Spend some more time reviewing your system events and more time on the joblog of the long running job. Ensure your message logging setting on the job is:
Message logging:
Level . . . . . . . . . . . . 4
Severity . . . . . . . . . . . 00
Text . . . . . . . . . . . . . *SECLVL
Log CL program commands . . . . *YES

I checked the following but still no success –

1. Changes the Run priority to max in the subsystem
2. Changes the time slice
3. checked the job log… couldnt fine anything..
4. Tried running the job in QINTER. As it is a nightly job.. there were no other jobs in the subsystem and gave it a run priority of 18.
5. There are no SQL or OPNQRYF used.
6. No change in the logical files used.
7. It still runas at its usual time.. sharing the memory with same jobs.. it used to from quite a while now.

And just to let you know it is a SYNON RPG program called by a driver CL program.

One of the file gets cleared first based on some RST parameters. and then the job builds the file from another file.

Please help.. if you have any other clue on how to get this done…

Did you had look at the priorities of the Jobs that are parallally running Batch.
How does this Jon work? I mean, will it updates Database Transactions or Generates Report?
I suggest, Compare the Job logs of the Current Job and the previous Job which used to take 20 mins.
If you have already compared let us know the results, that maybe helpful for tracking issue.

Pradeep.

It’s almost certain that something changed. That is, batch jobs that run for 20 minutes usually are processing significant amounts of data. If nothing changes in or about the data, there’s usually no point in running the job again — you’ll just get the same results if you run over unchanged data.

So, elements such as number of records or changes in authorities or just about anything may have effects.

One trouble is that if there is no benchmark to compare against, it can be difficult locating a bottleneck. Does the job do any kind of status logging as it runs? If it logged periodic messages (e.g., every 10000 records or every new job step), it might be possible to isolate where to look.

Tom

Its only this job that’s not getting CPU time..