IBM Mainframe Forum

Hi

Back Ground:-
I am working in an application where the DB is group of KSDS VSAM files. All these files are get synchronized during update by maintaining a sync number in header record of all the VSAM file called SYNC record. During the start of this application when update start through batch job this sync number get updated for all the files.

Problem statement:-
When this job encounters any system abend (like S0C7 or S0C4 ABEND), this updated sync number is not available in the few of the files, which leads to database out of sync. However If I am forcefully canceling this update job after it get started, this updated sync number is available throughout all the files. If this update job is successfully complete then obviously the updated sync number is stored in all the files.
Note this is the very first record in VSAM file and in this application data get handled using some intermediate buffer. I could not find out the root point where this buffer is getting handled. In this application the I/O operation of VSAM handled through one assembler routine instead of COBOL READ/WRITE.

Anyone please give me any hints, why this record in few of the file is not getting updated only when system abend is happening.
Any suggestion will be great help!!!

Thanks in advance,
Satya

This sounds like an application issue and, as such, is probably something you have to review the code to figure out.

Are all the VSAM file updates for the sync number being done at the same time? Could you be getting the system abend while updating the sync records -- in which some of them may not get updated due to the abend? There are many possibilities, most of which require code review to determine if they can happen (or did happen).

Hi Robert,
Thanks for giving your suggestion. First of all I would like give more back ground here on this issue.

As part of this application enhancement, we would like to implement one CHECKPOINT / RESTART facility in this VSAM application. For this reason we are testing it by forcefully abending this application for a particular record after taking a check point. Here our check point counter is 50. So when we are forcefully abend a record with S0C7(by moving low-values to a PD field), at that particular moment at least 50 records has been updated along with the very first sync record.

Yes, This sync record update is very first thing happen in this update job and this update happens for all the files simultaneously. I am still reviewing my application code, however I could not figure where it went wrong.

Thansk,
Satya