Using the ISPF Online Monitor in deadlock resolution
A deadlock is an error condition where processing cannot continue because of an
unresolved contention for a particular resource.
A deadlock is signaled when the deadlock detection cycle time has been reached. Ensure that
the resource timeout limit is at least twice that of the deadlock detection cycle so that
deadlocks are detected before a timeout occurs.
Deadlock detection
Start the data collector
to collect deadlock events.
Start exception event processing
from the Exception Processor panel to be notified when an exception
event occurs. See How to start exception processing.
Deadlock resolution
On detection of a deadlock,
use LOOK to display the participants in the deadlock.
If the 1 field is Datapage,
ensure that all applications (Holder and Waiter)
access the resource in the same order.
If the deadlock concerns
an index with high insert activity (the Locked resource Type field
is Index Page), set the SUBPAGES parameter for the
index to 1 or consider converting the index to type 2. If the problem
is not related to SUBPAGES, it could be that the commit frequency
for an update application is too low. The ratio of updates per commit
can be viewed in the Thread Detail panel.