Hi,

We're migrating to Enterprise and and experiencing seemingly random (but extremely frequent)  0C4/AKEA and 0C7/AKEA abends in CICS (only under Load, like in PROD where we already are now!). Almost always the transactions work the second time around, and they work in all Test regions.

In a previous post http://www.tek-tips.com/viewthread.cfm?qid=915037  someone named KIMMIEH explained exactly the same problem in some detail. The reply there doesn't seem to be of much help. I am sure Kimmieh must have solved the problem by now. Any ideas?

Or is there anyway that I can send a Private Message to Kimmieh via this forum? It doesn't seem that way.

Here is a snippet of a typical abend:
An abend (code 0C4/AKEA) has occurred at offset X'00004CC6' in program CRT190.
 029                                                                           
CICS symptom string for message DFHSR0001 is                                   
LS/640 MS/DFHSR0001 RIDS/DFHSRP PTFS/UK05645                                   
 RIDS/CRT190 ADRS/00004CC6                                                     
  A SYSTEM DUMP FOR DUMPCODE: SR0001  , WAS SUPPRESSED BY THE DUMP TABLE OPTION



Regards
Leon

I forwarded the text to Kim Hicks to see if she recalls anything.   It was a long time ago and she is no longer actively programming.
 

Frank Clarke
--America's source for adverse opinions since 1943.

Hi All,

I want to post the answer just in case someone has same problem in future (we spent 3.5 working days crashing all the time)! I don't understand it exactly, but it goes something like this:

What changed was:
Enterprise compiled programs replacing OS compiled code.
More of the remaining static mods were changed to CICS dynamic programs.

We first found that it was usually Controlled storage that got damaged. This storage probably gets stored somewhere else now. Spent some time making sure we free all storage etc, thought we had memory leaks somehow.

The problem:
Most of the programs are compiled as AMODE=31 RMODE=ANY.
This was the same for the old code, with only a few more are 31 now.
The Transactions were defined in CICS as 24.  For some reason this never made a difference, now it does!
The moment we bounced CICS with changed transaction definitions to be 31 where the program is 31, all our strange random crashes disappeared.
The theory is that Enterprise memory handling is different somehow, and that calling 31 Amode programs under a 24 transaction will mess up addressability somewhere along the line and eventually some sub program is addressing space below the line that was not allocated to it.

I hope this can someday help some other guy who stumbles accross this post in desperation

Regards
Leon

Have you enabled SUBSCRIPTRANGE, STRINGRANGE, AND STRINGSIZE
conditions?
Also for good measure FIXEDOVERFLOW and SIZE?

You're six months late...
 

Frank Clarke
--America's source for adverse opinions since 1943.