The database service queue holds a list of the IDs of all the AI blocks that must be replicated to the Target database. If the queue is full, all incoming ID entries are discarded, so information about AI blocks that are yet to be replicated is lost.

To prevent this from happening, you should set the database service communication area size (-pica) to its maximum size (1000000). The system cost is very low and the penalty for having too small a value can be high.

We recommend that when you set up the Source and the Targets for replication, make sure that you include -pica 1000000 in your list of startup parameters.