Jobs Contending with each other .

infotech58 · Posted: Tue Dec 04, 2007 1:46 pm

Friends,

I would appreciate your inputs for the following.

In our batch cycle four jobs (read only) A,B,C,D run in parallel and are using same program, PSB, PROC; and some time they abend due to contention of resources with each other. Could someone suggest me some way to avoid this contention ??
Could it be done by using four different PSB's with same access.

Sandy Zimmer · Posted: Tue Dec 04, 2007 5:21 pm

Did you ask us about running the same PSB and job at the same time before this posting? Did we ask why?

infotech58 · Posted: Tue Dec 04, 2007 7:04 pm

yes. I had asked. But that was a different problem. That was about splitting a job into four and running in parallel. I changed the design of that program and reduced the run time instead of converting that into multiple jobs running in parallel.

These jobs are already running in production and getting down oftenly. Any comments on this ??

Sandy Zimmer · Posted: Tue Dec 04, 2007 7:12 pm

Do these jobs have the same objective? If so, why are you running them parallel? If not, why are you using the same PSB? Each program should and MUST have it's very own PSB. You are getting yourself locked into contention with yourself.

babai043 · New User Joined: 19 Dec 2007 Posts: 19 Location: chennai

If jobs are updating database, definatly jobs needs to run one by one,

Regards,
Srinivas Jala

Bitneuker · Posted: Thu Dec 20, 2007 2:53 am

Check the PROCOPT of the PSB. Also, if any of the programs uses a HOLD or a command code to prevent any other program to access a segment......

Robert Clubb · Posted: Sun Dec 23, 2007 6:29 am

Chances are contention is at the database level or in an external subsystem such as DB2. If that is the case, breaking them up into separate jobs and separate PSBs probably will not help.

You need to find out what is going on: If you have a monitor such as Omegamon you should see resource contention while they are running. If not, you can determine this by running the IMS log utility DFSERA10 using the DFSERA30 exit and specifying DEADLOCK against the type 67'FF' log records. An example is in the IMS SYSTEM Utilities manual.

If you don't see anything in after the log utility, try turning on the IMS Monitor.

External issues not only would include looking for such items as a deadlock or resource contention in DB2, you also need to look to see if this is a JES issue such as waiting on a common dataset with a DISP=OLD. Check the JES output for any data set messages. For DB2 or CICS, look in the logs or journals to see if there are any messages showing up.

v4vicnme · New User Joined: 25 Apr 2008 Posts: 1 Location: Mumbai

Can this not be solved by defining the same DBD more than once (based on the no of pgms using the PSB parallelly) ? If not, what is the reason behind defining PSBs that way ?

Atul Tamboli · Posted: Tue May 20, 2008 12:38 am

I don't think your jobs are contending due to same PSBs. They must be contending due to same database(s) being accessed. This could be IMS, DB2, etc.

We have 15 jobs running parrallely 16 times a day and we hardly run into contention issues. These jobs use same PSBs and program. The key is that the each job accesses different partition of master database. Also there is some option that I don't remember that can make the PSB wait/retry in case of contention. You may want to explore it.

Another thing that you might want to research is the checkpoint frequency. The lower the frequency, more records are locked.

What is the abend you are receiving?

Thanks.
Atul