed that most of the experienced posters get very bothered when people request or demand "ASAP" or "IMMEDIATE" help to their "URGENT" problems. I, for one, put any urgent requests at the ...

If you have RT_LOGnnnn Type 30 hashed files that are not directories containing exactly 3 files - DATA.30, OVER.30 and .Type30 they are corrupt. There are various ways to recover; fortunately log files can be cleaned up and removed without affecting other things. The simplest approach is to export t...

No, the 'key' attribute has no effect in DataSets.

Why did you post in the PX forum, then?

Hint - go to the manager, create a new routine and enter the text Ans = Trim(oconv(date(), 'DMD[2,2]')," ","A") and compile then run it. You'll find out what that statement actually does.

That's an easy one - It has failed.

Without supplying more information there is no additional analysis possible.

This is like me asking you what is wrong with my motorcycle because I have a blinking light in the console.

Are you sure you are writing to a sequential file stage - the posted log shows a peek stage.

Is it reproduceable now? Do other jobs use that hashed file as well and could be affecting the contents?

Also, do you re-create the hashed file each run or append to it? Is the error visible on each run or just occasionally? Do other jobs use the same hashed file? If you turn off buffering on the write does the error remain?

Do you have a SELECT on your hashed file read or any warnings in the log file?

One of the conditions here is that the finished application be ready for distributed processing, so that precludes BASIC stages. I was hoping that I'd missed some glaringly obvious solution. I've now got an ugly amount of stage variables doing the comparisons and will probably leave that instead of ...

No, not with the large amount of data going through per day; this job needs to remain a PX one and putting basic transform stages in isn't an option in this case.

How many rows went "into" the OCI stage - perhaps you are getting insert errors due to constraints. Add a reject link in the job to ensure that this isn't causing your dropped records.

You can tell the stage to ignore the first line (state that it contains the column names). If the footer has a different format you can split the logic in a transform stage and use the @INROWNUM stage variable to determine if the record count is correct.

I have a job where I have about 20 columns containing timestamp values. I would like to output on column which contains the lowest value from those 20 columns. The only function I could find is the MIN() function, which operates on numbers. If I put the result of a derivation "Min(Min(Min...))) into...

Is this happening on the Windows client or the UNIX server? Where are you seeing the message? This is to the loopback IP address, if on the windows client do you have TCP/IP installed (i.e. try a "ping localhost" or "ping 127.0.0.1")