Problem with Multiple instance job

[yaminids]

Hello friends,

We have a multiple instance job which is triggered 3 times, in sequence, by 3 different Sequencers.

JOB A
SEQUENCE1
SEQUENCE2
SEQUENCE3

The problem is once in a while the control of the job is not transfered to the next SEQUENCER.

For example, on a given day
JOB A is triggered by SEQUENCE1 and after completion it (JOB A) is triggered by SEQUENCE2 and then by SEQUENCE3

The problem is sometimes JOB A which is triggered by SEQUENCE1 shows completed in DataStage Designer but the Director shows as it is still running.
Meanwhile SEQUENCE2 waits for ever to trigger the job

Any ideas what's causing this?

Thanks a lot in advance
Yamini

[ray.wurlod]

Look in the log for JOB A. Look for processes associated with JOB A (in Cleanup Resources). Has JOB A aborted without being able to update its status and log records? (If so, you won't find any process associated with it, in which case clear its status file and look in &PH& directory for any error report.)

Does each sequence (note: "sequence", not "sequencer") run the job with the same invocation ID?

[kcbland]

If you do find processes executing, look at before/after job/transformer routine calls. If the transformer is executing an after routine, the links may be green but the call could still be open. Also, look at any large log file purging that could be happening. Your &PH& directory could be full and you're waiting on the job to wrap up. Make sure that any before/after routines aren't trying to lock some resource and hanging on that.

We've seen/heard/read a lot of ways valid jobs "hang".

[yaminids]

Thanks a lot for your suggestions.
I am waiting for the scenario to reoccur so as to get more information

Yamini

[yaminids]

Hi Ray,

Can you please tell me how I can access &PH& directory?
I tried to enter the directory and got the following
[dstage@DSEngine]$ cd &PH&
[3] 28082
[4] 28083
-bash: PH: command not found
[3] Done cd

Thanks
Yamini

[ray.wurlod]

"&" is a meaningful character to the shell. You must escape or quote it.

Code: Select all

cd \&PH\&

or

Code: Select all

cd '&PH&'

[yaminids]

Hello there,

The problem occured again
The job does not have any before/after sub routines. Also, &PH& directory is not empty and I found only two files in the directory (.Type1 and .uvnlsmap)

Any idea of whats causing the problem

Thanks a bunch in advance
Yamini

[ray.wurlod]

It's extremely unusual that &PH& is empty but for these two files. Do not delete .uvnolsmap (since it specifies the NLS map to be used for file names in that directory). If &PH& is otherwise empty you can delete .Type1, which is an indicator to DataStage how to treat this directory in BASIC.

If &PH& is empty that suggests that CLEAR.FILE has been executed against it. Ordinarily every job run will leave a file in &PH&. The name of that file is DSD.RUN_xxxxx_yyyyy where xxxxx and yyyyy are the time and date (in internal format) that the job was run.

Can you check the individual job logs to determine whether they actually ran, and the invocation IDs? Can you also post the second last entry from the job sequence log (the "summary" event)?

[yaminids]

Hello Ray,

The following is the second last entry in the log of the sequence
Job control process (pid 20162) has failed

I think this occured after I killed the process as it was not doing anything

Also, can you please tell me how I can find the InvocationID of the job?

Thanks
Yamini

[ray.wurlod]

InvocationID should be in the "job started" entry in the log.

Please advise, when reporting errors, whether kill has been used. It opens an entire extra set of possibilities.

Even better, don't use kill on DataStage processes.