Dataset read - Unknown Error Reading Data

[csri]

Hi,

I created a job which reads from Oracle and writes to a dataset and on the dataset I have set the option to sort the data. Job finished successfully and dataset got created.

Problem: There are six partions of the dataset with some of the partitions having 0 rows. When I try to view the dataset with all records I can see the records. But, if I try to view data from a particular partition (0 rows or with rows existing in the partition) then I am receiving error "Unknown error reading data".

Did anyone came across this or can someone shed some light on this please.

Thanks.

[Nagaraj]

How to view data from a particular partition? i dont know how we do this!

[ArndW]

I don't know what is causing your error, but did you use a hash partitioning algorithm (which would explain the empty partitions). Perhaps the view data command just doesn't like empty partitions (just like nature abhors a vacuum). Also, have you tried using the command-line "orchadmin" command to display the dataset?

[csri]

did you use a hash partitioning algorithm (which would explain the empty partitions).

ArndW, I did do hash partitioning.

"Perhaps the view just doesn't like empty partitions ".

Do you think even view on non empty partitions also doesn't work?

How can I view the data then? Is there a way?

Thanks for your reply.

[csri]

Nagaraj, to view datasets and partions from designer under tools and data set management.

[ArndW]

I would recommend trying it from the command line with "orchadmin", if that brings an error then it needs to be reported to IBM as a problem.

[Nagaraj]

Thanks csri for your suggestion.

[csri]

ArndW,

I am looking into reading the dataset using orchadmin. I never used it. So, I had set some env variables and libraries before the command started working. I still got an error. I gave the same of the file with .txt extension and it may be incorrect. After looking at the dataset file name defination, I have change the name to .ds extension and re-run the job. Hope it will work. I will update once job finishes.

Thanks.

[csri]

ArndW,

I am still getting error from orchadmin for partition read and works fine for full read.

Error:

$ orchadmin dump -n 99 -part 0 /data/dstage1/project1/dataset.ds
##I TFCN 000001 14:57:25(000) <main_program>
Ascential DataStage(tm) Enterprise Edition 7.5.1A
Copyright (c) 2004, 1997-2004 Ascential Software Corporation.
All Rights Reserved


##I TFSC 000001 14:57:25(001) <main_program> APT configuration file: /tmp/aptoa37672107f704
##E TFPM 000040 14:57:27(000) <APT_PeekOperator,1> Operator terminated abnormally: received signal SIGSEGV
##E TFPM 000338 14:57:28(000) <main_program> Unexpected exit status 1
##E TFSR 000011 14:57:38(000) <main_program> Step execution finished with status = FAILED.
##I TCOA 000049 14:57:38(001) <main_program> The dump FAILED for /data/dstage1/project1/dataset.ds.

Any idea.

[ArndW]

If you change the partitioning to "round robin" to make sure all paritions have data, does the error still persist?

[ray.wurlod]

Moderator: please move to parallel forum

[csri]

ArndW, Thanks for your reply. If I change the partition type to round robin this error (reading individual partitions) does not occur.

Any idea why only with hash partition. I am running the job with hash partition and not sort to see what happens.

Thanks.

[csri]

ArndW, Thanks for your reply. If I change the partition type to round robin this error (reading individual partitions) does not occur.

Any idea why only with hash partition. I am running the job with hash partition and not sort to see what happens.

Thanks.

[ArndW]

Round robin "cheats", since it guarantees that all partitions have data (assuming you have records => number of partitions); but it doesn't solve the cause of the problem. I think that it would be a good idea to involve your support provider at this point in time, since you have a reproduceable simple test case on your system.

[csri]

Thanks ArndW.