Query about IMS extended checkpoint use in program

sivasaras · New User Joined: 29 Sep 2007 Posts: 93 Location: chenna/i-

Hi,

I have a db2 batch program which will link to IMS program.

Today we received nearly 840k records, job went late and completed 2 hrs late.

normally job ran within 20 to 30 minutes daily around 3 to 3.30.

when i saw it takes checkpoint correctly when it starts 3.05 am, then it didnt take for sometime, then 3.68 am it takes then stopped and then from 5.08 am it takes and completed .

Please guide me what i need to change in my program to make my program works fine so that next time it wont happen.

Rohit Umarjikar · Posted: Tue Sep 09, 2014 4:28 am

This could have happened because before two commits to DB2 whatever rows fetched are in process of IMS DB and hence you could not see the check point and during this window it could have been busy working with IMS DB (with ??? reason it took longer time ). So you basically needs to investigate what was happening between 3.05am till 3.68 am and then till 5.08 am? Check the IMS logs .

Sometimes it is scary like a Paranormal Activity...

Gary Jacek · Posted: Tue Sep 09, 2014 5:01 am

So, a DB2 batch program which links to IMS.

Do you mean a BMP batch job executing PGM=DFSRRC00,PARM='BMP,....' ?

You say you "received nearly 840k records".
Do you mean that todays input file contained 840k records?
Is this an unusually high number of input records?
Naturally, more input records would take longer to process.

Or did your program produce 840k output records?
Were they dump records?
Report records?
Something else?

Checkpoint frequency is extremely variable and certainly depends upon how you coded your program.
Did you see any messages indicating that your job was waiting for tape mounts?
Do you checkpoint based upon input record count? Database roots processed? Clock time?

Have you consulted your IMS sysprog to determine if your job was in lock contention with some other job? Do your IMS SLDS reveal an unusually high number of DLI calls from your job during the time there were no checkpoints? How does Workload Manager treat your job...is it getting enough service?

If you have some expectation that anyone here could help you, a complete listing of the job log, JCL and sysmsgs plus some sample output and your program source code would help a great deal.

As it stands now, there are very few clues at the crime scene.

Bill Woodger · Posted: Tue Sep 09, 2014 12:58 pm

3.68am? Well, there's your error, right there. Clock's broken.

sivasaras · New User Joined: 29 Sep 2007 Posts: 93 Location: chenna/i-

yes i am using DFSRRC00, PARM=BMP, yes my job output program produces 840k records

DBA told to change checkpoint logic

but stil i am unclear on that

Gary Jacek · Posted: Tue Sep 09, 2014 11:40 pm

Sadly, with the information provided, we cannot help you.