LOGMNR_MAX_PERSISTENT_SESSIONS

LOGMNR_MAX_PERSISTENT_SESSIONS enables you to specify the maximum number of persistent LogMiner mining sessions (which are LogMiner sessions that are backed up on disk) that are concurrently active when all sessions are mining redo logs generated by standalone instances. This pre-allocates 2*LOGMNR_MAX_PERSISTENT_SESSIONS MB of contiguous memory in the SGA for use by LogMiner.

LogMiner requires 2 MB of contiguous memory for mining the redo log stream generated by a non-cluster database instance. To mine redo logs generated by an n-instance cluster database system, 2 MB of contiguous memory is required for each redo thread (or, a total of 2*n). Because persistent LogMiner sessions stage all data in the SGA, read buffers must be allocated in the SGA. The required contiguous memory is allocated at instance startup.

LOGMNR_MAX_PERSISTENT_SESSIONS is normalized for redo logs generated by non-cluster database instances. If you wish to mine a redo log stream generated by a 3-node cluster database instance, you should set this parameter to 3 (because one persistent session mining redo log generated in a 3-node cluster database instance is equivalent to 3 persistent sessions, each with mining redo logs generated by a non-cluster database instance). The memory remains exclusively allocated for LogMiner persistent sessions. This parameter does not affect transient LogMiner sessions that are not backed up on disk and are staged entirely in the Process Global Area (PGA).