RE: Record Lock Problem

["John Taylor" <lists@xxxxxxxxxxxxxxxxxxxx>]

Mark:

> No. It didn't recognize the lock. The problem was that Mapics
> was releasing the lock within their maintenance program and 
> locking an ancillary file to lock the record. Then rechaining 
> to the Itembl record before updating. Not sure why but that's 
> what they were doing.


I don't know Mapics, but I'm familiar with one use of this technique.
It's a co-operative locking scheme used when you need to process a
transaction exclusively, yet cannot maintain a lock on the "master"
record for the duration of the transaction. A simple example might be
the posting run of a sales order:

chain to order header allocation (ancilliary) record with lock
if locked (no other job should be using this order)
   chain to order header master record without lock

   if order not posted to inventory yet
          chain to order header master record with lock
          execute routine to post data to inventory
          set inventory post completed flag (in header rec) to 'Y'
          update order header master record
   endif

   if order not posted to sales analysis yet
        chain to order header master record with lock
        execute routine to post data to inventory
        set inventory post completed flag (in header rec) to 'Y'
        update order header master record
   endif
   
   release order header allocation record
endif

Updating the master record after each step in the posting run allowed
the process to be rolled back and/or restarted if necessary.
Unfortunately, that also made locking the master record useless as an
allocation strategy.

Nowadays you'd use commitment control if possible for this type of
thing, but that wasn't always available.

Regards,

John Taylor