- Article Type: General
- Product: Primo
- Product Version: 2
In 2.1.2 we have made some changes to the FRBR process in order to make it run faster.
Since we run the FRBR with 5 (depends on configuration) different threads to make the process finish quickly there are cases where 2 threads try to update simultaneously the same record and this causes a "deadlock" in the database.
We are aware of such cases and treat it in our code.
In case of a deadlock we run a "one by one insert" which means:
Instead of trying to update a bunch of records in one query, we update each record separately hoping this will minimize collisions between different threads when trying to update.
Our defect was that our "one by one insert" wasn't never called because of a coding defect.
So whenever a deadlock occurred, there was no treatment for these records and they were never updated.
A hot fix for this problem was distributed. This problem is fixed from 2.1.3 and onward.
- Article last edited: 10/8/2013