Troubleshooting
Problem
During regular processing your database reports may show that the mpi_entlkiq_xx table (where xx is your entity type) is being locked excessively.
Cause
This is a perfectly normal process which occurs during regular entity management. The locks that you are seeing on the mpi_entlkiq_xx tables are used to insure that only one entity manager can retrieve queue records from the mpi_entique_xx table at a time. The lock lasts for a the duration of time required to retrieve and mark the mpi_entique_xx records which should be fairly quick.
NB: The fact that the lock on the mpi_entlkiq_xx is a table lock is irrelevant. The entlkiq table is always empty so the only type of lock that will work on this table is a table lock.
Resolving The Problem
Because a lock is placed on this table by an entity manager working to retrieve records from the relevant mpi_entique_xx table, excessive locking on mpi_entlkiq_xx could be due simply to the entity managers finishing their workloads too quickly and returning to acquire more queue entries for work. You may need to tweak the configuration parameters in the conf/com.initiate.server.entity.cfg file in order to optimize entity manager performance.
- The workUnit= parameter determines the number of queue items that the entity manager chooses to work on at a time. Try adjusting this gradually upward. Smaller numbers like 100 or below should typically be avoided.
- The maxQueueDepth= parameter represents the total amount of rows that your EM(s) will acquire. Adjust gradually upward.
- When the queue is full (maxQueueDepth has been reached), pollSeconds= configures the amount of "pause time" before EMs will look for new work. Don't set this too high or database performance may take a hit. Try 10-30 seconds.
Was this topic helpful?
Document Information
Modified date:
16 June 2018
UID
swg21692521