active-to-active link problem

sendmk · Post by **sendmk** » Sun May 07, 2006 2:25 am

yes this is the template we should follow

(hi ray i have re-edited hash file information in my previous post for ur clarification, thank you)

chulett · Post by **chulett** » Sun May 07, 2006 6:48 am

sendmk wrote:yes this is the template we should follow

Huh? Do you mean you should have made it two jobs? Are going to? Something else?

I, for one, will no longer comment on the design as posted until a screenshot of the actual job canvas is made available. The ASCII art just isn't cutting the mustard, at least not for me.

ray.wurlod · Post by **ray.wurlod** » Sun May 07, 2006 2:29 pm

I still can't tell whether "hash1" is being populated by or being used to serve a reference link by either of the transformer stages to which it is linked. Arrowheads (use "V" or "^") are needed. Or post the image, as Craig asked.

sendmk · Post by **sendmk** » Sun May 07, 2006 11:23 pm

This is the image url

http://i77.photobucket.com/albums/j79/sendmk/dsjob.jpg

i hope this is helpful.

thank you

ray.wurlod · Post by **ray.wurlod** » Mon May 08, 2006 12:24 am

Even with Photoshop applying maximum sharpening, I can not read the labels on that job design. It is, nonetheless, clear that you have a circular design.

Code: Select all

       ----->  TransformerStage1  ----->
               ^               |
               |               |
               |               V
           HFStage1         HFStage2
               ^               |
               |               |
               |               V
       ----->  TransformerStage2  ----->

which will never work. Please post a larger image, that humans can read without scanning electron microscopes, if you want more help.

sendmk · Post by **sendmk** » Mon May 08, 2006 3:16 am

hi ray u have misread the design, see the picture, it is very clear i guess, there are two transformers in place of transformerstage2 in the actual diagram

thanks for ur time

kumar_s · Post by **kumar_s** » Mon May 08, 2006 3:49 am

BTW what is the use of transformer4 at dead end?

sendmk · Post by **sendmk** » Mon May 08, 2006 4:01 am

kumar_s wrote:BTW what is the use of transformer4 at dead end?

it writes to an odbc output

ok

kumar_s · Post by **kumar_s** » Mon May 08, 2006 4:20 am

Have you tried the job design to split into two?

balajisr · Post by **balajisr** » Mon May 08, 2006 4:34 am

kumar_s wrote:Have you tried the job design to split into two?

Kumar,

That will work.Even If i put a sequential file or IPC stage between transformer1 and transformer3 it will work.

This is the job template provided to us. Changing the template would be the last option. As everyone pointed out this is the only option i guess.

kumar_s · Post by **kumar_s** » Mon May 08, 2006 4:59 am

Sorry I could not view the image you attached, but may I know the use of Hash1 to transformer1?
Though the suspection is at hash2, are you able to see the progress of the rows through the performance staticts, are you able to identify the bottleneck?

balajisr · Post by **balajisr** » Mon May 08, 2006 5:20 am

kumar_s wrote:Sorry I could not view the image you attached, but may I know the use of Hash1 to transformer1?
Though the suspection is at hash2, are you able to see the progress of the rows through the performance staticts, are you able to identify the bottleneck?

The job reads data from oci or odbc and populates the data to hash1 through transformer1. Hash1 acts as reference link to transformer2.
Transformer2 populates data to hash2 which acts as reference link to transformer3.

The job just hangs and there are no rows are populated. Once I unchecked "Create File" option in hashed file stage, Hash file is not created in that instance.

Looking at the log I feel no active stage has started.

Log only has only three entries:

1.Starting job xxxxx
2.Environment variable settings:
3.Set NLS locale to xxxxx

That's it no more entries in the log.

kumar_s · Post by **kumar_s** » Mon May 08, 2006 5:56 am

Not sure whether Server jobs is also a victim of fork lock conditions. Also Iam not that experience in server jobs with the design of writing into the hash and reading from the same stage in the same job. Perhaps experts can comment.
But do you mean to say that by removing hash2, i.e., connecting

Code: Select all

transfomer2-->transformer4-->odbc2

as well as

Code: Select all

transformer1-->transformer3-->seq2

without any reference link and this works?
If this is the case then the hash1 logic shouldn 't be under doubt.

ArndW · Post by **ArndW** » Mon May 08, 2006 6:52 am

I am somewhat confused by these continued posts on this subject. On Friday I mentioned why it won't work, and Ray stated it again this morning.

chulett · Post by **chulett** » Mon May 08, 2006 7:03 am

balajisr wrote:This is the job template provided to us.

Whomever provided this to you as a job template should be severly thrashed across the bum.

ArndW wrote:I am somewhat confused by these continued posts on this subject.

Only somewhat?