MVSFORUMS.com

star99 · Beginner Joined: 09 Mar 2004 Posts: 4 Topics: 2

is it his/her responsibility ?

kolusu · Posted: Tue Mar 09, 2004 3:13 pm Post subject:

Star99,

I can say to some extent it is indeed the programmer's responsibility to avoid dead locks. First take a look as to why the dead lock has occured in the first place. once you know the reason/culprit , then it is a simple task of how to handle the deadlock. If the transaction is not that critical , code a delay in the pgm and try the update after the delay. If this task is indeed critical , then you need to check with the other task and see if then can implement the delay logic. If you are dealing with DB2 the check the isolation level defined.

Hope this helps...

Cheers

Kolusu
_________________
Kolusu
www.linkedin.com/in/kolusu

star99 · Beginner Joined: 09 Mar 2004 Posts: 4 Topics: 2

sorry, but i wondered how can the programmer avoid dead locks before it happens ?

R.Nickel · Posted: Tue Mar 09, 2004 10:39 pm Post subject:

Deadlocks occur because 2 programs are updating 2 or more databases, if program PGM1 updates DB1 and DB2 in that order and program PGM2 updates DB2 then DB1 then you have a potential for a Deadlock. ensuring that this does not happen will ensure that you have no deadlocks and the subsequent deadlock ABEND
_________________
Rainer

Mike Chantrey · Posted: Mon Mar 15, 2004 8:32 am Post subject:

To expand/generalize the above comment, the general principle for avoiding deadlocks is:
"All programs which update more than one recoverable resource should update in the same order."

Why? If program A updates item X then Y, and program B updates Y then X, you can get a situation where A holds X and needs Y, and B holds Y but needs X. Neither can proceed - deadlock occurs.
This cannot occur if they both update X then Y.

Thje recoverable resources could be records on the same file/table or different files/tables. Applies to VSAM, DB2 whatever.