Need guidelines from CICS expert to debug storage violation

Gopalakrishnan V · Posted: Thu Dec 20, 2012 5:56 pm

Hi,

I am getting storage violation in and from the few dump got the module name in freemain step. due to this the region became unstable and fluctuating. After removing the code the region is stable.But i can't able to debug why this is happening. The same code working fine in testing environment, only getting SV in production environment due to multiple transaction at same time.

I need some help from CICS expert those who know very well about this SV debugging. I read the manual which is sharing in this forum for the same topic, but they are using some VERBXIT command using IPCS tool. Here we doesn't have that tool.

We are using abendaid, I know its a compuware tool and we should not share the manual directly. So atleast i just want some guidelines like from which option or area we can get the address details exactly causing storage violation and need to confirm because of which modules.

Please help me on this. here i am facing some difficulty to get a CICS expert guidance.

dick scherrer · Posted: Thu Dec 20, 2012 8:01 pm

Hello,

If you are using abendaid and have a license to use it, the material is available for free from Compuware Support.

You could also open an issue with Compuware Support. . .

Robert Sample · Posted: Thu Dec 20, 2012 8:04 pm

Actually, I believe IPCS is available at your site -- you just don't have access to it.

Your best bet would be to talk to your site support group and get a CICS trace run to analyze. Storage violations can be difficult to debug since the perpetrator may be located a long way from where the storage violation appeared. Abend Aid may -- or may not -- be much use in debugging the problem.

enrico-sorichetti · Posted: Thu Dec 20, 2012 8:05 pm

the usual advice when dealing with storage violation ...
the abending program might be an innocent victim

Garry Carroll · Posted: Thu Dec 20, 2012 8:45 pm

I would agree with Enrico's comment.
Many years ago, I debugged a storage violation which caused other transactions to abend and made the CICS region unstable (sometimes locking-up, sometimes abending). This only happened when we hit production volumes. In that case, the problem was caused by a mismatch in COMMAREA sizes where one program was overwriting an area which was beyond the actual end of the allocated COMMAREA. All was OK until that overlay acess corrupted an adjoining TCA in memory causing the 'victim' TCA-owning transaction to abend.

Garry.

Gopalakrishnan V · Posted: Fri Dec 21, 2012 2:58 pm

Thanks everyone..

Hi Garry,
Here also the prob due to COMMAREA it seems..it having nearly 300k bytes. Can you please just elaborate how you resolved your SV issue.

enrico-sorichetti · Posted: Fri Dec 21, 2012 3:07 pm

Bill Woodger · Posted: Fri Dec 21, 2012 3:09 pm

And since COMMAREA has a maximum length of around 32k, what exactly are you talking about when you say it is 300k?

prino · Posted: Fri Dec 21, 2012 3:23 pm

At a former client the decision was made to compile all programs with SUBSCRIPTRANGE condition enabled, even for the production environment. Yes, there is some overhead, although in recent versions from PL/I this has apparently been reduced, but the overhead was considered well worth it.

Garry Carroll · Posted: Fri Dec 21, 2012 3:45 pm

In our case the diagnosis was in-depth and required IBM assistance. Essentially, knowing the corruption in the TCA, we modified trace so that, every time an EXEC CICS was executed in ANY task, we checked for the corruption in the TCA chains and abended CICS with a dump once the corruption was found. Since everything was OK at one trace point, we know which task had caused the corruption and could home in on the program location.

Not an easy task and with considerable CPU overhead - a method that should be avoided unless all else fails.

Garry.

Gopalakrishnan V · Posted: Fri Dec 21, 2012 4:54 pm

enrico-sorichetti · Posted: Fri Dec 21, 2012 5:16 pm

nobody around here cares about the size You are using ....

the important thing is that You should ensure that ALL the copy books are included in ALL the programs involved

in order not to have any length mismatches !!!

what is that is not clear in the above concept ?