I have a transactional replication. The 'distribution clean up:distribution' alway was killed by systme as a victicm while deadlock happen involve 'log reader' and 'distributuion agent '. So 'distribution clean up:distribution' never succeeded once. The 'distribution_database' became biger and biger. It badly impact the replication performance.
Of course, I can stop 'distribution agent' or 'log reader' and launch it manually. but is there a smart way to resolve it. Thanks.