Application has a table where all service calls are logged. App sends up ad hoc insert statement to insert data. Different spids doing the same insert are deadlocking on different indexes in the table. The fragmentation on the indexes looks good. Any suggestion on how to prevent deadlocks in this type of situation?
The database is vendor schema but the application changes are possible.
asked Sep 10 '10 at 11:30 AM in Default
Once had a very similar issue - with a very high concurrency, high throughput session data table - resolved it with covering indexes, as the updates where deadlocking with the clustered index seek and bookmark (or key) lookups against the non-clustered indexes for the reads.
The covering index meant that the read only needed to use (shared lock) one index.